ayout 1 - EuroSLA
.)%(*$'" ,#
(-' *$ +
(! *( ++ $%$,/ # (*/
University of Verona
University of Western Sydney
-*(+% &('("* )#+
Eurosla publishes a monographs series, available open access on the association’s website.
The series includes both single-author and edited volumes on any aspect of second language acquisition. Manuscripts are evaluated with a double blind peer review system to ensure that they meet the highest qualitative standards.
*$ +
Gabriele Pallotti,
University of Modena and Reggio Emilia
++$+, ',
Amanda Edmonds,
Université de Pau et des Pays de l’Adour
Kevin McManus,
University of York
Fabiana Rosi,
University of Modena and Reggio Emilia
© The Authors 2015
Published under the Creative Commons
“Attribution Non-Commercial No Derivatives 3.0” license
ISBN 978-1-329-42765-5
First published by Eurosla, 2015
Graphic design and layout: Pia ’t Lam
An online version of this volume can be downloaded from eurosla.org
We dedicate this volume to
Maria Agata Di Biase,
who would speak to anyone with
little recourse to a second language,
and to Noel Macainsh,
who would not speak in
his second languages unless
he could use a well phrased sentence.
$,(*$ % ( *
Cecilia Andorno, University of Pavia
Dalila Ayoun, University of Arizona
Camilla Bardel, Stockholm University
Alessandro Benati, University of Greenwich
Sandra Benazzo, Université Lille 3
Giuliano Bernini, University of Bergamo
Camilla Bettoni, University of Verona
Marina Chini, University of Pavia
Jean-Marc Dewaele, Birkbeck College, UCL
Anna Giacalone Ramat, University of Pavia
Roger Gilabert, University of Barcelona
Pedro Guijarro-Fuentes, Plymouth University
Gisela Håkansson, Lund University
Henriëtte Hendriks, University of Cambridge
Martin Howard, University College Cork
Gabriele Kasper, University of Hawai’i at Ma-noa
Judith Kormos, Lancaster University
Folkert Kuiken, University of Amsterdam
Maisa Martin, University of Jyväskylä
James Milton, Swansea University
John Norris, Georgetown University
Lourdes Ortega, Georgetown University
Simona Pekarek-Doehler, Université de Neuchâtel
Manfred Pienemann, University of Paderborn
Leah Roberts, University of York
Jason Rothman, University of Iowa
Michael Sharwood Smith, Heriot-Watt University Edinburgh
Nina Spada, University of Toronto
Richard Towell, University of Salford
Danijela Trenkic, University of York
Ada Valentini, University of Bergamo
Ineke Vedder, University of Amsterdam
Christiane von Stutterheim, Heidelberg University
Johannes Wagner, University of Southern Denmark
Table of contents
Camilla Bettoni and Bruno Di Biase
Part I
Introducing and developing Processability Theory
Camilla Bettoni and Bruno Di Biase
Processability Theory: theoretical bases and universal schedules
Camilla Bettoni and Bruno Di Biase
Part II
The developmental path across languages
Camilla Bettoni and Bruno Di Biase
The development of English as a second language
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
The development of Italian as a second language
Bruno Di Biase and Camilla Bettoni
The development of Japanese as a second language
Satomi Kawaguchi
Part III
Exploring new issues
Camilla Bettoni and Bruno Di Biase
Acquiring case marking in Russian as a second language:
an exploratory study on subject and object
Daniele Artoni and Marco Magnani
The development of case in a bilingual context: Serbian in Australia
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
Exploring the acquisition of differential object marking (DOM)
in Spanish as a second language
Bruno Di Biase and Barbara Hinger
The development of constituent questions in Italian as a second language
Camilla Bettoni and Giorgia Ginelli
Acquiring V2 in declarative sentences and constituent questions in
German as a second language
Louise Jansen and Bruno Di Biase
Exploring Processability Theory-based hypotheses in the second
language acquisition of a child with autism spectrum disorder
Tonya G. Agostini and Catherine T. Best
Connecting CALL and second language development: online tandem
learning of Japanese
Satomi Kawaguchi
About the authors
Processability Theory (PT from now on) is a psycholinguistic theory of second language acquisition (SLA from now on), first formulated in a book-length publication by Manfred Pienemann in 1998. Since then, PT has earned a place in many
SLA introductions, handbooks, companions and encyclopedias such as Doughty
& Long (2003), Gass & Mackey (2012), Kroll & de Groot (2005), Macaro
(2013), Ortega (2009) and Robinson (2012). It is one of nine theories of SLA singled out and discussed by VanPatten & Williams (2007), along with four others
sharing a cognitive processing approach. Like other theories, PT has its strengths
and weakenesses, and does not deal with all the phenomena and processes constraining SLA or contributing to it. However, few theories seem to accommodate
such a variety of phenomena or offer the basis for so many new developments.
What PT offers is a principled transitional paradigm that deals specifically with
grammatical development, and accounts for it. It also contributes an explicit and
universal definition of developmental stages, which may be applied in principle to
any language-specific developmental trajectory, and which in turn offers a stable
point of reference for investigating typologically diverse L1-L2 constellations,
learning modalities, environments and populations. From a practical point of view,
PT can help in assessing language development in individual learners as well as in
constructing a syllabus appropriate for their stage of development. In terms of new
directions, as Jordan (2004: 227) remarked, PT “can be seen as ‘progressive’ […]
extending its domain, refining its concepts, making the variables more operational,
attracting more research.”
More than fifteen years have passed since the publication of Pienemann’s book
on PT in 1998; and before that, it took almost twenty years to mould into PT the
initial achievements of the ZISA (Zweitspracherwerb ausländischer Arbeiter)
Multidimentional Model (cf. foremost Meisel, Clahsen & Pienemann 1981). Over
this period, not only has the whole field of SLA grown exponentially, but PT’s feeder disciplines have also advanced significantly. PT has paralleled this growth, and
widened its scope in several directions. We very briefly retrace its history in order
to place our volume within this developing context.
In the early eighties the ZISA team proposed a Multidimentional Model to
account for their considerable body of data and, in particular, to explain the staged
development of German word order. The two dimensions of the Model were, on
the one hand, a psycholinguistic and potentially universal dimension, and on the
Grammatical development in second languages, 7-14
other, a socio-psychological one, no longer pursued. Within a broadly Chomskyan
generative framework, the ZISA team guided by Jürgen Meisel provided the SLA
field with two further fruitful contributions: first, while lively debates moved SLA
more and more towards processing approaches (e.g., McLaughlin 1978, 1980 vs
Krashen 1976, 1979), Harald Clahsen proposed an explanation for the apparently
invariant stages observed in the acquisition of German word order in terms of cognitive strategies, the accumulation of rules determining the learner’s grammatical
progress (Clahsen 1980); and secondly, Manfred Pienemann successfully addressed
the theory-practice link by focusing on the connection between development and
teaching with his Teachability Hypothesis (Pienemann 1984, 1985, 1989), showing
experimentally that it is not possible to alter the trajectory of acquisition through L2
teaching. This latter hypothesis, which has since remained unchanged, is still, in our
opinion, the key to the popularity of Pienemann’s approach, its potential already
recognised in Larsen Freeman & Long (1991: 270-287).
When Manfred Pienemann moved to Australia in 1982, Malcolm Johnston’s
large structural English L2 data set (Johnston 1985a) provided the two researchers
with the challenge to test out the assumed universality of ZISA’s cognitive account.
Assuming the universality of human cognition, if the constraints operative for the
development of German L2 syntax are cognitively based, then they should turn out
to be operative not only for German word order but also for other L2s and possibly other domains, such as morphology, provided the target structures meet the
requirements of those processable by each particular strategy (Pienemann &
Johnston 1986; 1987). In other words, this framework had a predictive potential,
as Larsen-Freeman & Long (1991: 287) pointed out. Thus, taking on board
Pienemann’s (1984) teachability experiments and other contemporary works
pointing in the same direction, the framework could predict for example that, no
matter what the teacher may try, the learners would not ‘skip’ a stage of acquisition
and could only learn what they are developmentally ‘ready’ to learn. The reader will
quickly notice that, having specified what belongs to what stage, this prediction is
falsifiable. Pienemann’s Teachability Hypothesis however did not dismiss the role
of instruction altogether. Rather it constrained its potential to improvements in the
rate, not the route, of acquisition. This ‘predictive framework’ provided key theoretical insights, and generated publications by both Johnston and Pienemann on a
range of issues of interest to SLA at the time (cf. Di Biase 2000), such as whether
there is a ‘natural’ progression in learners of English from different L1 background
(Johnston 1985b), what factors influence language development (Pienemann &
Johnston 1987), how learners develop their own grammar (Johnston 1987), and
what exactly is the influence of instruction on L2 processing (Pienemann 1987).
An early result of the Pienemann and Johnston predictive framework’s potential
was the application, in association with Geoff Brindley, of the hypothesised developmental stages to the construction of a principled and interlanguage-sensitive lan-
guage assessment and testing procedure for ESL (Pienemann, Johnston & Brindley
1988), providing the foundation for Pienemann’s later Rapid Profiling. The application of this innovative predictive framework to English, however, was not without problems, as pointed out by Larsen-Freeman & Long (1991: 275-ff), yet the
schedules for the development of English morphosyntax elaborated in the eighties
on Johnstons’s (1985a, 1985b) cross-sectional data from 12 Polish and 12
Vietnamese immigrants to Australia have provided key empirical evidence ever
since (cf., e.g., Pienemann 1998: 165-181; Pienemann 2011a: 8-11).
Over the nineties, the empirical base of PT widenend in several directions.
First, the structural results on German L2 and English L2 data were supported by
positively testing the Steadiness Hypothesis, according to which task variation does
not produce variation in the learner’s procedural schedules (Pienemann, Mackey &
Thornton 1991; Pienemann & Mackey 1993). Secondly, a fruitful connection was
established between input and interaction on the one hand and development on
the other in the area of question formation (Mackey 1999). Thirdly, the typological validation widened with initial work on the acquisition of Japanese (Doi &
Yoshioka 1987, 1990; Huter 1996, 1997), and Arabic (Mansouri 1995, 1997,
1999). Thus, the limitations of the original strategy-based explanation became
more evident, as much as the need for a more explicit approach to grammatical representation if the theory was to work cross-linguistically. PT came about when
Pienemann (1998), abandoning the strategies approach, grounded his new theory’s psychological plausibility on Levelt’s ‘Blueprint for the Speaker’ (1989), an
explicit model of language generation, itself a successful integration of several
strands of psycholinguistic research focusing on speech processing. At the same
time, following Levelt’s (1989) own example, as well as Pinker’s (1984), Pienemann
(1998) introduced a second crucial innovation that provided PT with the necessary grammar-theoretical basis to test and support its typological plausibility,
namely, Kaplan & Bresnan’s (1982) Lexical Functional Grammar (LFG from now
on). In Kaplan’s (1995: 7) own words the “LFG formalism, which has evolved
from previous computational, linguistic and psycholinguistic research, provides a
simple set of devices for describing the common properties of all human languages
and the particular properties of individual languages”. PT then further broadened
and consolidated its typological spread, moving from its German and English focus
towards a greater variety of languages, such as Arabic (Mansouri 2002, 2005),
Chinese (e.g., Zhang 2002, 2004), French (Ågren 2009), Italian and Japanese (e.g.,
Di Biase & Kawaguchi 2002; Kawaguchi 2005), as well as Swedish and other
Scandinavian languages (e.g., Pienemann & Håkansson 1999; Glahn et al. 2001).
Further PT developments over the first decade of this century include the
Developmentally Moderated Transfer Hypothesis (Pienemann, Di Biase,
Kawaguchi & Håkansson 2005) to account for L1-L2 transfer, following on from
work on typological proximity (Håkansson, Pienemann & Sayehli 2002). This
hypothesis claims that transfer from L1 to L2 can only happen if it is in accordance
with PT’s schedules, that is, in a nutshell, you can transfer only what you can
process. Furthermore, the plausibility of the theory was extended to acquisition
contexts other than classic adult L2 ones, such as bilingual language acquisition
(e.g., Itani-Adams 2008), children with Specific Language Impairment (e.g.,
Håkansson 2001, 2005; Håkansson, Salameh & Nettelbladt 2003), child L2 learning (Yamaguchi 2010), and in the emergence of creole languages (Plag 2008a,
2008b, 2011). Finally, the range of applications of the original PT to language
teaching and language testing has also expanded over the years, involving several
new languages, teaching situations and ways of testing, while generating new
hypotheses, such as the syllabus construction hypothesis (Pienemann & Keßler
2007), and the developmentally moderated feedback hypothesis (Di Biase 2002,
2008; Nuzzo & Bettoni 2011); as well as further work on question formation in
English as foreign language contexts (Sakai 2008).
Some ten years ago, PT’s framework (Pienemann 1998) widened substantially when Pienemann, Di Biase & Kawaguchi (2005) followed up on an idea first
introduced in 2002 by Di Biase & Kawaguchi in coming to terms with their two
nonconfigurational languages, Italian and Japanese respectively. By incorporating
LFG’s syntacticised discourse functions and Lexical Mapping Theory (Bresnan
2001; Dalrymple 2001; Falk 2001), PT added a new discourse-pragmatically
motivated syntactic component to its ‘classic’ syntactically motivated morphological module. Since the formulation of this 2005 extension, its two main hypotheses, the Topic Hypothesis and the Lexical Mapping Hypothesis, have been tested
on various languages (cf., e.g., Itani-Adams 2009; Zhang 2007; Yamaguchi 2010;
and Bettoni & Di Biase 2011 for the former hypothesis; mainly Kawaguchi 2007,
2009a but also by Wang 2009, and Keatinge & Keßler 2009 for the latter hypothesis). However, these works have just begun to explore the two hypotheses, and
more needs to be done.
Given the above historical sketch of PT, the aim of our volume is twofold.
First, we intend to provide a consistent, if concise, new presentation of PT’s main
tenets, as we see them in connection with our proposal for a broader ‘prominence
hypothesis’ to approach the syntax-discourse interface (cf. part I). We feel that a
new presentation is necessary in order to clarify some theoretical and terminological issues. For instance, in Pienemann & Keßler (2011), some of the original transformational terminology from much earlier work (e.g., Pienemann, Johnston &
Brindley 1988) is still used for describing English PT’s schedules. While this may
have some practical advantages in terms of continuity, we offer some proposals for
updating the terminology so as to incorporate current LFG into PT’s framework
in a consistent fashion. For example, connecting LFG’s discourse functions and its
Lexical Mapping Hypothesis to L2 development enables us to investigate their operation analytically, and interpret the learner’s ability, among others, to topicalise a
grammatical function other than the subject or choose between active and passive
constructions as something other than a purely structural operation. In other
words, the processing of discourse-pragmatic information does play a principled
role, and after all LFG is a declaredly nonderivational grammatical theory. As
Asudeh & Toivonen (2010: 454-455) put it, a basic principle of LFG theory is that
“grammatical information grows monotonically (Bresnan 2001: ch. 5), i.e., in an
information-preserving manner. […] One general consequence is that there can be
no destructive operations in syntax. For example, relation-changing operations,
such as passive, cannot be syntactic, because that would require destructive remapping of grammatical functions” (emphasis in the original).
The second aim of our volume is to make a contribution to theory construction. For example, we take a more analytical approach to PT and look first at morphology separately from syntax, at declaratives separately from questions, and then
at their interfaces. Furthermore, we propose to recast the 2005 Topic Hypothesis
as the Prominence Hypothesis, and reformulate the Lexical Mapping Hypothesis.
The changes in the staging of L2 development deriving from these proposals, broadly presented in part I, are then illustrated in part II through the recast learners’
schedules of three typologically distant and well tested languages in PT: English,
Italian and Japanese. Furthermore, we intend to contribute to theory construction
in PT by exploring new possibilities and providing a coherent context for current
work – that is, new work which draws on the consequences of developments in
PT’s feeder disciplines, and explores issues, languages and applications not previously treated in PT. Thus, in part III of our volume the scope of PT is widened in
several directions. Among them, Russian is a new language for PT, which exemplifies the way in which grammatical case can, and must, integrate morphological and
syntactic considerations. Another new area is the treatment of Differential Object
Marking, as it emerges in Spanish L2, itself a language scantily examined within
PT (but cf. Bonilla 2012, 2014). Constituent questions, an older SLA area, receive new treatment within our Prominence Hypothesis, thanks to Mycock’s (2007)
pioneering work within the nonderivational LFG framework. For instance, our
more analytical approach to PT enables a comparative treatment of questions and
declaratives, leading to results for German L2 development that are theoretically
interesting in the history of PT. Part III also includes explorations into whether
PT’s stages hold also with autistic L2 learners, and concludes with a sally into
technological innovations applied to language teaching and learning, and the role
PT may play in this important area. Our hope is that the exploratory nature of
some of this work will suggest potential lines of further development for both the
theory itself and its applications – thus contributing to PT’s promise as a ‘progressive’ theory. Where the evidence for our new hypotheses is sometimes based on few
learners – even one learner in longitudinal studies – we can only invite researchers
to test our claims on richer data sets.
The principal audience for this volume consists of SLA researchers and graduate students, advanced undergraduate students and their instructors. The editors
and authors assume little previous knowledge of PT on the part of the reader. On
the other hand, we must warn readers that this is not a comprehensive introduction to PT, because our volume does not discuss important areas such as the
Developmentally Moderated Transfer Hypothesis, learner variability or applications such as the Rapid Profiling. In addition to limitations of space in a single volume, these gaps may also be attributed to our wish to prioritise for PT fresh interpretations of earlier achievements, new ground, and new pointers for future
The ideas presented in this volume build on previous research on PT by
numerous colleagues, relying foremost, of course, on Manfred Pienemann’s (1998)
cornerstone publication. This does not mean that the originator of PT agrees with
all aspects of our interpretation of his theory or with all our contributing authors’
new developments. His reservations have helped us clarify our own thoughts, and
we thank him for sharing them with us. Care is taken throughout the volume by
editors and authors alike to signal to the reader differences or potential disagreement between previous versions of PT and the explorations presented here, and to
provide references to specific works whenever appropriate.
Among our many colleagues working in the PT framework, we wish to single
out Gisela Håkansson, Yuki Itani-Adams, Junko Iwasaki, Louise Jensen, Satomi
Kawaguchi, Jörg Keßler, Fethi Mansouri, Gabriele Pallotti, Manfred Pienemann,
and Yanyin Zhang among the older generation; and Daniele Artoni, Marco
Magnani, Lucija Medojevic, Elena Nuzzo, Jacopo Torregrossa, Yumiko Yamaguchi,
Kenny Wang, and Karoline Wirbatz among the younger generation. We thank
them all not only for their scholarly contribution but also for the personal friendships that have developed with them over the many PT meetings held in Australia
and Europe in the last two decades or so. As editors, we wish to thank foremost
Gabriele Pallotti and Manfred Pienemann, then, alphabetically, Cinzia Avesani,
Cathi Best, Gisela Håkansson, Barbara Hinger, Louise Jansen, Jörg Keßler, Marco
Magnani, Louise Mycock, Elena Nuzzo, Lourdes Ortega, Valeria Peretokina, Ingo
Plag, Ruying Qi, Peter Robinson, Jason Shaw for their help and encouragement,
and the anonymous reviewers for their thoughtful and constructive comments on
the volume as a whole and/or single chapters. Needless to say, remaining errors are
our responsibility.
30 May 2015
Camilla Bettoni
Bruno Di Biase
1st person
2nd person
3rd person
argument structure
Computer Assisted Language Learning
Computer-Mediated Communication
complementiser phrase
constituent structure
discourse function
Differential Object Marking
English as a second language
functional structure
grammatical fucntion
inflectional phrase
Japanese as a second language
first language
second language
Lexical Functional Grammar
noun phrase
secondary object
oblique family
object–verb–subject (word order)
Processability Approaches to Language Acquisition
preposition phrase
Processability Theory
question phrase
question particle
question feature
question word
second language acquisition
Specific Language Impairment
subject–object–verb (word order)
subject–verb–object (word order)
target language
verb phrase
verb–subject–object (word order)
Universal Grammar
open phrase
Introducing and developing Processability Theory
Camilla Bettoni* and Bruno Di Biase**
*University of Verona, **University of Western Sydney
This initial part of the volume includes a presentation of PT’s main tenets,
which is designed first of all to highlight particular developments and proposals
that have arisen since Pienemann (1998). Because much of the substantial
progress in PT relies on developments in its two theoretical sources, in our exposition we give priority to Levelt’s (1989) psycholinguistic Model and later developments (Bock & Levelt 1994; Levelt, Roelof & Meyer 1999; Levelt 2000) for
language production, and to current LFG (Bresnan 2001; Dalrymple 2001;
Dalrymple & Nikolaeva 2011; Falk 2001; Asudeh & Toivonen 2010) for language description, and allow them as much space as that reserved for the learner’s developing path. Furthermore, this initial part is designed to give conceptual coherence and terminological consistency to the whole volume, thus avoiding
repetitive introductions in each of the subsequent chapters. In this way, we hope
to achieve two further aims: to illustrate PT’s universal schedules of grammatical development and our ‘prominence’ proposal in a way that is consistent with
its feeder disciplines, and also, crucially, to try to explain further the reasoning
behind the schedules and the way they are connected, thus contributing to theory construction in SLA.
The progress in the two feeder disciplines that, in our view, bears most fruitful consequences for PT’s developments and proposals concerns, on the one hand,
(a) Levelt, Roelof & Meyer’s (1999) Theory of Lexical Access, and (b) Bock &
Levelt’s (1994) specifications about the sequencing of the formulator’s encoding
procedures, with regard to language production; and on the other hand, (c) the formal representation of discourse functions (DFs from now on), and (d) the Lexical
Mapping Theory in LFG’s framework (Bresnan 2001; Dalrymple 2001;
Dalrymple & Nikolaeva 2011; Falk 2001) with regard to linguistic knowledge.
Consequently, in this part I, the main novelty in terms of contribution to theory
construction derives from a coherent deployment of this progress in explaining the
learner’s developmental path by proposing our Prominence Hypothesis, as well as
in attempting to solve incongruities deriving from Pienemann’s continuing reliance
on the older version of Levelt’s Model, the earliest LFG, and derivational syntax
(cf., Pienemann & Keßler 2011).
Grammatical development in second languages, 15-18
Camilla Bettoni and Bruno Di Biase
In relation to Levelt’s Model, the Theory of Lexical Access has, crucially, introduced a third conceptual component to the original lemma-lexeme dichotomy.
This development has its main application for PT in the Lexical Mapping
Hypothesis by Pienemann, Di Biase & Kawaguchi (2005), whereby the (conceptual) lexical requirements of the verb contribute to driving the structural choices in
the clause. Moreover, Bock & Levelt (1994) clearly show that, in the temporal
course of language production, functional processing precedes positional processing. In PT terms this means that learners will first build up their syntactic frames
and learn to assign grammatical functions (GFs from now on) to the retrieved lemmas, and only later will they relate morphological inflection to the constituents.
In LFG, the formalisation of the DFs, particularly Topic (TOP) and Focus
(FOC) as syntacticised relations, has allowed for the development of a new dimension for PT, in so far as it can now formally represent promising areas – such as topicalisations, question formation, and other ways of attributing prominence – for
investigating the learner’s behaviour at the crucial intersection of syntax and discourse-driven choices. Furthermore, LFG’s Lexical Mapping Theory contributes to
explaining for PT how learners develop beyond rigid default mapping between thematic roles and GFs towards more flexible nondefault mapping in order to enhance
expressivity and establish a different perspective or point of view on the event they
intend to communicate. For instance, previous PT was silent on why learners who
have acquired canonical word order may fail to produce passives despite their
apparently pretty ordinary SV structure.
In presenting the learner’s progress in developing their L2 grammar, we introduce several main innovations. The first is the separation of morphological development from syntactic development. This is different from Pienemann’s own work
(e.g., 1998, 2005, Pienemann & Keßler 2011) and much of PT work so far. The
reason is that these two schedules appear to depend on two different sets of motivations. On the one hand, we have the original psycholinguistic procedures of
Kempen & Hoenkamp (1987) assumed by Levelt (1989) and adopted by PT in
Pienemann (1998), who shows how these processing procedures can be modelled
in LFG by the mechanisms of feature unification. These procedures trace the developmental path of the learner’s morphological marking beyond lexical learning over
the hierarchical (phrasal, interphrasal and interclausal) levels of syntactic organisation (cf. ch. 1, § 4.1). On the other hand, the development of syntax depends on
two different kinds of correspondences that formally relate three LFG parallel
structures: argument structure, functional structure, and constituent structure
(respectively a-structure, f-structure, and c-structure from now on). One set of
these correspondences – that is, the mapping of c-structure elements (NP, VP, and
so on) onto f-structure elements (SUBJ, OBJ, and so on) – describes the precedence relations of arguments according to the allocation of LFG’s grammaticised
DF, such as TOP and FOC (cf. ch. 1, § 4.2.1). The other set of correspondences
PART I. Introducing and developing Processability Theory
is guided by the principles of LFG’s Lexical Mapping Theory, which accounts for
the mapping of a-structure (a hierarchically organised set of semantic roles) to fstructure (a hierarchically organised set of GFs) (cf. ch. 1, § 4.2.2). Furthermore,
by keeping morphological development separate from syntactic development, we
are able to clarify and investigate issues involved in the interface between them (cf.
ch. 1, § 4.3).
Other important innovations in chapter 1 concern the three hypotheses proposed in Pienemann, Di Biase & Kawaguchi’s 2005 extension. In § 4.2, we abandon one of them, the Unmarked Alignment Hypothesis, and propose a reformulation of the other two, the Topic Hypothesis, and the Lexical Mapping
Hypothesis. First, the Unmarked Alignment Hypothesis is abandoned because, at
the initial grammatical stage, canonical word order and canonical (default) mapping between thematic roles and GFs do not necessarily entail each other. In fact,
under pragmatically marked conditions a thematic role other than the highest in
the hierarchy (e.g., the theme) may occupy the most prominent position in the
string (i.e., the first). Secondly, in § 4.2.1, the original Topic Hypothesis is recast
as Prominence Hypothesis, so that, parallel to the development of topicality in
declarative sentences, it now explicitly includes the development of focality in
interrogative sentences, thanks to the more general processing principle whereby
any constituent can be made prominent by grammatical means. Among these
means, together or separately, Levelt (1989) includes early appearance in the sentence (cf. ‘linear precedence’ in Sells (2001: 1), Choi’s (2001) work on information
structure, and Lee’s (2001) work on word order), and mapping the role with the
allocated prominence onto the highest-ranking GF (i.e., SUBJ) – as well as prosody, which however we do not treat because it clearly lays outside the scope of this
volume. Thirdly, in § 4.2.2, our Lexical Mapping Hypothesis now includes an
intermediate stage between the initial default mapping stage and the final nondefault mapping one. Thus our schedules for the development of syntax (based on
the Prominence Hypothesis and the Lexical Mapping Hypothesis) now both
hypothesise a middle stage which acts as a sort of lockpicker to open up their
respective higher stage.
Finally, our presentation of PT in chapter 1 attempts to reflect more consistently a basic assumption about language development shared not only with many
SLA researchers but also with L1 acquisition researchers and typologists. With
Andersen (1984), Brown (1973), Keenan & Comrie (1977), Krashen (1982), and
many others, PT assumes that the learner proceeds from least marked, feature-scant
forms and structures towards more feature-rich, more specified and more marked
forms and structures. Our contribution here, from a processability vantage point,
attempts to throw some light on how learners proceed from obligatory defaults
towards the skills needed to handle discourse-pragmatically induced and grammatically laden options in their L2 communication.
Camilla Bettoni and Bruno Di Biase
This path from obligatory grammar towards the deployment of a greater range
of grammatical options is a characteristic that distinguishes the more advanced
learners from beginner and intermediate learners, thus coming to terms with a perceived gap in PT, which has so far dealt more thoroughly with obligatory grammar
in early interlanguage.
Processability Theory:
theoretical bases and universal schedules
Camilla Bettoni* and Bruno Di Biase**
*University of Verona, **University of Western Sydney
1. Introduction
PT can be seen as a ‘progressive’ theory, that is, a theory capable of extending its
domain, refining its concepts, making its key variables more operational, and
attracting more research (Jordan 2004: 227). Even as we write, PT is expanding
rapidly, and our presentation must inevitably be limited. But it is limited in four
further ways. First, in this chapter we will mention PT’s history only when it
helps explain some of the incongruities we try to eliminate, or justify our own
choices. Secondly, our outline here is not intended as an independent introduction to the theory, in the sense that we will mention only minimally PT’s main
scope, constructs and processes. These can be found in the original works by
Pienemann (1998, 2005b; Pienemann, Di Biase & Kawaguchi 2005; including
the more recent Pienemann & Keßler 2011, 2012) and in Pienemann’s own
numerous shorter presentations (e.g., 2003, 2007) – although with regard to the
latter a note should be added to the effect that they mostly refer to the acquisition of English, and rely on older versions of PT, barely touching upon the 2005
extension. Thirdly, although we will mention some problems in the theory, we
do not intend to solve them all here. Nor, finally, can this chapter be read as a
full review of the rich and varied PT literature.
On the positive side, our ambition here is to offer an outline of PT which is
tightly anchored to its two updated psycholinguistic and theoretical linguistic bases
and less dependent on the English language for exemplification.1 So, we aim to
offer a balanced synthesis: both critical in pointing out areas of weakness in PT, and
enthusiastic in showing how its hypothesised universal schedule can generate paral-
1 Of course, we will continue to use English as the main l2 for illustrating PT, because
it is the language familiar to most readers and most studied acquisitionally. Yet when
we do so, care is taken to point out its typological peculiarities and present our discussion in such a way as to accommodate the widest possible crosslinguistic variation.
grammatical development in second languages, 19-80
Camilla Bettoni and Bruno Di Biase
lel language-specific schedules widely across languages (as part ii shows) and in a
variety of situations (as indeed the chapters that follow in part iii illustrate).
Needless to say, our focus here is not so much on the details of PT’s developmental schedules as on the reasoning behind them.
in sum, our first main focus here is on integration and coherence among what
is at times separately and varyingly treated in PT literature: to wit
• between PT itself and its two theoretical bases (i.e., levelt’s psycholinguistic model
for language production, and lFg for language representation);
• between the original 1998 version by Pienemann and its 2005 extension by
Pienemann, Di Biase & Kawaguchi; and
• among studies on different languages which are based on one or the other of these
two versions.
This will help us offer our own proposals as a contribution to theory construction,
which is the second main aim of this chapter. in turn, this will provide an opportunity for revisiting the earlier developmental schedules of three typologically
representative languages in part ii of the volume, and preparing the grounds for the
new developments and explorations discussed in part iii.
The remainder of this chapter is organised as follows: § 2 deals with language production (i.e., levelt’s model, in § 2.1) and linguistic knowledge (i.e.,
lexical Functional grammar, in § 2.2) in relation to PT; § 3 summarises PT’s
key concepts; § 4 traces the learner’s progress, first for morphology (§ 4.1), then
syntax (§ 4.2) according to the Prominence hypothesis (§ 4.2.1) and the
lexical mapping hypothesis (§ 4.2.2), and finally discusses how the morphological schedule and the two syntactic ones may interface; § 5 sets out some
methodological issues for testing PT’s hypotheses; and § 6 offers some concluding remarks.
2. Language production and linguistic knowledge
Closely paraphrasing Pienemann (1998, 2007), the underlying logic of PT is that
l2 learners, at any stage of development, can produce only those l2 forms which
the current state of their language processor can handle. This places PT among
those Sla theories that see language primarily as a mental construct, like other
generative approaches, but unlike most of them it considers the acquisition of an
l2 as connected to real time human performance constraints such as those regarding speech processing. it is therefore crucial to base our understanding of language development on PT’s two formal models, accounting for (a) language generation, that is, how the processor handles language, and (b) linguistic knowledge, that
1. Processability Theory: theoretical bases and universal schedules
is, what is language and how it may be represented in our mind/brain. given that
the anatomy and physiology of the human language processor are universal, if specific languages and their development are described according to a principled architecture, it may be possible to predict broadly similar grammatical trajectories for l2
development across languages.
For language generation, PT relies primarily on levelt’s model (1989), a
dynamic model accounting for language processing in real time and within
human psychological constraints, such as the requirement for a very fast word
access and, at the same time, the limitations of human memory. Thus a set of
psycholinguistic universal constraints comes to bear on l2 acquisition, and
provides a framework for PT’s universal hierarchy of processing procedures,
accounting for why learners seem to follow similar paths in their l2 grammatical development, across languages, whatever their l1 background. For linguistic knowledge, on the other hand, PT relies on lFg as originally conceived
by Kaplan & Bresnan (1982) and further developed by Bresnan (2001),
Dalrymple (2001), and Falk (2001) among many others over the last three
decades. lFg provides a generative, explicit and well-defined, formal theory of
language, which contributes towards solving for PT not only the problem of
relying on a plausible mental representation of grammatical structure, but also
Chomsky’s well known ‘logical problem’: what is the origin of linguistic knowledge? Why do people end up knowing more than they can hear? as Pinker
(2004: 949) puts it: “it is the question of how acquisition could work in principle – how a learner can correctly generalize from a finite sample of sentences
in context to the infinite set of sentences that define the language from which
the sample was drawn.” For example, how does a learner know that words can
be nouns, verbs, etc., or that in the sentence The boy who loves Therèse is
Indonesian, Therèse may not be indonesian despite the sequence Therèse is
Indonesian? The interface between these two formal theories, levelt’s model
and lFg, allows PT to make language specific predictions about l2 development which can be tested empirically.
These two source theories of PT interface well because lFg is a constraintbased theory of generative grammar (asudeh & Toivonen 2010) which intends to
be psychologically plausible. indeed, its suitability for psycholinguistic work is supported by the fact that lFg was chosen for language acquisition work, such as
Pinker’s (1984), as well as levelt’s model of the speaker. There are good reasons for
this compatibility. First, unlike other generativist frameworks, lFg’s approach is
decidedly lexicalist, declarative and nonderivational, which sits well with the psycholinguistic understanding that processing is formulated in one stage (Pickering,
Branigan & mclean 2002), i.e., transformations are unlikely. Secondly, it explicitly connects competence with performance phenomena, in so far as, for example,
Camilla Bettoni and Bruno Di Biase
lFg’s feature unification is related to performance in real time (cf. rothman &
VanPatten 2013). Thirdly, lFg’s parallel projection architecture iconically suggests
a representation of the dynamic processes temporally modelling language production. in particular, the fact that lFg’s interest in typological questions of similarities and differences among languages has been central to its theoretical development (asudeh & Toivonen 2010) finds parallels, mutatis mutandis, with PT’s own
history and current developments.
The next two sections (§§ 2.1-2.2) are meant to introduce the reader to those
key features of PT’s source disciplines that bear most directly on (i) its general
architecture, and (ii) some of the issues mentioned in the volume. in no way do
these two sections dispense the researcher interested in pursuing the finer details of
PT from reading the original works by levelt and Bresnan, and the ongoing updates by their respective teams – anymore than s/he is dispensed from reading
Pienemann’s own original works on PT.
2.1. Levelt’s Model and PT
The debt to levelt’s model is already fully acknowledged in the original PT version, when Pienemann (1998: §§ 2.4-2.5) highlights issues concerning the storage of grammatical information during language production, and the general psychological constraints that bear on language development. Thus we now focus on
elements of language generation that help us understand PT’s newer developments. These concern mainly two aspects of levelt’s model: the lexicon and grammatical encoding. First, following developments in levelt’s model itself (levelt,
roelofs & meyer 1999), a novel look at lexical access theory allows for a more precise characterisation of the lexicon and its central role in PT, which can now
account for the acquisition of a wider range of features and constructions. if
Pienemann’s main concern in 1998 was the establishment of minimal requirements for reaching a stage, PT can now explore how learners proceed from the
emergence of a structure to its mastery, or indeed from the emergence of one or
two ‘typical’ structures in a stage to the mastery of the range of structures in that
same stage. This contributes to accounting for development within a stage, which
mansouri & håkansson (2007) began to explore as intrastage phenomena and we
further characterise in this volume as progress through ‘soft barriers’ (cf. § 5 below,
and further ch. 2, § 2 for English, and ch. 3, § 2.2 for italian). That is, whereas
Pienemann (1998) proposed modules other than PT for handling the complexities within a stage, we propose instead to integrate them into PT’s explananda.
Secondly, we wish to outline that part of levelt’s model which bears more directly on PT’s extension, because the extension of the theory by Pienemann, Di Biase
& Kawaguchi’s (2005) dealt more fully with its lFg formalism rather than
discuss the details of language production.
1. Processability Theory: theoretical bases and universal schedules
levelt’s model, along with other speech production models that have been
proposed from various theoretical viewpoints over the last 25 years or so, share
the understanding that the speech production process can be described as consisting of three broad components: conceptualisation, formulation and articulation (cf. grosjean & li 2013: 51). levelt assumes that when we intend to say
something we select in the conceptualiser the information whose expression
may realise our communicative goals – the conceptualiser being the processor
where the preverbal message is generated and then fed to the formulator, as the
fragments of the preverbal message become available. Since any state of affairs
can be expressed in many different ways, in the conceptualiser we also plan the
form of the message, in the sense that here we select not only the language and
register but also the appropriate speech acts or rhetorical device required (assertion, question and so on), we tag topic and focus, mark the referents as given or
new, and so on. Thus, the conceptualiser’s output (i.e., the preverbal message)
already includes information on a number of choices, including the relative prominence of its elements. all this presents no problem for adult l2 learners, who
are already fully competent speakers of their l1, because, in order to produce a
preverbal message they can rely on the same conceptualiser for either language,
as De Bot (1992) maintains in adapting levelt’s model of language production
to bilingual speakers.2 On the other hand, grammatical learning begins when
then the formulator – which is language-specific – receives input from the conceptualiser, and has the task of mapping the preverbal message onto the appropriate linguistic form (i.e., the formulator is responsible for grammatical encoding), and preparing (through phonological encoding) the phonetic plan, as
represented in (1). This task is performed by retrieving out of the lexicon the
stored entries that best fulfil the conditions required by the preverbal message.
Thus, just as the learner has to learn, however gradually, a language-specific lexicon, so do they also have to build, contextually, a language-specific formulator
to process it. levelt (1989: 103) assumes, in fact, that “[t]here are different formulators for different languages”.
let us then take a look in turn at how lexical entries are stored in the lexicon,
and then processed in the formulator.
2 Once the language has been specified in the conceptualiser, the output is language-specific because “[w]hatever the speaker tends to express, it should ultimately be cast in
terms of lexical concepts, that is, concepts for which there exist words in the target language” (levelt, roeleof & meyer 1999: 8). Despite considerable growth in psycholinguistic experimental research of spoken language processing in bilinguals, “a comprehensive model remains to be developed” (grosjean & li 2013: 52).
Camilla Bettoni and Bruno Di Biase
(1) Language production from preverbal message to phonetic plan (after Levelt 1989: 9)
in the lexicon of the adult l1 speaker, according to levelt, roelofs & meyer’s (1999)
Theory of lexical access, words are stored with the full bundle of features involving
three types of information, distributed in a three-level system: the conceptual level,
the lemma level and the lexeme level. We should point out that the conceptual level
was not present in Pienemann’s 1998 application of levelt’s model to PT, nor has it
been incorporated in Pienemann’s subsequent versions, which retain levelt’s (1989)
two-tier system of the lemma and the lexeme. in levelt, roeleof & meyer’s (1999)
three-tier system, instead, conceptual information include elements associated, for
example, with the semantic roles of V (e.g., <eater, eaten> for eat), rather than infor(2) A fraction of the lexical network for the word goat (after Bock & Levelt 1994: 951)
1. Processability Theory: theoretical bases and universal schedules
(3) A fraction of the lexical network for the word escort (after Levelt, Roelofs & Meyer 1999: 4)
mation about the number and type of arguments only. Notably, PT’s extension
would not work without the conceptual tier of the lemma.
in (2)-(3) we give a simplified representation for two lexical entries, the N goat
and the V escort, following respectively Bock & levelt (1994: 950-952), and levelt,
roelofs & meyer (1999: 3-4).
First, at the conceptual level, knowing a word involves knowing its meaning.
about a goat we know it is a kind of domestic animal that produces milk, etc., and
also that it typically de-selects certain other words such as think or smile typically
reserved for humans, etc. about escort we know that it is an action related to
accompanying, guiding, etc., that it requires the two semantic roles of agent and
patient, etc. These are properties of our concepts gOaT and ESCOrT.
Secondly, at the lemma level, a word has syntactic properties, a bundle of
grammatical features – including also combinatorial information (cf. Sells 2001;
Kim & Sells 2008) – which place it in its syntactic frame. The English word goat
is an N. its italian equivalent capra is also an N, but in addition it has feminine
syntactic gender. The word escort is here a V, and Vs are specified for (in other
words, they select) the arguments they require, corresponding to their semantic
roles (<agent, patient> in this case); thus about escort we know that it typically
Camilla Bettoni and Bruno Di Biase
requires a SUBJ and an OBJ on which to map the required semantic roles. as will
be clear further on, these lexically determined requirements of the V for argument
and function specifications are of crucial importance for understanding, and operationalising, PT’s lexical mapping hypothesis, first advanced by Pienemann, Di
Biase & Kawaguchi’s 2005 (cf. § 4.2.2 below). moreover, many lemmas require socalled diacritic features that have to be learned. For example, in English V lemmas
have diacritic features of person, number, tense, and mood, which must be valued
for further encoding. hence the lemma escort will be realised phonologically as
escort, escorts, escorted, or escorting depending on the values of its diacritic features.
Some values of these features derive from conceptual representation, as when
English Vs are marked for tense, or Ns for number, others may be set in the course
of grammatical encoding, as we will shortly see below.
Thirdly, at the lexeme level, knowing a word means knowing its formal properties that is, its morphological and phonological shape. The word goat is
monomorphemic and consists of three phonological segments: /g/, /ou/, and /t/,
whereas the italian word capra consists of two morphemes, a stem (capr-) and a suffix (-a), and five phonological segments: /k/, /a/, /p/, /r/, and /a/. likewise, in (2),
the nodes at the form level represent phonemic segments.
in levelt’s model, it is the lexicon – with its associated semantic, grammatical
and phonological information – that primes the procedures and feeds forward the
encoders. as we move on to grammatical encoding, let us remember that all this
information is characteristically stored in the mature native speaker’s mental lexicon,
but learners build up their l2 lexicon gradually. if, on acquiring a new word, learners may quickly be able to associate a ‘meaning’ at the conceptual level with some
(phonological) ‘form’ at the lexeme level, the same is certainly not true for the
lemma level, where features and values may take a long time to emerge3 and even
longer to master. So, whereas it will not take long for English learners of italian, for
example, to learn the word capra and get to know what it means (goat), it will take
them longer to learn that the N capra has the value ‘feminine’ with respect to the N
feature ‘gender’, which in turns combines with ‘number’ information, and hence, if
they want to refer to more than one, they need to use the form capre (not capras).
So, unlike many of the components at the conceptual level, which may find
a surprisingly high degree of commonality across languages, diacritic features and
their values at the lemma level tend to pop up in peculiarly idiosyncratic mixes
and with surprisingly different exponents requiring unification or merger over
varying degrees of syntactic distance – all very much in language-specific ways.
These idiosyncrasies of the l2 lexical features fuel one of the least understood and
3 Emergence can be understood here both as annotation of each feature-value pair in the
lexicon and their retrieval during production.
1. Processability Theory: theoretical bases and universal schedules
perhaps most confusing issues in PT, as well as in other approaches. The question
is why is it that structures that are apparently at the same stage of development
(e.g., marking some sort of aspectual feature with –ing or tense feature with –ed,
where both are category level morphemes) should emerge so far apart from each
other? Within PT this has been variously cast as ‘intrastages’ or ‘steps within a
stage’. in our treatment we choose the term ‘soft barriers’ to indicate that the
learner, having overcome the ‘hard barriers’ and learned the skill to go past the
important constraints imposed by a stage, will then learn more fine-grained lexically-based distinctions. Similarly at the lexeme level, there may well be peculiar
lexical class distinctions interacting with specific features at the lemma level that
may be rather difficult to acquire for adult learners (cf. Di Biase 2008 for one
example of such lexical class-based distinctions, affecting number in italian nouns
and adjectives). We will see below in § 2.2 how the lexicalist orientation of lFg,
which indicates feature-value pairs formally, is particularly apt in providing an
efficient representation (presence or absence of a particular feature/value) of such
phenomena.4 and we will come across specific examples of soft barriers in the language-specific chapters in Part ii.
The formulator – to resume our presentation of levelt’s model – encodes the
utterance first grammatically and then phonologically, as shown in (4). We are interested here in grammatical encoding, whose processes create the skeleton of the
(4) Components of grammatical processing (after Bock & Levelt 1994: 946)
4 as well as exploited by our proposal in the context of l2 learners’ progress, lFg’s lexicalist orientation, efficiently represented in formal annotation, is pointed out by
Schwarze (2002) in the context of (historical) language change.
Camilla Bettoni and Bruno Di Biase
utterance. These processes are grouped into two sets – one functional and the other
positional – each set with its own subcomponents. it may be worth noting that in
(4) the arrows represent the flow of information during production (and, in reverse order, comprehension) and depict activation trajectories, whereas in (2)-(3) they
represent types of connections within the network and depict permanent relationships in a store.
Functional processing has two subcomponents: lexical selection and functional assignment. given a lexical concept to be expressed, lexical selection involves
retrieving a word, or more specifically a lemma, from the lexicon. Functional assignment involves creating the appropriate syntactic environment for the words by
assigning them their syntactic functions. For example, upon selecting the lemma
escort, its syntax – it is a transitive V with two argument positions, corresponding
to its two semantic arguments – will become available for further grammatical
encoding (levelt, roelofs & meyer 1999: 4): which of the two arguments will
serve as SUBJ, which as OBJ?
Functional assignment is controlled by two kinds of information represented
in the message. First, the eventuality conceived in the conceptualiser is associated
with thematic or event roles, such as agent (the instigator of an event), patient or
theme (the person or object that is affected or moved). This explains why, in organising an utterance, the V lemma chosen during lexical selection is central over
other lemmas. Second, the relative prominence among the participants in the event
is associated with discourse or attentional roles. These organise the informational
distribution in the utterance so as to direct the listener’s attention to its components. as Bock & levelt (1994: 964-965) comment, there are “seductive correspondences” between both thematic and discourse roles and grammatical functions
(gFs). That is, agent is most often SUBJ, beneficiary is OBJ, etc., as in (5a),
although violations to these default correspondences are possible, as in (5b) or (5c),
depending on discourse-pragmatic information tagged in the preverbal message.
(5) a. romeo gives a rose to Juliet
b. Juliet is given a rose by romeo
c. Juliet receives a rose from romeo
likewise, elements expressing given (or topical) information, which are more readily available, often appear early in the sentence and have great affinity with SUBJ.5
This is shown in (6), where the same propositional content is expressed with diffe5 The developments in lFg since Bresnan (2001) go a long way in representing these
processing computation issues by having incorporated hierarchical semantic roles in
lexical mapping Theory and syntacticised discourse functions in its f-structure, as discussed in § 2.2 below.
1. Processability Theory: theoretical bases and universal schedules
rent prominence in (6a) and (6b) by assigning the SUBJ function to a different
topical element:
(6) a. [what’s going on with the dog?]
b. [what’s going on with the cats?]
the dog is chasing the cats
the cats are being chased by the dog
For the number and type of functions assigned during functional assignment,
we refer to their lFg representation reported in § 2.2 and shown in (21).
Suffice it to say here that, although functions are universal, they may be marked differently in different languages: morphologically by case markers, clitic
pronouns, etc., or structurally by position, as shown in (23)-(24). These different ways are not mutually exclusive, and indeed many languages use a combination of means. Even a highly configurational language like English marks the
case of personal pronouns morphologically (e.g., by distinguishing between I
and me, she and her).
Finally, during functional processing, the combination of lexical selection and
functional assignment also specifies the value requirements for the diacritic features of individual lemmas. For example, if the speaker intends to produce the sentence in (6a) above, upon selecting the V chase and the Ns dog and cat for expressing this current eventuality in the present involving dog as agent and cat as patient,
functional assignment will determine not only the grammatical relations between
the lemmas (i.e., dog (the chaser) is SUBJ and cat (the chased) is OBJ of chase), but
also the values of the diacritic features (i.e., dog, referring to a single referent, is realised in its singular form as dog; cat, referring to more than one referent, is realised
in its plural form as cats; and chase is realised in its present, progressive form chasing with the auxiliary carrying the required SUBJ number information with singular value).
in sum, functional processing yields an available and activated set of lemmas
and a set of abstract syntactic functions, which are linked together via the argument structure of the lemmas, notably that of V (cf. (7) for a schematic illustration of the product of functional processing of the sentence the dog is chasing the
cats). all this material (i.e., abstract relations or linkages among elements) may
contain some indication of the relative prominence assigned to various components, but it is not ordered in any sequence, except for the order in which the fragments become available from the conceptualizer. To convert into an utterance, the
fragments of this partial, incomplete structure do not go into the phonological
encoder directly as they come out of functional processing, but they are stored
temporarily in the memory buffer. The product of functional processing must
now be processed positionally.
Positional processing, like functional processing, also has two subcomponents: constituent assembly and inflection. Both involve the creation of a set of
Camilla Bettoni and Bruno Di Biase
(7) The product of functional processing for the dog is chasing the cats (after Bock & Levelt
1994: 968)
slots which are ordered: the former for lexemes, the latter for morphemes (Bock
& levelt 1994).
Constituent assembly fixes the linear order of word production and captures
dependencies among syntactic functions. Ordering is necessary because the output
of functional assignment carries no intrinsic order.6 This becomes clearest not with
English, a highly configurational language which marks grammatical functions
(gFs) by position, but with less configurational languages, whose constituents may
appear in different positions serving the same gFs, often signalled by differences in
case marking. russian and latin are such languages. in (8) for example, by marking the SUBJ (Paul) of the V amo (‘love’) as NOm case by means of the –us morpheme, and the OBJ (Mary) as aCC case by means of the morpheme –am, latin
can place either anywhere, as required by discourse or pragmatics, without changing the propositional content of the message:
(8) a. Paulus mariam amat
b. mariam Paulus amat
6 also lFg’s f-structure is not intrinsically ordered (cf. § 2.2).
1. Processability Theory: theoretical bases and universal schedules
Establishing dependencies among words means organising phrase groupings in a
hierarchy. Without them, as Bock & levelt (1994: 969) point out, there would be,
for example, no means to segment sentences such as (9) appropriately, where the
listener knows that it is not to be understood that it is the boy who blushed, despite the linear sequence showing the boy blushed:
(9) the girl who kissed the boy blushed
This hierarchical organisation (namely, the phrase structure) is assembled bit by
bit under the control of the syntactic functions and the grammatical categories
of the lemmas that realise them. This means that, for example, given the nominative function and a N lemma to fill it, adequate information is available to
create a SUBJ NP in the appropriate position in an utterance. as we will see further on, Bock & levelt’s claim that functional assignment precedes positional
constituent assembly in the temporal sequence of language production by the
native speaker is crucial for understanding learners’ development beyond canonical word order.
inflection is the last grammatical encoding process, and involves the generation of fine-grained details at the lowest level of the hierarchy of phrasal constituents, as shown in (10). This is a thorny issue, not yet solved in all its facets in
levelt’s model. The debate is around two questions: first, whether in such cases as
the English handing or italian capra the mental lexicon stores the whole word or
its morpheme components; and secondly, whether to consider under inflection
not only inflection proper, but also the formulation of function words often associated with grammatical phrases such as determiners for NPs, auxiliaries for VPs,
and prepositions for PPs (Bock & levelt 1994: 972). Suffice it to say here that
lFg’s principle of lexical integrity considers words as atoms from the point of
view of syntax, that is, they are no further divisible into smaller syntactic units (cf.
§ 2.2). as stated by asudeh & Toivonen (2010: 430), “[t]he syntax is […] blind
to the internal structure of words and it sees only their category”. Furthermore, as
mentioned above, certain lemmas carry specifications about diacritic features to
be valued inflectionally. in some cases these specifications may be under the control of conceptual elements, as when Vs are specified for tense. in other cases the
control is syntactic, as when there are dependencies among inflectional features.
So, in the sentence in (10) speakers say she was handing him some broccoli, rather
than she were, because here two constituents of the sentence reflect a value (i.e.,
third singular) of some feature (i.e., person) that triggers inflectional variation.
These constituents need not be adjacent. What is necessary is that the agreeing
constituents stand in appropriate syntactic-functional relationships. in this
English example, the agreement operates between the head of the SUBJ NP and
the finite V.
Camilla Bettoni and Bruno Di Biase
(10) Constituent hierarchy of she was handing him some broccoli (after Bock & Levelt 1994:
in sum, we can say that, first, functional processing integrates a set of lemma specifications with a set of syntactic functions. its output is a set of abstract relations
and properties which guides the creation of a frame for positioning words.
Subsequently, positional processing places words and their inflections into that
frame. The output of functional processing, then, is an ordered set of lexemes, formally realising the abstract relations of the functional specifications.
Why do we need to comprehend this complex process in order to understand
the way PT explains the learner’s progress in acquiring an l2? Though the reason
may become clearer in §§ 3-4, we can anticipate here that, whereas mature l1 speakers are able to activate all the encoder’s components effortlessly, l2 learners must
build them up gradually. if De Bot (1992) is correct in saying that bilingual speakers operate with a different formulator for each language, while the l2 formulator is under construction learners will be able to produce only those structures that
can be processed by the components already in place – as well as the lexical material already stored.
let us now go back to the native speaker. in levelt’s model this complex process of grammatical encoding for language generation assumes that grammatical
information activated by one procedure needs to be stored temporarily in a specialised memory buffer in order to be used by another procedure, so that the two lots
of activated information can then be compared by yet another procedure that
builds the output of the first two procedures together. how does this process
unfold? Following the incremental Procedural grammar developed by Kempen &
hoenkamp (1987), levelt (1989) proposes that grammatical encoding in mature
monolingual speakers follows the sequence in (11):
(11) a.
the lemma
the category procedure
the phrasal procedure
the sentence procedure
1. Processability Theory: theoretical bases and universal schedules
(12) An illustration of processing hierarchy for Kim eats a pear: phrasal and interphrasal
Thus, upon selecting the lemma, the category procedure is instigated, assigning a
lexical category to the lemma. Then the category of the head lemma will instigate
a phrasal procedure, resulting in a phrase. With the activation of the sentence procedure, phrases in turn acquire their function according to the syntactic frame of
their head lemmas. To illustrate these steps with an example, in Kim eats a pear,
shown in (12), first the lemma Kim needs to be assigned to the lexical category N,
and its diacritic features (number and person) returned with their respective values
(singular and third person). Then the lemma eats needs to be assigned to the lexical category V, and its diacritic features (number, person, tense, and aspect) annotated with their respective values (singular, third, present, and noncontinuous).
Further, in order to achieve the agreement between the two phrases Kim and eats,
information must be exchanged between them, and the values of the features they
share (number and person) must be compatible. likewise, in generating the NP a
pear, the selection of the lemma a partly7 depends on the value (singular) of the
diacritic feature of the phrase head lemma pear, because the common values they
share must be checked against each other, at the phrasal node, for compatibility.8
in this case the value of the diacritic feature of pear is stored by the category pro7 The choice of a vs the also depends on the information tagged in the ‘discourse model’ (in
the conceptualiser) determining its new vs old, definite vs indefinite values, and so on.
8 This notational innovation showing oly the unifying features and their values at the
node where unification is assumed to occur was introduced in lFg by andrews &
manning (1999).
Camilla Bettoni and Bruno Di Biase
cedure until it is checked against that of the modifier a. Finally, in order to build
up the sentence, a gF must be assigned to the two newly created NPs, that is SUBJ
for Kim and OBJ for a pear. This matching, or exchange, of information regarding
the values of shared diacritic features among the elements of a sentence is called
‘feature unification’ in lFg terminology (cf. § 2.2). it is a key lFg concept used
in levelt’s model of the (monolingual) speaker and incorporated into PT to
account for the gradual process of l2 development.
Kempen & hoenkamp’s (1987) incremental Procedural grammar also assumes that the whole four-step sequence from activating the simple lemma to the
sentence procedure is implicational. This means, for example, that in order to activate the phrasal procedure, both the lemma and the category procedures must be
activated, but that the sentence procedure need not be active as yet.
Furthermore, the whole process of language generation is incremental. This
means that all processors can operate simultaneously in parallel, but they all work
independently on different language fragments of the utterance under construction. The final order of articulation may follow the sequence in which fragments
are conceptualised, as in (13a), or it may not, as in (13b). although here the speaker has decided to make some specific circumstance (in this case the context of the
event) more prominent, the English formulator will not allow the fronting of both
the time fragment and the place fragment (which would produce *last week in
Rome Ugo sang). So the formulator, being the specialist processing component
charged with generating grammatically acceptable sentences, will order the fragments accordingly, and produce last week Ugo sang in Rome. These types of asynchronies do not create a particular problem or cost for the native speaker, who has
automatised the necessary procedures to handle diverse ways of organising the mes-
(13) Incremental production (a) without and (b) with inversion of order (after Kempen &
Hoenkamp 1987, as cited in Levelt 1989: 25)
1. Processability Theory: theoretical bases and universal schedules
sage online and delivering it in real time.9 But the processing cost of this very fast
process, averaging between two and three words per second (levelt 1989: 22), can
be very high for the learner.
Finally, the theoretical assumption of incremental processing (i.e., of parallel
processing activities in the different components of speech generation) hinges on
automaticity. automatic processes have great advantages: they are executed without
conscious awareness; they are quick; and they run on their own specialised resources, which means that they do not share resources, so they can run in parallel
without mutual interference. if each processor were to require access to attentional
resources such as Working memory, speaking would not be fluent, and between
one articulated fragment and the next there would be silences devoted to processing. On the other hand controlled processing demands attentional resources,
because we can only attend to a few things at a time. attending to the process
requires a certain level of awareness of what we are doing. Thus human controlled
processing tends to be serial rather than incremental in nature, and is therefore slower. its advantage is that it is not entirely fixated in memory; in fact, controlled
processing is quite flexible and adaptable to the requirements of the task at hand.
levelt (1989) accounts for the amazing speed of the speech production process by
assuming it is largely automatic. Only the message generation and monitoring activities are controlled processes, that is, normally requiring the speaker’s ongoing
attention, whereas grammatical and phonological encoding, articulation, and lexical access are usually automatic – the exceptions being, for instance, the retrieval of
infrequent words (cf. Poulisse 1997). here is where the difference between the native speaker and the l2 learner is crucial. Whereas the native speaker effortlessly
generates fluent speech, the learner proceeds gradually from painfully slow retrieval of lexical items towards ever more complete grammatical encoding in an increasingly automatic way. The relevance of automaticity versus (attended) control in
speech production is a fruitful area for future research.
This is then, in its basic outline, levelt’s model for language generation in the
mature native speaker. What about l2 learners? Summing up, as they develop their
interlanguage, learners need to
(14) a. build up the lexical store, and include in it fully mature lexical items; that is, not
only increasing the number of words with their meanings and sound forms, but
also storing richer lemmas with their relevant diacritic features (semantic, grammatical and formal, as well as categorical and combinatorial) and specific values;
9 Even for native speakers processing less preferred options involves more time, however
minimally. E.g., Weyerts et al. (2002) report that german native speakers prefer SVO
orders to SOV orders and account for this in terms of the relative higher processing cost
of latter as against the former.
Camilla Bettoni and Bruno Di Biase
b. learn to retrieve and encode these lemmas functionally (lFg’s f-structure) and
positionally (lFg’s c-structure) (cf. § 2.2 below); and
c. automatise increasing numbers of retrieval and encoding processing components
for fluent speaking, so as to free up greater attentional resources for semantic and
discourse-pragmatic processing.
The sequence in which learners gradually learn to enrich their lexical storage, activate more grammatical procedures, bottom up, and automatise them is precisely
PT’s domain (cf. § 3). in 1998 Pienemann’s main concern was with the learner’s
progressive activation of grammatical procedures in the formulator. This allowed
for PT’s explanation of the development of obligatory morphological and morphosyntactic structures. Then in 2005 Pienemann, Di Biase & Kawaguchi began to
focus more sharply on the role played by (a) the conceptualiser in preparing its output with a variety of discourse-pragmatic perspectives, and (b) the lexicon in the
grammaticalisation process. Bringing into PT this element of choice the speaker
may have in selecting alternative discourse-pragmatic perspectives and alternative
lexical items allows for new explanations regarding optional syntactic structures. in
this volume the structural consequences of the speaker’s choices will be taken a step
further (cf. our own formulation of the Prominence hypothesis and the lexical
mapping hypothesis in § 4.2.1 and § 4.2.2 respectively). With regard to automatisation, very little work has so far been done apart from Pienemann (2002) and
Kawaguchi & Di Biase (2012). This is an important line of research worth pursuing within PT.
But before describing PT’s developmental stages based on implicational procedural skills identified in levelt’s model, we need to introduce some basic notions
of lFg, and integrate them – as far as possible – within this model.
2.2. Lexical Functional Grammar and PT
lFg10 is committed to the interface between linguistic knowledge and language
processing, and is therefore designed to account for linguistic knowledge in a way
that is compatible with the architecture of the language processor (Kaplan &
Bresnan 1982: 177). This must be the sort of commitment that prompted levelt
to use Bresnan’s lexically-based grammatical theory for his model’s lexicalist approach – in the sense that, as we have seen, levelt assumes that lexical selection activates syntax, that is, it drives the procedures for building the syntactic frame for the
current utterance (cf. Kormos 2006: 9-10). So, in the speaker’s mental lexicon each
10 Cf. asudeh & Toivonen (2010) for a concise, up-to-date presentation of lFg.
1. Processability Theory: theoretical bases and universal schedules
lexical entry is associated not only with meanings and forms, but also with the full
set of their syntactic information.
in addition to its lexicalist assumption, another important principle of lFg is
its clear distinction between levels of linguistic representation that a formal model
requires in order to describe the complex architecture of natural language adequately. Since the sentence is an expression of several types of linguistic information
(semantic, pragmatic, syntactic, phonetic/phonological, etc.), there can be several
theoretically distinct structures of formal representation: semantic structure, information structure, and phonological structure, as well as the three syntactic levels:
argument structure, functional structure and constituent structure. all these levels
exist simultaneously and in parallel,11 each with their own distinct grammatical
module, with characteristic primitives and formal representation. lFg’s architecture “postulates a number of simple data structures with mappings defining the
relationship between the structures […] (it) is thus typically referred to as a Parallel
Projection Architecture” (asudeh & Toivonen 2010: 426, original italics). We will
concentrate here on the three syntactic structures: a-structure, f-structure, and cstructure. lFg work on the other levels is more recent and not yet applied to PT,
so it will not be further mentioned here except for some rudimentary information
structure, which is important for our Prominence hypothesis (cf. § 4.2.1 below).12
We first consider the syntactic structures separately, with their specific properties
that make each level of representation different and independent from each other,
and then we show how they are linked, or mapped, onto one another.
Beginning with a-structure, this level of representation determines many of
the basic properties of the sentence in which a predicate occurs – the predicate
being the word, typically a V, which names the action, event or state described by
the sentence. The a-structure of a predicate encodes information about the number and type of arguments selected by that predicate, as shown in (15). arguments
are thus assigned lexically through the meaning of the V.
(15) run <agent>
eat <agent, patient>
love <experiencer, stimulus>
give <agent, theme, recipient>
11 a good discussion and motivation for this architecture may be found in Falk (2001).
12 Further lFg work in this direction might help PT’s development, especially concerning i-structure. Cf., e.g., the debt to Choi (1999, 2001) for the Prominence
hypothesis, and to mycock (2007) and Dalrymple & Nikolaeva (2011) for chh. 7 and
8, this volume.
Camilla Bettoni and Bruno Di Biase
Because argumenthood is a semantic concept, it is not always as easy to determine the arguments of a predicate as one might expect, and many label sets have
been proposed. Suffice it to say here that lFg generally follows Jackendoff
(1972), and others, in defining a hierarchy among them which is formally based
on two very broad aspects of the way we conceptualise the meaning of Vs: one
based on action, the other on space. in the action conceptualisation, an agent has
primacy over a beneficiary, because when both are present, the beneficiary is affected by something the agent does. in the space conceptualisation, the instigator has
primacy over the theme, which in turn is more prominent than the path, etc. By
stipulating also that the action conceptualisation takes over the spatial one, the
thematic hierarchy in (16) is derived, typologically validated by Keenan &
Comrie (1977) and hopper & Thompson (1980), among others, and used by
Bresnan (2001: 307):13
(16) agent > beneficiary > experiencer/goal > instrument > patient/theme > locative
moving on to f-structure, this level of linguistic representation includes, for every
sentence, all the grammatical information needed to interpret the sentence semantically. it consists of two types of information about the syntactic elements (i.e.,
words and phrases) of a sentence:
• information about grammatical relationships between them; and
• information about their grammatical properties (i.e., features).
That is, in f-structure, abstract gFs and diacritic features attempt to capture universal syntactic principles that may vary crosslinguistically at other levels of representation. We will now have a closer look at these two types of information.
in f-structure encoding, let us look first at gFs, which lFg, uniquely among
grammatical theories, considers as primitives of the theory.14 gFs are cross-classified in several ways. The most basic ones are the argument functions, which express
the arguments of predicates. indeed, they may be directly selected, governed, by the
predicate. They are: SUBJ (subject), OBJ (object), OBJθ(secondary object), the
OBlθ (oblique) family of functions, COmP and XCOmP (complement). among
them a fundamental distinction is made between core functions, which are SUBJ,
13 Because a thematic hierarchy is not a primitive construct, other rankings have been proposed – all of which, however, almost without exception, rank agent highest. For a
discussion of criteria used in constructing a thematic hierarchy and ranking semantic
roles, cf. levin & rappaport hovav (2005: ch. 6).
14 For basic definition and examples of lFg’s gFs, cf. asudeh & Toivonen (2010: 432).
1. Processability Theory: theoretical bases and universal schedules
OBJ and OBJθ, and noncore functions, which are OBlθ, COmP and XCOmP
(cf. also (21) below). Core functions are associated with the central participants of
the eventuality expressed by V, and are usually distinguished formally from noncore functions. in English, for example, core arguments have canonical c-structure
positions which are occupied only by NPs and DPs; noncore functions are generally expressed by other c-structure categories, such as OBls by PPs. For example,
in the sentence in (17a), Consuelo and her sentiments, respectively SUBJ and OBJ,
are core gFs subcategorised by V, associated with NPs in c-structure, and obligatorily present in the sentence. On the other hand, to Pablo, an OBlθ, is a noncore gF associated with PP in c-structure, and optional, as can be seen from the fact
that in (17b) it can be left out without affecting grammaticality.
(17) a. Consuelo expressed her sentiments to Pablo quite nicely in writing
b. Consuelo expressed her sentiments quite nicely in writing
argument functions, like the thematic roles of a-structure, are also hierarchical:
SUBJ > OBJ > OBJθ, etc. (Keenan & Comrie 1977; Bresnan 2001). The hierarchical organisation of these two levels of structures (i.e., a-structure and f-structure) is relevant from a processing point of view, as we shall see later.
all gFs other than argument functions are not specifically selected by the predicate but occur freely, subject to other constraints of the theory (asudeh &
Toivonen 2010: 431). One such nonargument function is aDJ (adjunct). Whereas
argument functions allow only single instances (e.g., there can be only one SUBJ
per sentence), aDJ allows multiple instances. For example, in the sentences in (17)
there are two aDJs (quite nicely and in writing).
all gFs mentioned so far, whether argument or nonargument ones, represent
the clause-internal aspect of syntactic organisation. however, gFs can also relate to
the wider discourse. So, as a secondary function, a syntactic element can also relate to the role its clause may have in the wider discourse structure. These secondary syntactic functions are discourse functions (DFs) or overlay functions. They are
three: TOP (topic), relating the topic of the discourse, mainly old or shared information; FOC (focus), expressing new information; and SUBJ, which is the only
DF that, being also an argument function, is also a governable gF (i.e., selected
directly by the predicate). SUBJ is often identified in many languages as the default
discourse topic.15 The TOP and FOC functions are not selected by the predicate
and map indirectly to f-structure in the sense that they must be co-referential with,
or anaphorically linked to, a nondiscourse gF. For example, in (17a-b) Consuelo is
SUBJ (and default TOP), whereas in (18) the same propositional content topica-
15 For a discussion of subjecthood within the lFg framework, cf. Falk (2006).
Camilla Bettoni and Bruno Di Biase
lises in writing, which is TOP and aDJ; and in Bresnan’s (2001: 97) example in
(19b), the preposed NP Rosie is both FOC and OBJ of its sentence, with FOC relating this sentence to a previous element in (19a):
(18) in writing Consuelo expressed her sentiments quite nicely
(19) a. what did you name your cat?
b. rosie i named her
Note that DFs are not part of discourse representation, any more than nonDFs
are part of lexical semantics, or any more than the notions of SUBJ and OBJ, or
‘nominative’ and ‘accusative’, need references outside the sentence in which they
operate. DFs are somewhat similar to pronouns, for instance, in so far as they
may have an argumental function in the sentence, but refer anaphorically to an
entity or referent mentioned previously. it may be worth reiterating here that, in
the lFg framework, DFs are overlay functions strictly within the sentence, and
that by default (i.e., in the absence of other gFs specifically marked as TOP),
TOP is overlaid to SUBJ, which is both a gF and a DF. These notions may become clearer in the context of levelt’s model. as we have already mentioned in §
2.1, there are seductive correspondences between thematic and discourse roles on
the one hand, and gFs on the other. That is, agents tend to be SUBJs, and elements expressing topical information tend to appear early in the sentence and
have great affinity with SUBJ, a function that allows them to lead in the utterance itself (Bock & levelt 1994: 964; 365). hence, for example, the sentences
below in (25)-(27) need not annotate TOP formally. On the other hand, in the
russian sentence in (28) the overlay of TOP to OBl (as well as of FOC to SUBJ)
must be formalised. Conversely, because TOP is an overlay DF function, it will
not be expressed if there is no nonDF to overlay it to. So, for example, the italian
sentence in (20) has neither TOP nor FOC because neither SUBJ nor OBJ is
expressed syntactically – that is, a head-marking language such as italian can
mark them both on V by morphological means, which are part of the lexical
entry (i.e., part of the word, rather than the sentence).
(20) lo bevo raramente
[‘(i) it drink rarely’ = ‘i drink it rarely’]
The reader would thus appreciate that in lFg the DFs TOP and FOC, which
have a formal definition in the grammar and are subject to explicit constraints,
are quite different from the ‘topic’ and ‘focus’ notions in other frameworks that
may treat them ambiguously as discourse (rather than exclusively sentence) elements.
1. Processability Theory: theoretical bases and universal schedules
The lFg formalisation of DFs plays an important role in solving the functional uncertainty of the clause-initial phrase. in (17), for example, Consuelo is simultaneously TOP and SUBJ by default, so there is no need to mark TOP redundantly because SUBJ is already a DF (as well as a gF). On the other hand, in (18) the
phrase in writing is not SUBJ but an aDJ formally marked as TOP. in this particular case, functional uncertainty is constrained in the interpretation of the sentence because a PP in the c-structure clearly points to an aDJ (or at least a noncore
OBl). Uncertainty is more likely to arise, for example, with the topicalisation of a
core argument such as OBJ, where TOPOBJ may appear early in the sentence as a
bare NP in c-structure. For an italian example, we refer to (19) in chapter 3; in that
sentence, because the first element in c-structure bears both the DF as TOP and
the nonDF as OBJ formally in f-structure, lFg’s Extended Coherence Condition
is satified16 (Bresnan 2001: 69).
There is a futher crucial clarification we wish to make in order to understand
what DFs are in lFg, and hence in PT. On the one hand, TOP and FOC are universal gFs; on the other, their expression varies crosslinguistically.17 When a declarative sentence expresses TOP, more configurational languages, such as English,
and head-marking languages, such as italian and Spanish, tend to mark it, by
default, by placing it in the most prominent position in c-structure, namely the
first in the sentence, and may then place FOC after TOP. and this is how TOP is
operationalised for these languages in this volume.18 however, besides position in
c-structure, other types of languages may prefer to express TOP by morphological,
lexical and, notably, prosodic means, or by a combination of means. For instance,
16 Completeness and coherence are general well-formedness conditions on f-structure
(Bresnan 2001: 63-69). Completeness requires that every gF designated by a PrED be
present in the f-structure of that PrED. The Extended Coherent Condition requires
that all syntactic functions be integrated appropriately into the f-structure. This means
that a TOP or FOC function must be integrated either by identification with an integrated function (in the case of SUBJ or aDJ which have their own PrED) or by anaphoric linking to an integrated function (in the case of argument functions other than
SUBJ, such as OBJ and OBl).
17 in lFg’s work, DFs have been investigated linguistically by many scholars (e.g.,
Bresnan & mchombo 1987; Kroeger 1993), and formally implemented as the mapping of c-structure onto f-structure by Bresnan (2001), Dalrymple (2001), and Falk
(2001). They are also sketched in levelt (1989).
18 in italian, for example, TOP may be encoded also in nondefault positions (i.e., noninitial) in c-structure (e.g., li ha mangiati Pierino i fichi, ‘it was Pierino who ate the figs’).
in this case, however, sentences are highly marked pragmatically and display specific
prosodic contours. Since no learner considered in this volume produces any of them,
they will not be further mentioned.
Camilla Bettoni and Bruno Di Biase
Japanese and Korean express TOP primarily by morphological means, namely the
–wa marker and the –un marker respectively (for Japanese, cf. ch. 4). in interrogative sentences, which must express FOC, this DF can be expressed by prosodic,
lexical (e.g., wh- words), syntactic and/or morphological means. more about FOC
will be said when questions are treated in language-specific chapters because other
important issues are at play here besides configurationality and head/dependent
marking, such as whether questions are polar or constituent ones, and whether the
latter are fronted or in situ in the specific language. as the reader will appreciate
over the remainder of this chapter and the following ones, DFs play a key role in
our understanding of the learner’s syntactic development when they are used together with levelt’s processing principles, such as the attribution of prominence to an
argument function, or to an aDJ – cf. our Prominence hypothesis in § 4.2.1,
in sum, (21) shows how gFs in lFg are subdivided into two major dichotomies: argument and nonargument functions on the one hand, and discourse and
nondiscourse functions on the other.
(21) Grammatical functions and their subdivisions (after Bresnan 2001: 96-97 and Falk
2001: 60)
let us now turn to the second type of grammatical information needed to interpret the sentence semantically, that is, the information conveyed by grammatical
properties (or features). as we have seen earlier, these properties are part of lexical
entries, and include diacritic features such as number, person, gender, definiteness,
case, and tense, which all have their own values: in English, for example, singular
and plural for number (e.g., cherry, cherries); first, second, etc. for person (e.g., I,
you); masculine and feminine, etc. for gender (e.g., gentleman/he, lady/she); definite and indefinite for definiteness (e.g., the banana, a banana); nominative, accusative, etc. for case (e.g., he, him): and present, past, etc. for tense (e.g., sing, sang).
Thus in lFg, grammatical information in f-structure is represented by a set
of attribute-value pairs; that is, given a particular f-structure, each attribute is
always assigned a specific value. There are three types of values: (a) atomic symbols,
1. Processability Theory: theoretical bases and universal schedules
e.g., Sg for singular; (b) semantic forms, e.g., love <x, y>, which stands for a kind
of activity involving two arguments; and (c) f-structures, which themselves consist
of attribute-value pairs. The f-structure for the sentence lions live in the forest is illustrated in (22).
(22) F-structure for lions live in the forest (Bresnan 2001: 46)19
Turning now to c-structure, this is the overt expression of the functions and features that make up a syntactic expression (Falk 2001: ch. 2). it encodes three types of
information: (a) word order, (b) constituent boundaries, and (c) the syntactic category of each word and constituent in the sentence – that is, whether a word is a
noun, a verb, an adjective, etc., and whether a phrase is NP, VP, aP, etc. it is the
level of representation of phrase-structure trees.
in contrast to f-structure, which encodes the invariant (universal) aspects of
grammar, c-structure encodes properties that vary a great deal across languages. in
this regard, we follow Bresnan (2001) in comparing, for example, English with
Warlpiri, an aboriginal language spoken in northern australia. in English, c-structure is strictly organised, both linearly and hierarchically. a sentence can thus be
made up of identifiable constituents, such as NP or VP, which are placed in specific positions; and gFs are encoded in c-structure configurations, with SUBJ outside of the VP and OBJ inside. languages such as English are called configurational
languages. On the other hand, in Warlpiri gFs are not encoded in c-structure; cstructure is flat, and all arguments are sisters of V. Thus word order is free, “so long
as the auxiliary tense marker occurs in the second position” (Bresnan 2001: 6).
languages such as Warlpiri are called nonconfigurational languages. an English
phrase structure (hierarchical and endocentric) and a corresponding Warlpiri one
19 The internal structure of PrEDS is omitted here, as in the original.
Camilla Bettoni and Bruno Di Biase
(flat and lexocentric) are illustrated in (23)-(24) respectively.20 This alternative
mode of c-structure organisation, lexocentricity, “associates syntactic functions
directly with features borne by words rather than with the configurational relations
of phrases in syntax” (Bresnan 2001: 109).
(23) Phrase structure of the English sentence the two small children are chasing that dog
(Bresnan 2001: 5)
(24) Phrase structure of the Warlpiri sentence glossed in English as the two small children are
chasing that dog (Bresnan 2001: 6)
20 The c-structure representations in (23)-(24), as well as others in (25)-(28) in this chapter, keep formalisation to a minimum in order to increase comprehension of the gist of
our general presentation for readers who may be unfamiliar with the fine details of lFg
annotations. Needless to say, we are aware that there is a great deal of variation in cstructure organisation across languages, and that in many of them what is here marked
generally as a sentence (S) – as it is in Bresnan (2001) as a matter of fact – corresponds
to iP. Dealing with specific languages, the following chapters will provide more precise
and updated lFg formalisation.
1. Processability Theory: theoretical bases and universal schedules
in the Warlpiri sentence in (24), the actual word order is ‘two-small are chasing
that two-children dog’, with the NPSUBJ split by the V complex ‘are chasing’ and
‘that’, and with ‘that’ referring to ‘dog’ rather than to the adjacent ‘two children’.
So, it is clear that (in Warlpiri, as in other nonconfigurational languages) the
coherence of a conceptual unit is indicated by means of word shapes rather than
word groups. Noncontiguous words that form a conceptual unit must share the
same formal endings marking agreements. indeed, the richness of the inflectional endings marking relationships between words and groups allows for great
permutation of words and phrases in the sentence according to the speaker’s
discourse-pragmatic needs. Such ‘sharing’ of inflectional features is the basis for
the exchange of information that allows for feature unification. in terms of
levelt’s model for language production, the greater the distance (in terms of syntactic levels21) between the words needing feature unification, the higher the
cognitive cost of unifying them.
Typological variation between configurational and nonconfigurational languages sets up “competition between words and phrases expressing the same fstructure information” (Bresnan 2001: 101-102). This means that morphology
and syntax play complementary roles, in the sense that morphology-rich languages
show preference for lexical over syntactic expression for grammatical encoding, and
vice versa: morphology-poor languages show preference for syntactic over lexical
expression. however, as Bresnan (2001: 132) is quick to remark, along the typological continuum from strictly configurational to strictly nonconfigurational languages, natural languages may freely mix modes of organisation.
So far we have presented a-structure, f-structure and c-structure separately.
This is possible, because each of these levels of representation is independent from
the others, in the sense that they exist simultaneously and none is derived from
any other. however, because each provides only a partial description of a sentence, it is important to specify the correspondence, or mapping, between the elements of these parallel structures. according to lFg, the main problem of a syntactic theory is to “characterise the mapping between semantic predicate-argument relationships and surface word- and phrase-configurations by which they are
expressed” (Kaplan & Bresnan 1982: 174). Thus lFg grammatical formalism is
essentially based on the correspondence architecture with which a sentence maps
a- and c-structures onto the grammatical relations and properties in f-structure.
Because each structure has its own primitives and hierarchy, the mapping between them can align in more than one way (Sells 2001), both across languages and
21 as asudeh & Toivonen (2011) point out, even such extreme nonconfigurationality as
found in Warlpiri appears to preserve some kind of ordering principle; e.g., in this language, aUX must be in second position.
Camilla Bettoni and Bruno Di Biase
within a specific language. For example, on the one hand, the default correspondences between the SUBJ function and the arguments can vary typologically
along several dimensions; that is, in active languages like English, SUBJ maps by
default onto the semantically most prominent role in a-structure; in accusative
languages like Japanese, onto the argument in control of the eventuality; or in
ergative languages like Dyirbal, onto the argument most affected by the eventuality. What all these possibilities have in common is the hierarchical prominence of
the SUBJ argument on the selected dimension compared to other arguments
(Bresnan 2001: 95). On the other hand, the SUBJ function takes no single universal form. Expression of SUBJ includes the NP in a certain phrase-structure
configuration, as in configurational languages like English (cf. ch. 2); verbal
inflection morphology, as in head-marking nonconfigurational languages like
italian (cf. ch. 3); and nominals bearing a specific case, as in dependent-marking
nonconfigurational languages like Japanese or russian and Serbian (cf. chh. 4, 5
and 6 respectively).
The mapping among the three structures is unmarked (or default) when the
most prominent thematic role (i.e., agent) is encoded in the highest available gF
(i.e., SUBJ) and occupies the most prominent position in c-structure, that is, the
first position (Choi 2001). levelt (1989: 266) calls this “congruent encoding”. an
illustration of such mapping convergence is provided in (25) for the sentence
(25) Canonical correspondences of a- and c-structures onto f-structure for the sentence
romeo kisses Juliet
1. Processability Theory: theoretical bases and universal schedules
Romeo kisses Juliet, where the most prominent argument in a-stucture, the agent,
and the most prominent constituent in c-structure (the one in first position), both
map onto the most prominent function in f-structure (the SUBJ). less prominent
thematic roles, if required, link onto less prominent functions in less prominent
however, correspondences among the three structures can vary a great deal,
and be more or less marked. in any language, for a variety of pragmatic reasons, the
same propositional content can be expressed taking different perspectives, as we
have seen in § 2.1. Such perspectives trigger different structural realisations. in
most languages, sentences may vary between declarative and interrogative, between active and passive, and so on. Speakers may also choose to place a constituent
in prominent position by topicalising or focusing it, or they may choose not to do
so. many of these structural choices are devices for directing the hearer’s attention
(levelt 1989), and contribute to the representation of meaning, making communication more effective. however, how and how often these devices are deployed is
Correspondence between the elements of the three levels of representation can
be between arguments and gFs (mapping of a-structure onto f-structure) or between constituents and gFs (mapping of c-structure onto f-structure). The technical, formal details of these linking rules are complex, and well beyond the scope of
this chapter (for their formalisation, cf. Bresnan 2001; Dalrymple 2001). We will
illustrate in (26)-(27) one example for each type of correspondence, bearing in
mind that gFs are here considered the ‘relators’ of c-structure to a-structure.
With regard to mapping a-structure onto f-structure, lFg proposes the
lexical mapping Theory (Bresnan 2001: ch. 14; Dalrymple 2001: ch. 8, Falk
2001: ch. 4), which systematically explains how the conceptual representation of
the thematic roles, represented by a-structure, is mapped onto gFs. in (25) we have
seen an example of how this linking is predictable. But the eventuality described
there can be realised differently, if speakers wish to establish a different perspective
or point of view on the event they intend to communicate. For example, they may
express the matter from the recipient’s point of view, and prioritise Juliet over a
demoted Romeo. This can be done by choosing a different lexical item, such as an
intrinsically ‘exceptional’ V or an alternative ‘nonbasic’ V form. let us clarify these
two expressions, both borrowed from Pinker (1984).
Exceptional Vs are such because they “fail to respect canonical correspondences between thematic roles and grammatical functions” (Pinker (1984: 300). Receive
is such a V. Other common exceptional Vs in English describe a psychological state
or reaction, and include please, delight, bore, and bother. So, the eventuality of
romeo kissing Juliet, as well as canonically by the V kiss in (25), can be expressed
noncanonically by means of the exceptional V receive (a kiss) in (26). On the other
hand, it can be expressed also by a nonbasic V form, in this case the passive V be
Camilla Bettoni and Bruno Di Biase
given, as in (27). like exceptional Vs, also nonbasic V forms fail to respect canonical correspondences between thematic roles and gFs. So what is the difference between them? Without venturing into the intricacies of lexical-semantics, it may be
said that nonbasic V forms require nondefault mapping of a-structure to f-structure
and may be morphologically derived from a more basic form which exhibits default
mapping. For instance, in English the (active) V forms give and scold are morphologically related to the alternative (passive) forms be given and be scolded respectively, or in italian the V forms lavare (‘wash’) and abbracciare (‘embrace’) are morphologically related to their reflexive and reciprocal nonbasic alternatives forms lavarsi
(‘wash oneself’) and abbracciarsi (‘embrace each other’) respectively. Exceptional
verbs also require nondefault mapping, but their alternative forms, if available, are
not morphologically related. For instance, in English receive may be a nondefault
mapping V alternative to the default mapping V give, like in italian insegnare
(‘teach’) may be to imparare (‘learn’). another exceptional V in italian is pentirsi
(‘repent’), but in this case, despite the –si form, no more basic form pentire exists.
in any case, with both exceptional Vs and nonbasic V forms, it is important to
note that – whether or not alternative forms are available, and whether or not alternatives are morphologicaly related – current lFg considers them all as separate lexical V entries which select their own set of arguments. in essence, a phenomenon
such as passivisation “is a lexical relation change, not involving syntactic transformation, in that it can feed lexical processes of derivational morphology” (Bresnan
(2001: 30). in fact, lFg follows a standard assumption in morphology: word formation processes such as derivation, compounding and conversion are morpholexical and not syntactic, and are thus the inputs to those processes. Consequently,
in many languages passivization, causativization, and other relation changing
processes are inputs to lexical processes of derivational morphology such as
nominalization, adjective formation, and compounding. it is not just the verb
forms that are input to the lexical processes, but all their attendant syntactic
effects such as changes in transitivity. (Bresnan 2001: 30, original emphasis –
cf. also pp. 31-36)
This is fully compatible with Bock & levelt’s (1994) and levelt, roelofs & meyer’s
(1999) description of the nature and internal structure of the lexicon and its role
in the formulation of the message in speech processing (cf. § 2.1 above).
going back to our examples in (26) and (27), in either case, nondefaulteness
is due to the fact that it is the recipient, a less prominent thematic role than the
agent, that is linked to SUBJ, the most prominent gF. The difference between
receive in (26) and be given in (27) is that the passive V suppresses the agent, which
then may be optionally mapped as aDJ (Bresnan 2001: 310). Yet notice that, in
either case, with regard to c-structure, both sentences exhibit basic, canonical word
order pattern – which, in the case of English, is SVO.
1. Processability Theory: theoretical bases and universal schedules
(26) Nondefault mapping of a-structure onto f-structure for the sentence Juliet receives a kiss
from romeo
(27) Nondefault mapping of a-structure onto f-structure for the sentence Juliet is given a kiss
by romeo
Camilla Bettoni and Bruno Di Biase
With regards to mapping c-structure onto f-structure, we reiterate that, whereas fstructure functions are largely universal, c-structure configurations are languagespecific. language specificity is manifested in two ways. First, all languages have
their typical (unmarked) canonical word order for core functions. For example,
canonical order is SVO for English and italian, SOV for Japanese, and VSO for
moroccan arabic. Secondly, languages can be placed in different positions along
the typological continuum from configurational to nonconfigurational languages,
as we have just seen with the examples of English and Warlpiri in (23)-(24) at
opposite ends. among the European languages, russian is less configurational than
English, so we will use it to illustrate an example of noncanonical word order. in
the eventuality of romeo giving a rose to Juliet, if russian speakers wish to assign
prominence to Juliet (or better, Džul’etta) over Romeo, besides choosing the exceptional V polučat’ (‘receive’) in a similar way to English, they can also choose to topicalise OBl and focalise SUBJ, as in (28). This involves a marked correspondence
between c- and f-structures, with Juliet realised as TOPOBJ-rECiP preverbally, and
(28) LFG: noncanonical correspondences between c-structure and f-structure for the Russian
sentence Džul’ette daët rozu Romeo [to Juliet gives a rose Romeo]
1. Processability Theory: theoretical bases and universal schedules
Romeo as SUBJ postverbally. Thus word order is noncanonical, and the argument
function SUBJ is no longer associated by default with the DF TOP. On the other
hand, notice that the mapping of agent as SUBJ and recipient as OBl remains
in concluding this brief presentation, we may summarise lFg as a lexically
driven, psychologically plausible grammatical theory which provides an architecture for describing typologically diverse languages in a formal way. lFg provides PT
with two fundamental concepts, ensuring that the different parts of a sentence
actually do fit together:
(29) a. the different syntactic levels – i.e., lexical level, phrasal level or sentence level –
within or across which their elements require unification or merging of diacritic
features and values, a process which iconically reflects performance; and
b. the correspondences among a-, c- and f-structures; or more precisely, the ways in
which elements in a-structure map onto those of f-structure, and elements in cstructure map onto those of f-structure.
The first of these two concepts, already incorporated into levelt’s 1989 model,
provided Pienemann in 1998 with the means to describe the learner’s progress
in the development of obligatory morphological and morphosyntactic structures (cf. § 4.1). Then in 2005, thanks to the second of these lFg’s key concepts,
Pienemann, Di Biase & Kawaguchi began to focus more sharply on optional
syntactic constructions afforded by the alternative ways in which languages
handle mapping their c-structure and their a-structure onto f-structure (cf.
PT’s 2005 Topic hypothesis, which we propose to expand into the
Prominence hypothesis in § 4.2.1, and the lexical mapping hypothesis in §
4.2.2). We should note, in concluding this part, that DFs and lexical mapping
Theory were ‘officially’ formalised in lFg around 2001, that is, after two decades or so of research into many typologically disparate languages, including
australian aboriginal languages. Yet, levelt (1989) had already incorporated
into his processing model of the speaker considerations (a) about the default
way in which, for example, the topical fragment available early from the conceptualiser corresponds to the earliest retrieved lemma(s), which in addition
may receive primacy in grammatical encoding and is commonly generated as
SUBJ NP in sentence-initial position, and (b) about how the speaker’s perspective may direct attention and establish prominence and/or speaker’s perspective otherwise. Furthermore, we should also note that the hierarchies of lFg’s
parallel structures (a-, c- and f-structures) iconically represent cognitive processes, and as such lend themselves to supporting processing interpretations of
Camilla Bettoni and Bruno Di Biase
3. PT’s key concepts
On the basis of its feeder theories, PT describes, explains and predicts the development of morphology and syntax for any typologically different l2 by focusing on
the development of the processing procedures (described by levelt’s model) required for the production of l2 structures (described by lFg).
One of PT’s central claims in Pienemann (1998) is that the sequence with
which learners develop their grammar is not at all arbitrary but follows the time
course with which the grammatical encoding of the lexicon unfolds in levelt’s
model. hence the language processing sequence described for l1 mature speakers
in (11) above foreshadows the developmental progress described for l2 learners in
§ 4 below. The sequenced activation of the processing procedures allows for the
production of language structures, first those structures that do not require any
exchange of information among constituents, later on those that do require it at
the phrasal level, and finally at the sentence and higher levels. Exchange of information is a key concept here.
PT, then, spells out the hypotheses for the developmental sequences of l2
morphology and syntax in learners’ interlanguage. That is, if learners are able to
apply processing procedure x, they will be able to produce morphological or
syntactic structure y using procedure x. Subsequently, if learners are able to use
processing procedure xn to produce structure yn using that xn procedure, it
means that they are also able to use the preceding procedure x and produce
structure y accordingly; that is, the process of building up procedures and corresponding structures is cumulative. implicational hierarchy is a key concept
Based on the activation of implicationally arranged processing procedures,
PT conceives l2 acquisition in terms of sequential progression through a series
of stages. For morphology, progression is operationalised in terms of feature
unification, and measured by the syntactic level on which lie the elements requiring feature unification, that is, exchange of grammatical information, in the
target language (Pienemann 1998). For syntax, it is operationalised in terms of
word order in c-structure, and the degree of flexibility in the mapping of astructure onto f-structure measured by the canonical/unmarked vs noncanonical/marked sets of correspondences among these structures (Pienemann, Di
Biase & Kawaguchi 2005).
implicit in the amount and type of information that needs to be grammatically encoded, and then finally produced, is the cognitive cost of temporarily
storing in the syntactic buffer the bits of partially encoded information that will
be later required for further encoding. Processing cost is a key concept here. The
more information needs to be exchanged, and the longer it needs to be kept
available in the syntactic buffer of working memory, the greater the processing
1. Processability Theory: theoretical bases and universal schedules
cost for the learner.22 Crucially, the more costly the encoding of the structures,
the more difficult the learning, and the later these structures will develop in the
learners’ interlanguage.
however, the cost of grammatical encoding to learners decreases as processing
procedures become automatised through frequent activation. Thus the learners’
progress depends on both the ability to activate new procedures along their implicational sequence, and the automatisation of previously emerged components. in
the meantime, while more advanced procedures are not yet available and earlier
ones not yet automatised, the least costly solution for learners is to resort to default
(or unmarked) structures involving the simplest one-to-one relationship between
form and function, as long as there are words in the lexicon that match the conceptually generated message. Defaultness – (un)markedness, or canonicity – is yet
another key concept in understanding PT.
4. The learner’s progress
having briefly summarised PT’s theoretical bases in §§ 1-2 and key concepts in §
3, we now look at how these shape the learner’s progress along the development
Briefly stated, PT hypothesises that, in converging towards the target language, l2 learners initially encode words and formulas phonologically but not yet
grammatically. That is, they are produced in the order in which they become available in the conceptualiser, as long as the lexicon stores lemmas (in whichever state)
that match (or approximate) the preverbal message. grammatical encoding begins
with words exhibiting minimally categorial features, and conceptual structure mapping onto surface structure canonically according to “prominence hierarchies”
(Choi 2001: 24). This corresponds to levelt’s notion of “congruent grammatical
encoding” (1989: 266), which entails the mapping of the highest role in the thematic hierarchy (the agent) onto the highest gF (the SUBJ) in the most prominent
c-structural position (the first in the clause). grammatical encoding develops further when words acquire more features, and learners become able to process
discourse-pragmatic requirements that require noncanonical correspondences
among the three hierarchies. a case in point is a passive structure, where a thematic role lower in the hierarchy (e.g., the patient) maps onto SUBJ); another is topi-
22 For the native speaker hardly any cost is involved here because this process is proposed
by levelt (1989) to be largely automatic and hence requires no attentional resources (cf.
Poulisse 1997: 204).
Camilla Bettoni and Bruno Di Biase
calisation, when a gF lower than SUBJ (e.g., OBJ) is placed in the most prominent c-structural position (first in the clause).
in (30) we reproduce the components of grammatical processing already illustrated in (4) as fully activated in mature native speakers. here, a grey shadow in
the relevant components shows graphically the early learners’ limited annotations
in the lexicon and their limited ability to assign functions to thematic roles and
constituents, and values to features. learning progresses as the l2 formulator is
built up (cf. De Bot 1992), and these areas become gradually clearer and more actively involved. The task of building up an l2 formulator – or indeed the very existence of a totally different formulator for each language – may vary according to
the learners’ l1, as De Bot (1992) notes. if the two languages are typologically
unrelated the task is more arduous and the learning slower. if they are closely related, it may even be doubtful whether a whole new formulator is needed, and in any
case many of the categories and procedures needed for speech may already be in
place and operative. This would result in faster learning.
(30) Levelt’s Model: components of grammatical processing at an early stage of L2 acquisition (after Bock & Levelt 1994: 946)
although the sequences in the parallel learning of morphology and syntax certainly interface in important ways (cf. § 4.3), we present them separately in §§ 4.1 and
4.2 respectively.23 in the introduction to this part i of the volume we have already
23 There is some evidence across languages for l2 syntax to emerge, consistently, before
morphology. Bonilla (2014) for instance found that syntax emerges before morphology in all her Spanish l2 learners.
1. Processability Theory: theoretical bases and universal schedules
stated that the reason for doing so is that the two developmental sequences, as proposed for PT in Di Biase & Kawaguchi (2012), depend on two different sets of
motivations. On the one hand, we have the original psycholinguistic procedures of
Kempen & hoenkamp (1987) and levelt (1989) modeled in lFg by the mechanisms of feature unification, and adopted for PT in Pienemann (1998). On the
other hand, we have the different kinds of correspondences among the three parallel lFg a-, c-, and f-structures, adopted for PT in Pienemann, Di Biase &
Kawaguchi (2005).
in our presentation of both the universal schedules in this chapter and the language-specific ones in the following chapters, we prefer to avoid the use of numbers for identifying stages, unlike most tables presenting PT sequences in the considerable volume of PT literature since Pienemann (1998). Two reasons guide us
in this decision. First, although conveniently synthetic, numbers are not used consistently, especially whenever authors feel the need to highlight stages within a
stage. This is particularly evident when, in languages such as English, the activation
of phrasal procedure includes the emergence of both NP morphology and VP morphology, whose relative sequence is not clear (cf. ch. 2, § 2). This may not be relevant in languages without VP. Secondly, as well as the sequence for morphological
development, PT has now added two other sequences for syntactic development,
as we shall see in § 4.2 in this chapter. Because the correspondences among these
three parallel lines of development are very much an empirical issue crosslinguistically, and may well turn out to exhibit differences, these should not be pre-empted.
Thus the conveniently synthetic use of numbers for stages may become cumbersome if it were necessary to specify which of the three sequences was actually being
referred to, and whether or not it may correspond to the same-numbered stage
across lines of development.
4.1. Morphological development
PT hypothesises that the availability of increasingly more demanding processing
procedures defines the learners’ progress through a sequence of stages which
depend on the increasingly greater syntactic distance (i.e., in terms of hierarchical
levels) between the linguistic elements requiring exchange of information for their
approriate grammatical production.24 This postulated universal sequence is shown
in (31). Of course what exactly is unified over what syntactic distance must be defined language-specifically. Once the linguistic exponents are hypothesised – that is,
what language-specific structure requires what procedure (belongs to what stage) –
then these hypotheses are tested on spontaneous speech produced by learners of the
24 Franck, Vigliocco & Nicol (2002) managed to separate experimentally the role of syntactic distance vs linear distance in sentence processing.
Camilla Bettoni and Bruno Di Biase
particular language. language-specific sequences and examples will be presented
and discussed in §§ 2.1, chapters 2, 3 and 4 for English, italian and Japanese
respectively. Sequences for other languages (german, russian, Serbian and
Spanish) are also hypothesised, though less extensively, in subsequent chapters.
(31) PT: hierarchy of processing procedures – morphological development (after Pienemann
2005b: 14)
initially, as soon as one or more words are appropriated by the learners, the only
l2 procedure they can activate is lemma access, which is sufficient to allow the production of the word. Unable to activate any further grammatical procedure, with
regard to morphology they produce only formulaic expressions or single words
with no overtly meaningful formal variation. The main reason for this inability is
that, at this early stage, the l2 word or formula is learned as a whole and is not further analyzed. Or, putting it in another way, and paraphrasing Bock & levelt’s
(1994) and levelt, roelofs & meyer’s (1999), the three-level system of the mature
l1 speaker’s lexicon, represented in (2)-(3) above, is reduced to two out of the three
levels: the conceptual level (meaning) and the lexeme level (sound).
at the next stage, learners begin to annotate their lexicon, and develop a
system of lemmas whereby lexical concepts acquire a syntactic category and later its
characteristic diacritic features. This feeds the process of syntacticisation and activates the category procedure to grammatically distinguish, for instance, Vs from
nonVs. The values distinguished earlier by learners are usually those of diacritic features expressing the more transparent conceptual representation with phonologically consistent and possibly frequently occurring forms – such as plural marking
for Ns, or aspect/tense for Vs (in languages that do mark such differences). at this
1. Processability Theory: theoretical bases and universal schedules
category procedure stage, then, overtly meaningful formal variation begins to
emerge but the grammatical information thus annotated does not carry beyond the
word level. in the V lemma, in languages such as English for example, tense and
aspect markers remain contained within the word and may serve more as categorial markers to differentiate Vs from Ns than as actual markers of tense, aspect (cf.
Johnston (1994) for a discussion of the categorial role of –ing in early ESl learners). This means that the relevant diacritic feature is available in the same location
where the morpheme for the marking of aspect (-ing in English, for example) must
occur. Because there is no exchange of information taking place, nothing is stored
for further use somewhere else in the sentence. For those lexical entries which are
not yet fully annotated, the learner will likely use default or citation forms, the least
marked and most available in the input, such as (but not necessarily for all lexemes)
the singular form in languages that mark number, the nominative form in those
that mark case, non-past forms for tense, and so on.
With the next step forward learners reach the phrasal procedure stage. if the
category of the head lemma is N, then the NP procedure can be called in order to
produce an NP. if a determiner or modifier is added, the value of the head plays a
key role. That is, the grammatical information of the head lemma N must be temporarily stored in the NP procedure to be checked against that of the other constituent(s) within the phrase. in order to do this, information must be exchanged
among the words that in the target language require feature unification. For example, in the English phrase a pear in (12) above – here reproduced in (32) with the
addition of arrows showing the path/process of the unification –, DET and N share
the feature number, and require feature unification (represented at the NP node in
the lFg diagram), which in this case concerns the value singular.
(32) An illustration of processing hierarchy for Kim eats a pear showing phrasal and interphrasal procedures
Camilla Bettoni and Bruno Di Biase
at the sentence procedure stage, the phrase needs to be attached to the S(entence)node (i.e., the mother node in the tree structure – although, notice that S is not
necessarily the appropriate mother node in all languages or for all sentence structures; in English, e.g., it is typically iP – cf., e.g., Dalrymple 2001: 53-54, 60-61;
Falk 2001: 40), with the sentence procedure ensuring, in any case, the functional
destination of the NPs associated with the argument roles of the V, such as NPSUBJ
or NPOBJ. here again, the required information relating to a phrase’s values must
be stored until the diacritic feature is assigned to the appropriate place elsewhere
the sentence and the values checked for compatibility. For example, in the English
sentence Kim eats a pear in (32), the NPSUBJ’s number (Sg) and person (3rd) values are kept in the syntactic buffer (short-term memory store) until the appropriate V form, with the bound morpheme –s is retrieved for the V eat. The S-procedure then checks the compatibility of the information coming from N and the V
phrases. in this example, number and person come from the N Kim inside the
NPSUBJ and the V eat, with its person and number values, from inside the VP. as
the arrows indicate, this information travels over different phrasal nodes first, is carried up to the S-node and then checked for compatibility, that is, the person and
number values in the V must be compatible with those of the NPSUBJ, for the
required interphrasal exchange of information.
Further along the morphological developmental path, at the last stage, learners activate the subordinate clause procedure (if the target language requires one at
all). They thus become able to exchange information about the values of relevant
diacritic features between elements in different clauses related to each other by subordination (i.e., the main clause and the dependent one). For example, interclausal
information exchange is required in the subjunctive mood reading of the rather
rare and formal English sentence in (33), where the information exchanged is that
the subordinate clause does not require feature unification between SUBJ and V
(Kim eat), whereas the main clause does (Ann suggests).
(33) The doctor suggests that Kim eat less
in this volume the complex area of subordination will not be addressed. it is an
important research gap in PT, which requires further theoretical elaboration and
focused empirical investigation.
as we can see from the examples above, the exchange of information within
and across constituent boundaries can entail unification involving grammatical
operations of different nature. Unification includes morphological concord (e.g.,
fiori-PL.MAsC profumati-PL.MAsC, ‘scented flowers’, in italian), semantic compatibility (e.g., these books; three cats; many people in English), and grammatical goverment (e.g., has gone vs is going in English; s babuškoj-INsTR., ‘with grandmother’ in
1. Processability Theory: theoretical bases and universal schedules
4.2. syntactic development
For syntax, as for morphology (cf. § 4.1), PT hypothesises a staged development
(Pienemann, Di Biase & Kawaguchi 2005). This comes about as learners gradually develop from default, fixed solutions in linking arguments and constituents to
gFs towards noncanonical, more flexible solutions. That is, they gradually learn to
(a) annotate ever more detailed and specific lexical requirements and (b) grammaticalise discourse and pragmatic information. after the initial stage of single words
and formulas, there are two parallel paths ahead for syntactic development, both of
them leading to the enrichment of f-structure: on the one hand, after learners begin
to map c-structure to f-structure in a canonical way, they start ordering constituents
more freely, as required by discourse and pragmatic reasons. On the other, after
they begin to map a-structure to f-structure in a default way, they then startmapping semantic roles more freely on gFs as required by exceptional lexicon and/or
by discourse and pragmatic reasons.
in other words, one path develops the ability to assign gFs to constituents by
anchoring them at first to a fixed position in the linear order of the clause (SUBJ in
initial position then OBJ), and later on disengaging them from it – thus allowing,
for example, the learner to place a gF other than SUBJ in first position. This path
was originally formulated as the Topic hypothesis in Pienemann, Di Biase &
Kawaguchi (2005). however, two important problems arise from that formulation.
One is the operationalisation of TOP as a gF different from SUBJ, and the other is
the neglect of FOC, the third DF. after mycock’s (2007) work on constituent questions in the lFg framework and its deployment in recent work within the PT framework (e.g., Bettoni & Di Biase 2011; Di Biase & Bettoni 2015), it may be useful to recast the original Topic hypothesis more generally, such that it embraces all
three grammaticalised DFs. This would then include, besides TOP for the development of declaratives, also FOC for the development of interrogatives. in an attempt
to account also for processing as a basic constraint within which speakers – and
naturally enough, learners – construct their utterances from intention to articulation
(to paraphrase levelt), we propose to replace the 2005 Topic hypothesis with our
Prominence hypothesis. in previous drafts of this volume, as well as in Bettoni &
Di Biase (2011) and Di Biase & Bettoni (2015), this latter hypothesis was called the
Discourse Functions hypothesis. its current fine-tuning as Prominence hypothesis
owes a lot to the stimulus provided by discussions with gabriele Pallotti. in changing its label we realise, of course, that the lexical mapping hypothesis may also be
motivated by discourse concerns with prominence. however, as we will see, in the
lexical mapping hypothesis there is more at stake than just prominence. The
Prominence hypothesis deals, then, only with such prominence as achieved by
precedence relations on c-structure alignment, that is, by adding the required discourse information in f-structure and the corresponding c-structure elements.
Camilla Bettoni and Bruno Di Biase
The second path is heralded by the 2005 lexical mapping hypothesis, which
traces the ability to assign gFs to thematic roles first in a default way (e.g., agent
to SUBJ, theme/patient to OBJ) and later in a nondefault way. What motivates
nondefault lexical mapping? Unlike the Prominence hypothesis path, this path is
primarily lexically motivated. So, non-default mapping is lexically required either
by specific verbal predicates, such as intrinsically exceptional Vs or alternative nonbasic V forms, to use Pinker’s (1984) terminology, but recalling that it is not just
the ‘form’ of the V entry to play a role but, primarily, its attendant conceptual-syntactic information (Bresnan 2001: 30), as we have seen in § 2.2 above. Then, of
course, languages may well allow speakers to combine both types of resources; this
would require operating both on word order in c-structure and on argument mapping, which entails a high level of development and automaticity. We will not go
into this presently.
Some readers may have concluded by now that this treatment of syntactic
development along the two schedules based on the Prominence hypothesis and
the lexical mapping hypothesis makes one of the hypotheses in Pienemann, Di
Biase & Kawaguchi’s (2005) extension, namely, the Unmarked alignment
hypothesis, redundant. This is, indeed, a desirable result in theory construction
because it is more parsimonious to put forward fewer hypotheses to cover the same
(or a greater) area. Furthermore, as we said in the introduction to this volume, the
Unmarked alignment hypothesis actually conflates c- to f- and a- to f-structure
mapping at the first grammaticalised stage following the single-word and formulaic one. This, however, may be misleading conceptually because – as Conroy
(2007) remarks – it assumes that canonical word order in c-structure necessarily
entails default mapping of thematic roles onto gFs, thus conferring an unintended
universality to that particular convergence (i.e, agent-SUBJ mapping in first position, followed by the theme/patient-OBJ mapping). On the other hand, the
Prominence hypothesis and the lexical mapping hypothesis account for their
own respective paths independently from each other from the very beginning of
the grammaticalisation process.
in the next two sections, we spell out the two hypotheses in further detail and
illustrate the two developmental sequences respectively in (34)-(35) and (42)-(43).
Then in § 4.3 we discuss the interfaces between them and with morphological
4.2.1. The Prominence Hypothesis
This hypothesis accounts for the staged grammaticalisation of the DFs TOP and
FOC at the sentence level. When the activation of the category procedure takes
place and distinguishes at least between Vs and Ns, Vs begin to acquire their pivotal role in the sentence frame, and participants in the eventuality begin to be
1. Processability Theory: theoretical bases and universal schedules
aligned according to the specific canonical word order of the target language, as
learners follow positive evidence in the input. But, what is canonical order? is there
a universal one?
The notion of canonical order has been hotly debated at least since greenberg
(1963) – see Song (2012) for a historical and theoretical account. in a 402-language sample based on a database of 1063 languages, Tomlin’s (1986) typological
work found two prevalent linear orders for SUBJ, V and OBJ: SVO and SOV, with
no statistically significant difference between the two orders (41.79% vs 44.78%
respectively). The third order was VSO (9.2%) leaving the remaining three possible orders (VOS, OVS, OSV) with less than 5% between them. On the other
hand, working on language genetic groupings (genera) rather than languages as single entities, Dryer (1989) found a preference for SOV, confirmed with an extended database of 1377 languages in the World Atlas of Linguistic structures: 41% SOV
languages, 35% SVO, 7% VSO and only about 3% distributed over the other
three possible order, with the remaining 14% or so languages displaying no dominant word order (Dryer 2013). all these findings nevertheless confirm that in the
great majority of languages SUBJ precedes OBJ. So the typology of word order
would tend to support Keenan’s (1979) subjects Front Principle. and this principle
can be further supported if we consider the order of precedence in the widely
accepted gF hierarchy proposed by Keenan & Comrie (1977): SUBJ > OBJ >
OBJθ > OBl > aDJ (cf. also Choi 1999, Bresnan 2001, and § 2.2 above). in sum,
excluding the position of V, on typological accounts (whether numerical or
genealogical) and hierarchical accounts, SUBJ should be structurally more prominent than non-SUBJ gFs, and, in the absence of other arguments, OBJ will follow as the closest sister of V.
Canonical word order is then the optimal (and prevailing) order with which
each language organises its basic constituents in the c-structure of the prevalent
type of strings (i.e., simple, active, affirmative, declarative, minimally presuppositional, and pragmatically neutral sentences). Thanks to its very predictability, l2
learners can produce canonical word order with minimal processing cost, and
minimally specified functional assignment.25 That is, at an early stage functional
assignment may be purely positional, with a nonhierarchical, flat structure. Not
surprisingly, the l1 or l2 learner’s initial reliance on the canonical word order of
the target language is well attested by a large number of corpus-based longitudinal
and cross-sectional studies of typologically different languages, outside of the PT
framework (e.g., Bever 1970; Sano 1977; Slobin & Bever 1982; Clahsen, meisel
& Pienemann 1983; Pinker 1984; Johnston 1985a; Cook 1993; Sasaki 1998), and
25 readers may also note in this connection that all learners will already have the notion
of SUBJ and OBJ in their l1 though with their own language-specific characteristics.
Camilla Bettoni and Bruno Di Biase
within PT (e.g., Di Biase & Kawaguchi 2002; itani-adams 2009; Kawaguchi
2005, 2011; mansouri 2005; medojević 2014; Pienemann & håkansson 1999;
Yamaguchi 2010; Wang 2010; Wirbatz 2014; Zhang 2005) – where l1-l2 transfer occurs it is developmentally moderated, as Pienemann, Di Biase, Kawaguchi
& håkansson (2005) show across a range of typologically different languages. We
also assume that canonical order is language specific: for example, SVO in English
and italian, and SOV in Japanese.
We also admit that all languages studied so far exhibit the canonical SO order
– and that it would be extremely interesting to test PT’s hypotheses on OS languages, which are in any case very rare. rather than the nature of canonical order itself,
however, what we wish to stress here is that canonical correspondences between linear positions and gFs and DFs are typically learned before noncanonical ones.
The assumption that at the initial syntactic stage learners can produce canonical word order – whose default solution is to make the first NP the SUBJ/TOP,
and the second NP the OBJ/FOC – would imply that the S-procedure is already
operative in their interlanguage. This creates some confusion (cf. also Kempen’s
1998 critical remarks), which Pienemann (1998: 87) tries to disentangle by
labelling this canonical ordering principle a ‘simplified’ S-procedure. Without
entering into detail about procedures, which crucially determine morphological
development (cf. § 4.1), we make the point here that the kind of gFs emerging at
this stage are primarily positionally determined. That is, although learners are now
able to assign gFs to sentence constituents, they are unable to operationalise the
full range of functional assignment exponents apart from position. in other words,
at this first syntactic stage of development, they can produce only one frame: a
canonical string mapping c-structure to f-structure in a fixed correspondence.
Unable to mark gF/DFs by other means than position, the only frame learners can
produce is fixed: SUBJ/TOP precedes OBJ/FOC.
in configurational languages such as English, position may be more or less the
full story, but not so in nonconfigurational languages. The learner’s gradual progress
from functional assignment dependent on position in a fixed frame to functional
assignment independent from position depends on the availability of the necessary
resources (e.g., inflectional morphology). in this regard, it is useful to keep in mind
the temporal sequence in which Bock & levelt (1994) clearly indicate that functional assignment precedes inflection (cf. (29) above), and our interpretation of the
interfacing between syntactic development and morphological development in §
4.3, as well as language specific exemplifications in part ii of the volume.
let us then move on to consider richer frames and more flexible word
orders. in order to enhance their expressiveness, speakers may wish to give prominence to a particular entity in the eventuality they intend to communicate to
their interlocutor. There are various ways in which they can do this, such as by
prosodic, lexical or syntactic means (levelt 1989: 266-67) in various combina-
1. Processability Theory: theoretical bases and universal schedules
tions. We would add here, morphological means (e.g., topic marking in Japanese,
Corean, etc.). The way learners progress through the staged development of syntax – away from the rigidity of canonical word order towards the flexibility of
optional choices allowed by their l2 – was spelled out by Pienemann, Di Biase
& Kawaguchi in their 2005 Topic hypothesis, which we now propose to replace with a more inclusive Prominence hypothesis in (34). This is schematically
summarised in (35).
(34) Prominence Hypothesis. in second language acquisition learners will initially not differentiate between grammatical functions (gFs) and discourse functions (DFs), for
example, between SUBJ and TOP. Differentiation begins when an element such as an
XP, or other lexical material, is added to the canonical string in a position of prominence in c-structure, that is, the first in the sentence. This element may be TOP in
declaratives or FOC in interrogatives leaving, crucially, the canonical string unaltered.
at the next stage, learners will be able to construct noncanonical strings assigning prominence to any constituent in an unequivocal way.26
(35) PT: syntactic development based on the Prominence Hypothesis
Our formulation of this hypothesis represents a significant expansion on its 2005
version, in so far as (a) it incorporates a processing principle explicitly, that is, the
26 This formulation of the hypothesis may not account for all aspects of languages with
in-situ constituent questions, an issue which should be resolved empirically.
Camilla Bettoni and Bruno Di Biase
fact that the sentence-initial position has higher prominence; and (b) it coherently
includes interrogative as well as declarative sentences, and may thus be applied to
a wider typological range of languages.
But what exactly may the learners be learning when attributing prominence?
here we follow Choi (1999: 133), who uses i-structuring constraints operating on
c-structure to identify the contextual (discourse) dimensions of novelty and prominence, and marks each of them by a binary feature, respectively [±NEW] and
[±PrOm], as shown in (36).
(36) Information structuring constraints (i-structure/c-structure correspondence)
These feature-value pairs would enable the identification of SUBJ in its default
initial position as [–New] and [+Prom] preceding a default OBJ [+New] and
[–Prom], whereas a topicalised OBJ would be assigned [+Prom]. So the informational status of the V arguments in a given f-structure, characterised by their
informational feature structure, should predict the optimal alignment of the
arguments in c-structure. readers may recall that the f-structure does specify
functional assignment and, if necessary, the overlaid DFs (TOP, FOC), but fstructure is unordered, hence underspecified in terms of alignment. if the arguments do not differ in informational status, the canonical word order prevails,
but if they do differ, the optimal structure should result from the appropriate
ranking of the constraints. See Choi 1999, 2001 for a treatment of information
structuring within Optimality Theory and lFg. The information status of the
arguments cannot be derived on the basis of their referential (semantic) features alone, but also by the speaker’s perspective. as Dalrymple & Nikolaeva
(2011: 14, original emphases) put it, topicality, for instance, “is not an inherent
property of the referent, and although it correlates with the role played by the
referent in the preceding discourse […] it depends on the speaker’s assessment
of its saliency within a given communicative context”. They then assert that
“[s]ince we view information structure as part of grammar, we treat grammaticality as including pragmatic well-formedness.” an example of the different
information status of the elements in the same eventuality is shown in (37). in
this example by Choi (1999: 46-47), in answer to the where question in (37a),
the FOC on the table is the only element marked with the [+NEW] feature in
(37b-d). But why are there three different answers? The differences are due to
the speaker’s manipulation of prominence through the values of the [±PrOm]
1. Processability Theory: theoretical bases and universal schedules
(37) a.
where did you put the knife?
i put [the knife] [on the TaBlE]
[the knife] i put [on the TaBlE]
[on the TaBlE] i put [the knife]
[the knife]
[on the TABLE]
let us now proceed to illustrate the Prominence hypothesis summarised in (35) in
a general way, mentioning only some representative structures of English. For more
detail and language-specific schedules and examples, the reader is referred to the
chapters in parts ii and iii.
at the very first grammatical stage after the single-word and formula stage, we
include the possibility for learners to also produce questions. Constituent questions
may be produced by placing the wh-word in situ, that is, in the same position as
their declarative equivalents, irrespectively of whether their target language marks
FOC in situ or clause initially. Polar questions are grammatically marked by prosodic means, represented in the table by a prefixed QUE (for the development of
questions, cf. Yip & mathews (2006), and chh. 2 and 8, this volume, for English
and italian examples respectively). Typical additions to declarative sentences at this
stage are time and place specifications (cf. the notion of Stage Topic in ErteschikShir 2007), normally placed to the right of the clause, thus leaving unaltered the
default association of the DF TOP with SUBJ, as in (38).
(38) i eat lunch in park
The next stage of XP canonical word order comes about when, for discourse or
pragmatic reasons, learners become able to add to the canonical string an XP (that
is, a phrase of any category) in the sensitive initial position. Using English to exemplify, the DF assigned to this phrase will be TOP in declaratives or FOC in interrogatives, as in (39a-b).
(39) a. in libya the people not drink beer
b. where you buy this?
This fronted addition plays a crucial role in the learner’s development because it promotes a differentiation between the DFs TOP or FOC and the canonical string.
That is, in declarative sentences the TOP function will now be assigned to this new
Camilla Bettoni and Bruno Di Biase
constituent, rather than to SUBJ by default. The less costly choice in declaratives is
that this new constituent be aDJ rather than an argument of V. and indeed
Pienemann, Di Biase & Kawaguchi (2005: 232) report that empirical studies of the
development of a range of languages identify aDJ among the first nonSUBJ constituents to occur in sentence-initial position – but see chapter 9, this volume, for a
study including both declaratives and questions. What matters most, at this XP
canonical word order stage, is that, whereas earlier on the identification of
SUBJ/TOP by (first) position on c-structure did not allow for the differentiation of
SUBJ and TOP (and indeed there was no need to distinguish between them: if there
is only canonical order, TOP is SUBJ), now the appearance of an XP in the most
prominent position before the canonical string triggers a dislocation of SUBJ from
its canonical first position which interferes with the close connection between SUBJ
and TOP: the TOP function is assigned to the prominent constituent, which is followed by SUBJ and the remaining constituents in canonical order.
although the fronting of a circumstantial element bearing the nonargument
function aDJ (usually expressed with an adverb or a PP) may be common, learners may occasionally introduce in first position yet a different new element
expressed with a bare NP, co-referential with one of the argument functions such
as SUBJ and OBJ, as shown in (40a-b). Such topics, however, at this stage, are
‘external’ (cf. Bresnan 2001: 69) because the remaining part of the sentence (they
read the newspaper) not only displays fixed canonical word order, but it is also complete and fully coherent on its own.27
(40) a. the women, they read the newspaper
b. the newspaper, the women read the newspaper
in constituent questions, the added XP includes the DF FOC. asudeh (2004: 49)
uses the term UDF, unbounded dependency function – FOC or TOP – to generalise across unbounded dependencies. Fronted FOC typically occurs with languages such as English, as shown in (41b, d).
(41) a. Carmen is licking an ice cream-FOC/OBJ in the garden
b. what-FOC/OBJ is Carmen licking in the garden?
c. Carmen is licking an ice cream in the garden-FOC/aDJ
d. where-FOC/aDJ is Carmen licking an icecream?
27 When the topicalised NP is internal to the clause and not in its canonical position (and
hence ambiguous), its functional uncertainty (Bresnan 2001: 67) is resolved differently
in different languages, e.g., by clitic marking on V in italian (cf. ch. 3), and case marking on N in Japanese (cf. ch. 4).
1. Processability Theory: theoretical bases and universal schedules
and, indeed, it is a well-attested fact in English interlanguage that learners can
front their question words at this stage. What they cannot do yet is disrupt the
canonical word order frame.
at the next stage, the crucial step forward is the ability to scramble the elements of the canonical word order. What enables this to happen is that the learner can now assign a gF to each of the arguments autonomously, that is, abstracted
from the fixed position they occupy in the canonical order frame (e.g., by unequivocally identifying SUBJ through agreement with V). This makes argument functions other than SUBJ sufficiently independent as to receive, themselves, the
assignment of TOP or FOC. all this certainly needs the S-procedure to be firmly
in place. if no functional assignment markers (such as those provided by morphological resources, cf. § 4.3) were to signal to listeners that the first NP is not SUBJ,
they might interpret it as SUBJ, and misunderstand the message. Conversely, the
S-procedure being in place does not guarantee that the learner is also able to topicalise OBJ or other argument functions.
in sum, in order to capture the learner’s syntactic progress from the simple and
strict canonical word order frame to a richer and more flexible one, the Prominence
hypothesis predicts three stages: first, word order is canonical, any other additional information may be placed after it. TOP and SUBJ coincide by default, with
minimal (positional) specifications. Second, word order is still canonical, but
because an XP is added as TOP or FOC, DFs and gFs are differentiated. Finally,
word order may be other than canonical, and gFs are further specified.
Empirical evidence for the Prominence hypothesis is reported in several studies across different languages and situations, such as Kawaguchi (2005, 2010) for
Japanese, Di Biase (2007) and Bettoni & Di Biase (2011) for italian, Zhang (2007)
for Chinese, Yamaguchi (2008) for English, and itani-adams (2009) for JapaneseEnglish bilingual acquisition, all of them within the PT framework.
4.2.2. The Lexical Mapping Hypothesis
as we have just seen in dealing with the Prominence hypothesis, the speakers’
assessment of which referent may be salient in the current communicative context
will contribute to the assignment of prominence to a particular argument, whose
alignment may yield a noncanonical word order construction aiming to achieve a
given pragmatic effect. Similar discourse-pragmatic effects may also be achieved by
selecting a verbal lemma or construction that requires nondefault mapping between thematic roles and gFs. importantly, though, topicalisation or focalisation
actually preserve the semantic role hierarchy. So, if agent is mapped on SUBJ, topicalisation does not affect this relationship, as we have seen in the russian sentence
in (27) above, which topicalises OBlgOal. On the other hand, according to lFg
lexical mapping Theory, the default mapping of thematic role onto gFs may be
Camilla Bettoni and Bruno Di Biase
altered when, for example, the patient role (rather than the agent) maps onto
SUBJ, at the same time leaving canonical word order unaffected (i.e., SUBJ may
still be in first position). an important set of verbs marking saliency through noncanonical mappings are the passive verbs, which are present in many languages.
also the so-called exceptional verbs with intrinsic lexical features that require noncanonical mapping as their ‘normal’ (i.e., pragmatically unmarked) realisation are
present in many languages. We have already illustrated both an exceptional verb
and a passive one in (26) and (27) respectively in § 2.2.
With regards to the learner’s progress, the lexical mapping hypothesis was
originally proposed by Pienemann, Di Biase & Kawaguchi (2005: § 3.8) to trace
the way learners progress through the staged development of syntax beyond the
rigidity of canonicity towards a fuller flexibility of the optional choices allowed by
their l2 in assigning gFs to thematic roles. We now formulate an explicit and substantially expanded lexical mapping hypothesis interfacing syntax with both
discourse and semantics in (42)28, summarise it in (43), and then comment on it.
(42) Lexical Mapping Hypothesis. Second language acquirers will initially map the highest
available role in the thematic hierarchy (e.g., agent, experiencer) onto minimally specified SUBJ/TOP. We call this default mapping. Next, they learn to add further arguments mapped onto grammatical functions (gFs) differentiating them from SUBJ
(and OBJ, if present). They may also learn some exceptional verbs at this second stage.
Finally, they learn to impose their own perspective on events, that is, to direct the listener’s attention to a particular thematic role lower in the hierarchy by promoting it
to SUBJ, and defocus the highest role by mapping it onto a gF other than SUBJ, or
suppress it altogether. at this last stage learners may add further role information
regarding causality, benefit, or adversity. They may also add to their lexicon particular
subsets of Vs, such as unaccusatives, as well as further intrinsically exceptional Vs
requiring their own mapping schema. We call this nondefault mapping.
(43) PT: syntactic development based on the Lexical Mapping Hypothesis
28 an earlier formulation of the lexical mapping hypothesis is found in Kawaguchi &
Di Biase (2005).
1. Processability Theory: theoretical bases and universal schedules
To begin with, soon after the stage of single words and formulas, a default choice
for learners is to map the agent or experiencer role as SUBJ (positionally prominent) and the patient or theme, if present, as OBJ. This indicates, minimally, that
SUBJ and OBJ are differentiated.
Then, after the initial default mapping stage and before the final nondefault
mapping stage – differently from the Pienemann, Di Biase & Kawaguchi (2005)
hypothesis, repeated almost unaltered in Pienemann (2011c: 48, fig. 3.13) – we
hypothesise an intermediate stage. at this stage, crucially, learners add a further
argument role such as a goal/beneficiary, instrument, or locative, mapped onto
an OBl gF. Should learners add this further argument without functional marking, confusion could arise as to who does what to whom in the eventuality,
especially if there are two animate or human participants as in (44a). But if they
add it with functional marking, they are able to differentiate between core and
noncore arguments: typically NP vs PP, as in (44b), but also by means of position, as in (44c).
(44) a. *romeo gives a rose Juliet
b. romeo gives a rose to Juliet
c. romeo gives Juliet a rose
at its final stage, the lexical mapping hypothesis is quite broad and undifferentiated, as can be gathered by the wide range of verbs and verbal constructions listed
in the highest row in (38), which includes causatives, benefactives, and unaccusatives, among others. These verbs and constructions are arguably among the most
language-specific. Within this framework, what unites them, both within a language and across languages, is that the nondefaultness of the mapping of a-structure
to f-structure is determined either by the speaker’s lexical selection alone or by lexical selection in conjunction with discourse-pragmatic reasons. By contrast the noncanonicity in the last stage of the Prominence hypothesis is determined exclusively by discourse-pragmatic reasons. What unifies them from a language acquisition
point of view is that, until gFs are unequivocally assigned to thematic roles, canonical word order associated with nondefault thematic mapping can mislead the
hearer, especially when both participants are animate, as in (45a), where the speaker’s intention is to communicate the message in (45b).
(45) a. *[call your mother,] she worries you
b. you worry her / she worries about you
The sequence in which all these exceptional V and nonbasic V forms are learned
within the nondefault mapping stage is a widely open empirical question. it may
well be, for example, that learners acquire at least some exceptional Vs earlier than
Camilla Bettoni and Bruno Di Biase
passive Vs, because in the case of exceptional Vs the motivation for the lexical
mapping hypothesis is triggered exclusively by their intrinsic lexical nature (features) – that is, each particular V must be learned as an exemplar –, whereas in the
case of nonbasic passive V forms learners may find out, eventually, that the passive
construction is generalisable across transitive Vs. however, the assumed sequence
needs to be tested crosslinguistically.
Empirical evidence for the lexical mapping hypothesis comes mainly from
Kawaguchi’s study of Japanese l2 (cf. Kawaguchi 2005, 2007, 2009a, 2010), two
studies on the acquisition of passives in English l2, namely Keatinge & Keßler
(2009) and Wang (2009), and one by Bettoni, Di Biase & Nuzzo (2009) on the
acquisition of postverbal SUBJ in italian l2. less robustly supported by crosslinguistic empirical evidence than the Prominence hypothesis, this lexical mapping
hypothesis needs further testing across different languages and communicative situations.
4.3. Interfacing developmental schedules
as we have seen, with regard to morphology PT claims that progress is constrained
by the syntactic level of the constituents whose features require unification (cf. §
4.1). This means that learners develop first lexical, then phrasal, and finally interphrasal, morphology – we do not deal here with interclausal morphology because,
as mentioned in § 4.1, the complex area of subordination remains a gap in PT, and
requires further theoretical elaboration and focused empirical investigation. With
regard to syntax, progress is determined by the type of correspondences that map
constituents (c-structure) and argument roles (a-structure) onto gFs (cf. § 4.2).
This means that learners develop from a rigid canonical word order to more flexible choices, and from default mapping between thematic roles and gFs to nondefault mapping. how then may the parallel schedules we presented separately in
(31), (35) and (43) interface? When in 2005 Pienemann, Di Biase & Kawaguchi
added their extension to Pienemann’s original PT version, the interface between
“morphosyntax” in 1998 and syntax in 2005 was often mentioned, but not worked out in any detail, nor has this been treated further by Pienemann (2011a, b, c)
more recently. here we can only state the problem in its broader form, because the
details will necessarily be language specific (cf., e.g., the early mention for l2 italian
in Di Biase & Kawaguchi 2002; the treatment of case in l2 russian and l2
Serbian, and of Spanish Differential Object marking, chh. 5, 6 and 7 respectively,
this volume).
We assume that initially, when the activation of the category procedure distinguishes at least Vs from Ns, the general tendency for learners would be to organise
their utterances syntactically by mapping their a-structure roles onto gFs in a
default way, and sequencing them in canonical order in c-structure. however, on
1. Processability Theory: theoretical bases and universal schedules
the one hand, the category procedure (i.e., differentiation in the lexicon) is a necessary resource for both canonical word order and default mapping. On the other
hand, canonical word order and default mapping do not necessarily share the same
constraints. as a matter of fact, under pragmatic pressure, learners may not map
the highest thematic role available (e.g., the agent) onto the most prominent position or the highest gF, and this provides a good reason for us to abandon the 2005
Unmarked alignment hypothesis. For example, under clear pragmatic conditions
urging the learner to topicalise the theme argument rather than the agent argument, learners at this early stage may well produce sentences such as (46a), where
the SUBJThEmE V OBJagENT string would show nondefault mapping with canonical word order – leaving the interlocutor with an ambiguous or uninterpretable
meaning, although in this particular example animacy reduces the conflict. (For
further examples and discussion on this point, cf. ch. 3 on italian, § 3.1, this volume).
(46) a. *the lettuce eat the goat
b. the goat eats the lettuce
at an intermediate level, in order to progress beyond canonical word order, learners need to differentiate between SUBJ and TOP or FOC. This may be achieved
lexically. That is, as long as the lexicon includes temporal or spatial adverbials such
as yesterday and in Libya (cf. (39a) above) to express TOPaDJ in declaratives and
question words such as what to express FOC in interrogatives, learners may not
need to be beyond the category procedure stage. likewise, progress beyond
default mapping may be achieved lexically in many languages by distinguishing
between unmarked OBJ and an OBl marked by a preposition. On the other
hand, the phrasal procedure would be necessary within the VP in languages which
mark gFs by case rather than by PP (cf. the discussion for russian l2, ch. 5, § 3,
this volume). Finally, with regard to the interface between the two syntactic schedules, there seems to be no reason why the addition of an XP before canonical word
order should presuppose the ability to add another argument to the verbal frame,
or vice versa.
as we have just seen, language specificity plays an important role in the interfaces among the schedules at the intermediate level. This is even more evident at
the advanced level. here the key issue is mature, unequivocal functional assignment. The means – or, more precisely, the mix of means in competition with
each orther (cf. Bresnan 2001) – that languages deploy to assign gFs and DFs to
thematic roles and constituents include, typically, morphological means, but also
lexical, syntactic, as well as prosodic means. Prosody is a big player here (cf. levelt
1989), but this area is beyond the current scope of PT. This variety of means
entails that progress beyond the initial stage along the three schedules may vary
Camilla Bettoni and Bruno Di Biase
substantially from language to language, as the three chapters in part ii of this
volume will illustrate.
Even within a single language it may be premature to predict implications
among the schedules. in English, for example, learners will have to reach the sentence procedure stage in order to produce sentences with either noncanonical word
order or nondefault mapping, but with regard to the interfacing between the
Prominence hypothesis and the lexical mapping hypothesis (Kawaguchi & Di
Biase 2012), it is empirical evidence that should show the sequence, if any, in which
structures such as those in (47b-d) will be learned after (47a).
(47) a.
canonical word order
passive verb
exceptional verb
Juliet kisses romeo
romeo Juliet kisses
romeo was kissed [by Juliet]
romeo received a kiss [by Juliet]
Suffice it to mention here three issues affecting the interface between the two
hypotheses. First, what motivates them? We assume that, with regard to the
Prominence hypothesis, canonical vs noncanonical word order is triggered only
by discourse-pragmatic motivation, independently from the feature structure of
the V lemma. On the other hand, with regard to the lexical mapping
hypothesis, default vs nondefault mapping is triggered either exclusively by the
intrinsic lexical nature of V (e.g., in the case of exceptional Vs), or by the lexical features of V in interaction with discourse-pragmatic attribution of TOP or
FOC (e.g., in the case of passives). Secondly, Vs requiring ‘only’ intrinsically
nondefault mapping may be relatively easy to acquire. The English V receive is
one of them. although it requires that the beneficiary, rather than the agent, be
mapped onto SUBJ, the beneficiary is the highest role in its <beneficiary,
theme> a-structure, and the theme (which is hierarchically lower) is regularly
mapped onto OBJ. moreover, word order is canonically SVO. Conversely, some
other Vs are much harder to acquire. The italian V piacere is one of them
(Bettoni, Di Biase & Nuzzo 2009). This may be related to the complexity of
their lexical features, which must be learned V by V. Exceptional Vs, in fact,
involve what Skehan (1998) calls “exemplar-based knowledge”, as opposed to
“rule-based knowledge”, because the behaviour of each of them cannot be rulegenerated and is not generalisable to other Vs. Since exemplar-based knowledge occupies a larger memory storage than rule-based knowledge, it is less economical for online language production. Thirdly, the alternatives in (47), as well
as others not exemplified there, are not all available in all languages, nor do different languages show the same preferences among all available options – as the
awkwardness of some of the English ones in (47) imply.
1. Processability Theory: theoretical bases and universal schedules
5. Methodological issues
Parts ii and iii of this volume will empirically test, language-specifically and on a
wide typological range of l2s, the broad learner’s progress outlined in this chapter
within its universal, processability-based framework. Before proceeding, however,
we need to mention at least three methodological issues of crucial importance to
PT: (a) the type of data required for testing the hypotheses; (b) the type and number of structures which can be diagnostic for a given stage; and (c) the operational
criteria used for determining the acquisition of specific structures.
The first issue concerning the type of data needed to test PT hypotheses is the
least controversial. PT is a theory of Sla which explains the learner’s development
on the bases of speech processing procedures – and can thus, in Ellis’ (1994: 682)
words, be defined as a “transition theory” (i.e., concerned mainly with the dynamic mechanisms responsible for the changes in the development from one system
to the next) rather than a “property” theory (i.e., concerned mainly with describing
static systems of linguistic knowledge). as such, PT relies primarily on spontaneous speech data, produced online, in order to infer developmental patterns, as
already proposed in Pienemann (1998), and – we may add – in most l1 acquisition research, where there might be very little point in trying, for instance, grammaticality judgment methods. indeed, also linguists interested in knowledge acquisition, whether l1 acquisition or biligual l1 acquisition, rely heavily on children’s
performance data to infer knowledge patterns. Thus PT does not use as primary
evidence other types of data such as grammaticality judgements or introspective
comments used for instance by property theories working within Typological
markedness or Universal grammar frameworks (cf., e.g., Doughty 1991; gass
1994; Sorace 2003; White 1989). Such data, usually cross-sectional, is elicited in
order to assess the speaker’s grammatical knowledge (loewen 2009), whereas PT is
more interested in accounting for performance. it would be also rather difficult for
grammaticality judgements (which are comprehension-based) to deal, for instance,
with optional DFs or features that rely on speaker-induced choices in production.
Sure enough, measures from laboratory data such as reaction time, used also in
Pienemann (2002), may be very useful for testing whether some particular elements are, or have become, part of learners’ procedural knowledge, but they may
not be as useful for looking at developmental patterns. inferences about development are ideally drawn from oral longitudinal data, although cross-sectional data
is also often used, recruiting informants over a range of proficiencies. again, ideally, cross-sectional data may best be used to confirm developmental patterns inferred from longitudinal data.
PT’s primary reliance on spontaneous spoken data does not rule out the use
of specifically devised communicative tasks to elicit structures which occur less frequently in natural conversation or emerge only in highly specific contexts, as is
Camilla Bettoni and Bruno Di Biase
often the case for optional structures which are highly marked and depend on the
speaker’s particular pragmatic choices. What PT requires is the assurance that the
sentences uttered by learners are computed online thanks to their speech-procedural skills, at whatever stage of development they may be. For this reason spoken production, which moves at a rate of between two and three words per second (levelt
1989) is privileged in PT over written production, which allows for much longer
processing time.
The second issue concerns the type and number of structures needed as evidence that a learner has reached a given stage. The answer to the number of
structures is fairly straightforward: one is sufficient. in fact, according to PT,
learners are deemed to have reached a certain stage when they show their ability to activate the relevant procedures in their l2, and such evidence can be
obtained by producing a single type of structure, for example, single words and
formulas, canonical order, and so on. The answer to the question of which type
of structures the researcher chooses is more delicate because not all structures
belonging to a stage will be acquired at the same time, and we may well find
that the ‘earliest’ (i.e., easiest) structure at the following stage is acquired before
the ‘last’ (i.e., most difficult) of the previous stage. This is what Pienemann
(1998: 153), following larsen-Freeman & long (1991), calls ‘scouting’ and
‘trailing’. The best choice for a diagnostic structure on an untried language
should fall on a structure that displays possibly the clearest one-to-one relationship between form and function, or the most representative, or default, structure of a stage in a particular schedule, the one with the most transparent conceptual meaning. language specificity plays a crucial role here. let us take morphological development in a language such as russian as an example. russian
requires number, gender and case agreement between noun and adjective within the NP. So, in order to place a learner at the phrasal procedure stage, any of
these three features would do, and there is no theoretical need to limiting or
pre-empting the search – provided, of course, that in a fusional language such
as russian the concept of “factorisation” (Pienemann 1998: 159) is used to
disentangle the three features. however, the number feature is conceptually
more transparent than either gender (which is lexically assigned) or case (which
is structurally and/or lexically assigned). Number is also more common typologically. So it would seem to make sense to pick this feature to begin with. if
we were to choose the gender feature, for example, we may well find that it is
acquired after the emergence of number agreement at the higher sentence procedure stage, which may lead researchers to conclude that PT’s prediction is
wrong. Charters, Jansen & Dao (2011) is a case in point: they notice that
Vietnamese children learning English produce numerical phrasal plural before
categorial plural, and claim that this conflicts with PT’s hierarchy (i.e., categorial before phrasal procedure). But, before venturing into that claim and inval-
1. Processability Theory: theoretical bases and universal schedules
idate PT, they would have to show that when numerical phrasal plural is produced no other categorial marking (e.g., –ing for V-like items) has emerged in
the l2. it could well be that Vietnamese children may have a further barrier
with English plurals but not with categorial marking tout court.
all this does not mean, of course, that researchers should focus only on the
‘earlier’ structures of a stage. it is indeed desirable to gain an understanding of the
range of structures that may belong to a particular stage, but the bottom line is
that the implicationality of the staged development can be tested on just one
structure for each stage in a kind of ‘rapid profile’ (cf. Pienemann & Keßler 2012).
On the other hand, this cannot be the full story for PT. indeed some researchers
(e.g., mansouri & håkansson 2007) working on arabic and Swedish respectively, found that some structures within a particular stage appear to emerge regularly after others, which prompted them to conceptualise an ‘intrastage’ sequencing.
We believe that intrastages are likely characteristic of almost any stage in any language. These may be due to a range of reasons: intrinsic lexical features, such as
with exceptional Vs or N classes in particular languages (cf., e.g., Di Biase 2008
specifies extra features for the production of certain plurals due to N class in
italian), or particular extra steps linked to the processing of certain structures (e.g.,
computing tense and aspect choices in VPs). in this volume we propose to call
these ‘extra’ processing barriers within a stage ‘soft barriers’. This means that, once
the hard barriers across stages have been crossed, language specific soft barriers
within that stage can be identified, and attendant hypotheses may be entertained
and tested. in later chapters language-specific patterns may be found to exemplify this phenomenon. For a puzzling example, see in chapter 2 the late acquisition
of the so-called ‘regular’ past marker –ed in English, which belongs to the categorical stage but whose emergence coincides with much later stages. in other words,
as lardiere (2009) puts it, “(a)ssembling the particular lexical items of a second
language (l2) requires that the learner reconfigure features from the way these are
represented in the first language (l1) into new formal configurations”. Within
PT, we would wish to know the relative position in the sequence of any l2 grammatical structure and, possibly, pinpoint the feature (bundle) that may cause them
to occur when they do.
The third issue is more controversial and concerns the definition of acquisition criteria that allow us to determine operationally whether a given structure
has been acquired or not. Since its inception, PT has used the emergence criterion, that is, “the point in time at which certain skills have, in principle, been
attained or at which certain operations can, in principle, be carried out”
(Pienemann 1998: 138). For a critical review of Pienemann’s theoretical and
empirical construct and a thorough discussion of how the emergence criterion
can be best formulated and operationalised with regard to morphology, we refer
to Pallotti (2007). Suffice it to say here that the authors of the following chap-
Camilla Bettoni and Bruno Di Biase
ters in this volume all follow the emergence criterion. however, generally following Pienemann (1998: 133), their way of operationalising it for morphology differs from the one used in syntax. On the one hand, the obligatory nature of morphological structures allows us to determine the contexts in which a particular
structure occurs, and hence to observe and compute their distribution over the
corpus, that is, whether they are supplied or not supplied in obligatory contexts,
and/or produced where they are not required. On the other hand, this is not possible for syntactic structures, whether default or optional, since they mostly
depend on pragmatic choices. hence only the positive evidence of the number
of their valid instances can be counted.
in general terms, we may say here that, for morphology, we accept as evidence
of emergence the production of two occurrences of a structure, provided there is
formal and lexical variation among the two. This means both morphological variation (e.g., go vs goes, a kind of structural minimal pair) and lexical variation (goes
vs eats, a sort of lexical minimal pair), as Pienemann has always stated – although
Keßler & Pienemann (2011: 95) do not commit themselves to any number, simply saying in their example that “various” Vs are needed for lexical variation, and
“different morphological forms” are needed for “some of these verbs”. This principle of structural and lexical variation is designed to flush out exclusively formulaic
use. When agreement structures involve more words, a greater number of occurrences may be in order, but the design principle remains the same: minimal evidence of structural and lexical variation is required. For syntax, following a long tradition from ZiSa (meisel, Clahsen & Pienemann 1981) to PT, as in Pienemann,
Di Biase & Kawaguchi (2005) we accept one example as evidence of the emergence
of a structure. For further details, we refer to the individual studies reported in the
chapters to follow.
let us then finally describe in (48) how the implicational scales in the subsequent chapters of this volume organise the empirical data for testing PT’s
(48) a. We follow Pienemann’s long tradition of indicating progress vertically by placing
the earlier structures at the bottom row, and then proceeding upwards to the most
advanced ones at the top. Similarly, the learners’ progress in time is shown by proceeding left to right, that is, from t1 to tn in longitudinal studies, and from the
least advanced learner to the most advanced one in cross-sectional studies.
b. The lighter horizontal lines mark the divisions between the stages, thus grouping
together all the structures within each stage. We call these interstage divisions hard
barriers for the learner to negotiate.
c. The numbers of occurrences for morphological structures are entered in the tables
with a plus (+) sign if the structures are supplied in obligatory contexts; with a
1. Processability Theory: theoretical bases and universal schedules
minus (-) sign if not supplied in obligatory context; and with the greater-than (>)
sign if supplied in a context that does not require it; an empty cell marks no context for a structure. as for syntactic structures, their optional nature makes it futile
to establish obligatory contexts. So the relevant cells only mark their presence, and
show either clear numbers or empty cells, unless a structure is attempted but produced in an interlanguage form, in which case the number is preceded by a minus
(-) sign. For example, in the green fish eaten by red fish, a passive structure is attempted, but the auxiliary is not supplied.
d. in order to highlight the implicationality of the learners’ progress, thicker vertical lines mark at which time (longitudinally) or by which learner (cross-sectionally) a particular stage has been reached – in other words, whether the acquisition criteria are not satisfied (to the left), or are satisfied (to the right). These
vertical lines are drawn for the whole stage, that is, regardless of whether only
one or more of the stuctures belonging to that stage have emerged for that stage.
Because – as we have just mentioned above – one contrastive token for syntactic structure (or two lexically contrastive tokens for morphology) is sufficient
evidence for declaring that a stage has been reached, it may well be that negative evidence for a different structure belonging to the same stage appears to the
right of the vertical line. When this is the case, a zigzag vertical line marks at
which time (longitudinally) or by which learner (cross-sectionally) a particular
structure has been acquired. We have called these intrastage divisions soft barriers, which the learner has to negotiate further within the stage. To illustrate this,
we offer the following table:
here stage 1 is reached at t1, stage 2 at t2, stage 3 at t3, and stage 4 at t4. at t2,
however, only one (structure B) of two structures in the same stage is acquired. The
other (structure C) is acquired at t4, that is, after the soft barrier is overcome, and
after structure D of stage 3. in English l2, for example, the –ed morpheme for regular past, which belongs to the category procedure stage, is usually acquired only
when higher stages are reached (cf. ch. 2).
e. Where appropriate, a scalability statistical test (cf. hatch & lazaraton 1991: 204216) is used to show the existence of hierarchical relationships among structures.
Camilla Bettoni and Bruno Di Biase
6. Conclusion
in 1998 Pienemann defined his theory as ‘lexicalist’. Yet, PT’s main concern was
then limited to the procedural activities of levelt’s formulator and did not deal
with the lexicon in sufficient detail except for its diacritic features. Further, 1998
PT was powerful in explaining obligatory morphological variation, but did not tackle syntactic issues of marked word orders or mapping. Thus it could not explain
why structures such as topicalisation or passives emerge in interlanguage much later
than their canonical SVO/SOV or active counterparts respectively. Pienemann, Di
Biase & Kawaguchi’s 2005 extension brought into the scope of PT structural consequences triggered by discourse-pragmatic and lexical requirements, which inevitably interface with syntax. in this chapter, we have proposed an updated PT formalism, and made more explicit connections between PT and Bock & levelt
(1994), as well as levelt, roelof & meyer’s (1999) Theory of lexical access on the
one hand, and Bresnan’s (2001) and other lFg scholars (e.g., Dalrymple, Falk)
formalisation of grammaticalised DFs and lexical mapping Theory on the other.
We have thus coherently broadened the scope of PT to include FOC in addition
to TOP in the syntactic schedule, as well as recast the syntactic hypotheses, and
generally tidied up the schedules resulting from the Prominence hypothesis and
the lexical mapping hypothesis, both of which could not be coherently understood, nor formalised in fact, without the later psycholinguistic and grammar-theoretical works just mentioned.
We have also proposed a more explicit interface between morphology and syntax. The acquisition of obligatory morphology is explained in Pienemann (1998)
by the sequential activation of levelt’s category, phrasal and inter-phrasal procedures modelled in lFg by the formalisation of feature unification. Ultimately, this
also explains the later acquisition of structural alternatives, because their noncanonicity is marked by some morphological or other (nonsyntactic) means in many
languages, including those with scant morphology, such as Chinese, for instance,
where a small number of particle markers become crucial as indicators of mature
functional assignment.29 indeed, it would be interesting for research to look at the
applicability of our Prominence hypothesis to so-called topic-prominent languages (li & Thompson 1981) such as Chinese. Compared to their canonical or
default counterparts, the later acquisition of the noncanonical or nondefault structures accounted for by the Prominence hypothesis and the lexical mapping
29 Prosodic means are of course important. however, in order to deal with them in a formal way, PT researchers will require further work within lFg and Prosodic Phonology
(Nespor & Vogel 2010).
1. Processability Theory: theoretical bases and universal schedules
hypothesis does not by itself determine higher stages in interlanguage; noncanonicity and nondefaultness however are dependent on morphological resources, in
the sense that their fluent deployment implies the automatisation of morphological components.30 So Pienemann’s (1998) PT remains, in any case, foundational.
The crucial question in order to move from ‘descriptive adequacy’ to ‘explanatory adequacy’ is what motivates alternative syntactic structures. according to
levelt’s model, the preverbal output of the conceptualiser guides the order and type
of lexical choices. When learners must compute them online, including their phonological and prosodic shapes, along with discourse-pragmatic information and/or
lemmas with richer feature structures, and integrate them all within an executable
frame for oral production, these extra processing computations add to the cognitive cost for the learner (and natives, though minimally, cf. Weyerts et al. 2002).
however, the cost of effective communication is reduced by automatisation, and if
one can handle the phonology and syntax automatically, then more attention can
be paid to processing semantic, pragmatic, and sociolinguistic levels of communication (cf. Segalowitz 2003). This is a processing area that cries out for integration
in a theory which has processability at its heart.
Part ii of the volume will now illustrate how our view of PT interacts with
language specificity.
30 This point is not uncontroversial and invites focused research.
The developmental path across languages
Camilla Bettoni* and Bruno Di Biase**
*University of Verona, **University of Western Sydney
After focusing on the universality of the theory and on the integration of its 1998
and 2005 strands into a more coherent whole in part I, this volume, part II draws
the consequences of these two foci, and reconceptualises the staging of L2 development with reference to three typologically distant languages covering a good chunk
of typological space between them: English, a configurational language; Italian, a
null-SUBJ head-marking language; and Japanese, a zero-anaphora, dependentmarking language. The latter two languages are all placed towards the less configurational end of the continuum, as shown in (1) below.
The universality of PT “universal schedules” is based on speech processing
procedures, which are cognitive, and hence universal. By that, however, we do not
mean that every language will have the same developmental schedules. Rather, we
mean that the universal schedules can only be interpreted in a language specific
way. Thus, every language has its own schedules reflecting its own typology. This
is why part II of the volume describes the development of three typologically different languages.
There are two principal sources of language specificity that the learner must
acquire (aside from phonological considerations): the lexicon and c-structure.
These are linked via f-structure, which is largely universal, but expressed in a language-specific lexicon and aligned according to language-specific constraints interfacing with discourse-pragmatic preferences. In this regard, there are two important typological distinctions – or rather continuums, because natural languages
may freely mix their modes of organisation (Bresnan 2001: 132). The first continuum – as we have already seen with the two extreme examples of English and
Warlpiri (cf. ch. 1, § 2.2) – is configurationality, which distinguishes between languages expressing GFs (principally the relationship between the verb and its arguments) by position, and those expressing them by morphology. The second important typological continuum relevant to our volume distinguishes between languages marking the relation between the constituents and the head morphologically
on the head (such as Italian, and to a lesser extent English), or on the dependent
(such as Japanese). This characterization as head-marking or dependent-marking,
first introduced by Nichols (1986) for any kind of phrase structure, indicates for
Grammatical development in second languages, 81-84
Camilla Bettoni and Bruno Di Biase
us mainly whether GFs are marked inflectionally on the head element (typically the
V, or the predicate) or on the dependent element (typically the nominal
arguments). For example, a language is head-marking if it overtly marks the SUBJ
function in a clause by means of the agreement of V with its SUBJ; on the other
hand, a language is dependent-marking if it marks the NP argument by casefeature. Some languages may use both agreement and case marking (e.g., Serbian
and Latin), others hardly any (e.g., Chinese). In (1) below we have added the three
languages treated in this part of our volume to the schema introduced by
Nordlinger (1998).
(1) Basic typology of expressing grammatical relations (after Nordlinger 1998: 49)
With regard to configurationality, represented on the horizontal continuum in (1),
we have already shown in chapter 1, (23), how a highly configurational language
like English uses hierarchical phrase structure to encode GFs such as SUBJ and
OBJ. English in fact is one of those languages where OBJ belongs under VP and
is strongly related to V, and ADJ may not be freely interposed between V and OBJ
(unlike Italian or Spanish). On the other hand, SUBJ is outside VP and precedes
V. English SVO word order is fixed, to the extent that, if the NPs before and after
V are swapped, the meaning of the clause changes, as in (2).
(2) a. Jane hits Tarzan
b. Tarzan hits Jane
At the other end of the configurationality continuum, as we have shown in chapter 1, (24), Warlpiri uses morphological case marking on NPs, rather than syntactic phrases, to encode GFs. This type of marking allows for a highly flexible word
order, though a positional point of reference remains even in radically configura-
PART II. The developmental path across languages
tional languages, like Warlpiri where, for instance, AUX must be in second position (cf. Asudeh & Toivonen 2010), thus retaining a certain positional organisational principle.
Like Warlpiri, Italian and Japanese are also nonconfigurational languages,
although less radically so, in so far as they do exhibit a canonical word order (SVO
in Italian and SOV in Japanese), and neither allows elements belonging in the same
NP to be easily separated. Both these languages allow for some flexibility in word
order. However, they differ from each other because they represent a case of headmarking and dependent-marking languages respectively. We illustrate this
difference by looking at morphological encoding of the two core GFs SUBJ and
OBJ by means of agreement marking on V in Italian, the more head-marking language, and of case marking on NPs in Japanese, the more dependent-marking. For
example, in the two Italian sentences in (3), word orders are SVO and OVS; yet
their referential meaning is the same. This is so because, when OBJ topicalisation
disrupts canonical word order, the functions of both NPSUBJ and NPOBJ are
identified morphologically by two inflections of V: one, which marks SUBJ, is
identified by the V morpheme –a, agreeing with postverbal SUBJ; the other, which
marks OBJ, is identified by lo, the ACC clitic marker coreferential with preverbal
TOP (for further details on OBJ topicalisation in Italian L2, cf. chh. 3 and 8, this
(3) a. Desdemona
picchia Otello
Desdemona-3.SG hit-3.SG Otello
[Desdemona hits Otello]
b. Otello
picchia Desdemona
Otello-3.MASC.SG he-3.ACC.MASC.SG hit-3.SG Desdemona-3.SG
[Desdemona Otello hits]
Likewise, in the two Japanese sentences in (4) from Kawaguchi (2008: 96), word
orders are SOV and OSV respectively; yet their propositional meaning does not
change. However, unlike in Italian, this is so because, irrespective of their positions,
the function of NPSUBJ is identified morphologically by the case-marking –ga for
NOM, and the function of NPOBJ by the case-marking –o for ACC.
(4) a. Mari-ga
Mari-NOM Takashi-ACC hit-PAST
[Mari hit Takashi]
b. Takashi-o
Takashi-ACC Mari-NOM
[Mari hit Takashi]
Camilla Bettoni and Bruno Di Biase
In sum, different languages encode GFs by different means. English, a
configurational language, does it mainly through configurationality, even though it
uses some vestiges of head-marking morphology (e.g., 3rd person singular is
marked on V in the present tense). Italian and Japanese, both nonconfigurational
languages, overtly mark GFs mainly through morphology, and reserve positional
options for DFs. PT claims that the learner’s morphological and syntactic development can be predicted by
• interpreting the different means by which a target language specifies its grammatical information, representable by an LFG description; and
• identifying the procedural skills required for a particular linguistic operation, as
indicated by Levelt’s Model.
The developmental hypotheses for English, Italian and Japanese discussed in the
following three chapters are not entirely new in themselves, but to a large extent
their illustration here is. The changes introduced are consistent with our presentation of PT in part I: they are not mere terminological formalities, but – as we have
already mentioned – substantial innovations derived partly from coherently adopting relevant advances in PT’s two source disciplines, and partly from our own and
the authors’ contribution to theory construction and new interpretations of the
results in L2 description work. The language-specific developmental schedules presented have been tested to a large extent, albeit some with more robust empirical
evidence than others. Where evidence is still scant, it will be indicated in order to
identify gaps and promote further work.
Theoretical progress specific to each of the three languages, within a broad
roadmap of how they develop in learners, concerns mainly the following areas: a
revisitation of the morphological schedules focusing on some neglected areas (e.g.,
the role and position of the VP procedure), the treatment of questions, both polar
and constituent, in English L2 (cf. ch. 2); the identification of soft barriers, or steps,
within the stage boundaries, and their explanations for English L2 and Italian L2
(cf. chh. 2 and 3 respectively); and a focus on the Lexical Mapping Hypothesis for
Japanese L2 (cf. ch. 4).
The development of English as a second language
Bruno Di Biase*, Satomi Kawaguchi*
and Yumiko Yamaguchi**
*University of Western Sydney, **Tokai University
1. Introduction
Many researchers have contributed to understanding the development of English
(cf. Finegan 2011 for an overview of this language from a typological point of
view), possibly the most studied L2, both outside PT, as can be gathered from well
known introductions to the field (e.g., Doughty & Long 2003; Ellis 2008; LarsenFreeman & Long 1991; Gass & Selinker 2001; Ortega 2009), and within PT. The
history of English within PT starts with its precursors (e.g., Johnston 1985a, 1997;
Pienemann & Johnston 1987; Pienemann, Johnston & Brindley 1988;
Pienemann & Mackey 1993), who found, by and large, morphological sequences
similar to those found in early SLA research (e.g., Dulay & Burt 1974; Bailey,
Madden & Krashen 1974). However, whereas Krashen (1977), for example, gave
no explanation for the sequences except for postulating that the order is ‘natural’,
PT (i.e., Pienemann 1998) – through ZISA first, and later Levelt and LFG (cf. ch.
1) – added speech processing and principled staging to the ‘nature’ of morphological sequences, and decidedly also grasped the syntactic nettle. All this produced
testable (i.e., falsifiable) developmental hypotheses, as more recent work amply
demonstrates for English (e.g., Sakai 2008; Dyson 2009; Zhang & Widyastuti
2010; Charters, Jansen & Dao 2011; Kawaguchi 2013; Pienemann, Keßler &
Lenzing 2013; Yamaguchi & Kawaguchi 2014).
We will not review here the variously detailed descriptions of L2 English
already offered by these and other authors. We will, instead, ground the theoretical changes proposed in chapter 1, this volume, by tracing the morphological and
syntactic development of this configurational language using a detailed, longitudinal, case study. For the development of mophology, we will note three areas
where our treatment of L2 English differs most from previous PT ones. First, we
re-examine the categorial stage with reference to four structures, that is, –ing as a
verbal marker, past –ed, plural –s, and possessive ‘s, a previously neglected area. We
discuss possessive ‘s in some detail together with possessive determiners and pronouns, but we exclude both the latter from the PT schedule for English. Second,
Grammatical development in second languages, 85-116
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
we place the unification of AUXs (be, have and modals) with lexical V within the
VP in the phrasal stage, unlike Pienemann who since 1998 deals with English VP
only in the context of questions. Third, we adopt for English the concept of ‘soft
barriers’ (cf. ch. 1, § 5) within stage boundaries to deal with intrastage sequences,
such as the very late acquisition of –ed within the category stage.
For the development of syntax, we first explore the proposed Prominence
Hypothesis, and treat declaratives separately from questions (both polar and content). Then, we explore the Lexical Mapping Hypothesis and test, on English for
the first time, the transitional step between default and nondefault mapping proposed in chapter 1, this volume. We attempt to do all this by thoroughly reanalysing the database from Yamaguchi’s (2010, 2013) two-year longitudinal study
of a Japanese L1 primary school-aged child learning English as an L2 naturalistically in Australia.
To make sense of our changes we must refer to Pienemann’s own presentations
of the PT hierarchy for L2 English. Among his several schematic versions, substantially unchanged since 1998 (cf. Pienemann 2011a, b, tab. 4.7), we refer to
Pienemann’s (2005: 24), reproduced in (1).
(1) Pienemann’s (2005: 24) hierarchy as applied to English
Our main point of contention here concerns Pienemann’s apparent assumption
that morphological information and syntactic arrangements depend on the same
processing procedure. In fact his presentation suggests that the same processing
procedure (column 1) and the same information-exchange process (column 2)
apply for both morphology and syntax. So, in stage 4, for instance, the reader
would be justified in assuming that ‘tense agreement’ in the morphology column
calls for an interphrasal procedure as much as ‘Y/N inversion’ or ‘copula inversion’
in the corresponding column for syntax. This assumption may turn out to be correct, but it clouds the issues under observation, because the morphological ‘tense
agreement’ structure refers to V constructions and does not involve the position of
SUBJ on the syntactic alignment, and may hence not require the same procedure
as Y/N or copula inversion, which does involve the position of SUBJ. In fact some
contradiction arises in (1) as to the nature of stage 4: does the VP-procedure actually belong to phrasal procedure or interphrasal procedure? The morphology column would make it ‘phrasal’, whereas the syntax column would make it ‘inter-
2. The development of English as a second language
phrasal’ because it involves the noncanonical position of SUBJ with respect to auxiliaries or copula. In line with chapter 1, this volume, we adopt an analytical
approach that looks at morphological development separately from syntactic development, which helps solve this apparent contradiction. Our study points to a clear
answer: the VP-procedure belongs to the phrasal procedure in the morphological
schedule, as we will see in § 2.
In any case progress in one grammatical area does not guarantee progress in
another. As a matter of fact, recent PT longitudinal studies of English L2 (e.g.,
Yamaguchi 2010, 2013) and of bilingual L1 acquisition involving English and
Chinese (Qi & Di Biase 2005; Qi 2011), and English and Japanese (Itani-Adams
2009) report faster growth in syntax than morphology, sometimes even by two PT
stages. Bonilla (2014) also found syntax to emerge earlier than morphology in all
of her Spanish L2 learners. These are the sorts of reasons that prompt us to present
the implicational hierarchy for morphology and its distribution over syntactic levels (cf. § 2) separately from that of syntax (cf. § 3), as Bettoni and Di Biase argue
in chapter 1, § 3. This does not exclude that phenomena across morphology and
syntax may appear at the same time. We are simply making the point that, as a matter of analytical procedure, it is desirable to separate phenomena that reflect different explanatory principles, leaving their temporal progression to a separate step in
the analysis.
A further concern with Pienemann’s presentations of L2 English, including
recent work (2011a, 2011b: 51-56, tabs 4.1-4.2, 4.4), relates to the fact that morphological development relies on cross-sectional data – chiefly Johnston’s (1985a)
SAMPLE Report for adult ESL data, and Pienemann & Mackey (1993) for adolescent ESL data. We wish to add here significant support for PT’s hierarchy for
morphology with genuinely developmental (i.e., longitudinal) data (cf. § 2).
Our proposed changes to Pienemann’s English schedules are supported by
the evidence afforded by the longitudinal corpus of Yamaguchi’s PhD dissertation
and subsequent publications (2010, 2013). This data, like most longitudinal data,
belongs to one single (child) learner, and provides findings which are hardly generalisable. Yet they are remarkably compatible with those provided by the crosssectional data from both Johnston’s (1985a) adult learners and Pienemann &
Mackey’s (1993) adolescent learners. What Yamaguchi’s longitudinal data contributes, beyond what cross-sectional data possibly can, is to provide critical evidence on the controversial status of some stages (as we have just seen) or particular morphemes, such as plural –s, which straddles uncomfortably between categorial and phrasal stages, as well as an idea of the status and temporal sequence of
morphemes acquired not only across stages but also within a stage, as we shall see.
The empirical data at hand is produced by a Japanese child, codenamed
Kumi, who had been learning English in a naturalistic environment in Australia
since her arrival at age 5;7;15. Kumi was first recorded four weeks after arrival (t1),
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
and then every two weeks for the first 2 months (t2-t5), every two months for the
remainder of the first year (t6-t11), and every three months in the second year (t12t14). Our new analysis of the 14 sessions’s data, presented in the next section, covers approximately two years – a 100 week period, up to age 7;8;15. In these sessions various tasks were performed in the child’s home with native or near-native
speakers of English, including adults and children, who had a close social relationship with her. In order to maintain consistency of data sets across interview sessions, similar tasks were used for each session, including communicative tasks such
as ‘story telling’ tasks, riddles and ‘spot the differences’ tasks, to elicit questions and
declarative sentences. The ‘pictures without words’ storybook was Frog where are
you? (Mayer 1969), often used in linguistic research with children, notably by
Berman & Slobin (1994). In Yamaguchi’s (2010) study this particular task was
used in four occasions over the 100 weeks of data collection. Other tasks (e.g., the
spot-the-differences task) were also used on more than one occasion, but the stimuli were different each time. The collected data consists of 2,151 turns by the
informant with a total of 11,951 word tokens and 3,625 word types (excluding
uhms, ers, and other backchannelling), ranging from a minimum of 176 English
words per session at t1 to 1,783 at t14.
2. Morphological development
In (2) we present PT’s original and well known universal hypotheses1 for morphological development applied to L2 English, with morphological structures distributed hierarchically, and implicationally, over phrasal, interphrasal and interclausal
levels of information exchange (cf. the universal hierarchy in (31), § 4.1, ch. 1,
this volume). Then we follow the learner’s progress stepwise, noting the raw
occurrences in (3a), and their calculated scalability in (3b). However, within this
basic arrangement, in addition to our focus here on the morphological domain to
the exclusion of questions and other syntactic structures, we introduce several
changes. First, possessive determiners and pronouns do not appear in our hypothesis schedule because they belong to the lexicon, that is, they are not generated by
morphological rules. Second, in (2) the order of structures within stages follows
the temporal order in which they emerge, when they do emerge, from Kumi’s longitudinal corpus shown in (3). So perhaps the most obvious difference between
Pienemann’s schedules and ours concerns not so much the choice of structures but
1 Needless to say, ‘universal hierarchy’ in PT never refers to linguistic structures per se,
but to their processing, hence the hierarchy may only be expressed linguistically when
it is applied to a specific language, English in this case.
2. The development of English as a second language
rather the addition of language-specific intrastage sequences within stages that
have a range of exponents (cf. the notions of ‘hard’ and ‘soft’ barriers introduced
in ch. 1, § 5, this volume, and illustrated empirically also for Italian, ch. 3, § 2.2,
and for Japanese, ch. 4, § 2.2). For instance, within the category procedure stage,
we order the English morphemes as follows: –ing emerges before possessive ’s,
which in turn precede plural –s, leaving past –ed last.
(2) Developmental stages hypothesised for L2 English morphology (after Pienemann 1998,
2005, 2011a, b)
(3a) Kumi’s morphological development
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
(3b) Scalability matrix of Kumi’s morphological data2
At the first lemma access stage, the L2 lexicon contains hardly any grammatical
annotations and so learners can activate only the lemma access procedure, and produce morphologically invariant forms. Kumi already produces a narrative of 176
words at t1, hence she is already beyond the one-word/formulaic stage. For this reason, the lemma access stage is not shown in (3a) and is also, naturally, discounted
from the scalability calculations in (3b).
Category procedure stage. At the category procedure stage, learners are hypothesised
to start annotating lemmas with their characteristic diacritics. These annotations
lead the charge towards language-specific differentiation as they are the basic building blocks of the L2-specific grammatical system, starting from the formal differentiation of broad categories such as V-like words from other lexical items. This
categorial stage is the most crowded stage for English: PT hypothesises no less than
four separate morphological structures to emerge, two characterizing nominal
items and two verbal ones, each annotated with its own nature and function, and
likely to be used by the learner in ways which may be different from native use.
These structures are not required to emerge all at the same time (and they don’t, as
we shall see) nor are they all meant to emerge before any morphological structures
of the next stage. As far as PT’s predictions are concerned, it is enough for any one
structure belonging to a stage to emerge before, or coincidentally with, any one
structure hypothesised at the next stage. If the data shows that any structure
hypothesised to be at a higher stage appear before any structure from a lower stage
then the theory is falsified.
2 In a longitudinal study, once the emergence criterion is met in a session, insufficient
evidence in subsequent sessions is irrelevant. Hence, some cells with ‘+1’ in (3a) are
interpreted as ‘+’ in (3b).
2. The development of English as a second language
So, what unites the four morphemes within the categorial stage, despite their
differences? And in what way are they different from morphemes hypothesised at
higher stages? The PT answer3 lies in whether or not they require grammatical unification with other lemmas in the phrase or sentence structure (Pienemann 1998:
97; and (12), § 2.1, ch. 1, this volume). The four morphemes at the English categorial stage do not require grammatical unification with other constituents of the
phrase or sentence in which they may appear, so they belong to the categorial stage.
We will now proceed to discuss them following their order of emergence as shown
in (2). This order is not a necessary part of the PT hypothesis but a matter of observation in Kumi’s data. However, we do attempt to explain, in a principled way,
early appearance of some morphemes in this learner, such as –ing and possessive ‘s,
as well as the late appearance of other morphemes, such as –ed.
The earliest lemma annotation in English is –ing, the key for the L2 learner
to appropriate the minimal cognitive resources necessary to differentiate lexical
items explicitly: one category will be marked with –ing, and others will not, as in
(4a-b) where V-like items are marked with –ing and N-like and other items are not.
(4) a.
and deer running
mm. girl looking in the hole […] and girl climbing on rock
and my dad love er.. using this for the drinking
he just kept walking
he saw the man bringing it
This is consistent with Johnston’s (1994: 15) observation that “the first function of
the –ing marker is to enable the learner to build up a prototype grammatical category”. Verbal aspect or tense is not crucial yet, so there is little point in calling the
–ing morpheme a PROG(ressive) aspect marker at this stage. In fact, learners will
typically begin by marking V-like items categorially with the –ing morpheme (go
vs going) regardless of tense and/or aspect, and without the native auxiliaries. Hence
this go vs going alternation may capture differences in function which, initially, do
not correspond to the use native speakers make of these two forms. English –ing is
in fact an example of a form covering many functions and entertaining relationships within different sorts of structures (e.g., with AUXs for aspect, but also as
complement, ADJ and so on). Learners will continue to develop these associations
over time, eventually achieving quite sophisticated use, as (4c-e) testify. In the case
3 Morphemes within one and the same categorial stage in PT are notoriously placed at
the extremes of the time range for instance in Krashen’s natural hypothesis based on the
‘morpheme studies’: –ing is at the bottom, plural –s in the middle, and regular past –ed
at the top of the temporal range.
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
of Kumi, –ing is confirmed as the earliest morpheme to appear with five occurrences on different Vs already at t2, that is six weeks after her arrival in Australia.
Other diacritics such as those for marking possession and plurality on Ns or tense
on Vs will soon follow, but the important first step has been taken, and the categorial stage can be said to have emerged.
To continue with the order in which morphology emerges in Kumi at this
category procedure stage, we now deal with the possessive marker’s, and look at it
– as Pienemann and Johnston do – in comparison with possessive determiners,
which are its lexical counterpart (cf. the table in (5) below).
Before looking at Kumi’s data, let us however consider for a moment the position of possessive ’s and possessive pronouns in Pienemann’s English schedules.
Way back, in his ground-breaking SAMPLE report (1985a), Johnston places possessive pronouns at his stage 2 (now category procedure stage) and possessive ’s at
his stage 4 (now phrasal procedure stage), presenting substantial data for possessive
pronouns (pp. 340, 343) but lamenting the paucity of occurrences for possessive ’s
(19 instances sparsely and unevenly distributed over 12 of his 48 interviews), as
well as the lack of ‘developmental patterning’ and the ‘equivocal’ nature of the evidence for this form in his cross-sectional study (p. 256). Note that Johnston’s learners were of Polish and Vietnamese background, in equal numbers. As we will see
later, our Japanese background child informant presents abundant data for this
structure. Pienemann & Johnston (1987) and Johnston (2000) keep possessive ’s
at their stage 4.4 From 1998 onwards, Pienemann places possessive ’s at the categorial stage whenever it is mentioned. However, this structure is not mentioned consistently: for example, in Pienemann (2011b) it first appears in the hypotheses (tab.
4.1, p. 51), then it disappears from the tables reporting the adult SAMPLE data
(tab. 4.2, p. 52), and the child ESL data (tab. 4.4, p. 56). Possessive pronouns are
another area for which the treatment is rather erratic: in his ‘possessive pronouns’
tables, Johnston (1985) conflates both possessive determiners (or adjectives) such
as my and her, and possessive pronouns proper such as mine, hers. This may be the
meaning of ‘possessive pronouns’ in subsequent works such as Pienemann &
Johnston (1987) and Pienemann (2011a, b). In the latter, possessive pronouns are
ignored in tables 4.1-4.2 but appear in table 4.4, where however only a + sign (with
no figures) is given.
We prefer to keep the term possessive determiner for adjectives alone, which
in Kumi’s data emerge earlier and far more forcefully than pronouns, as (5) shows.
Also, as proposed earlier, we exclude both possessive determiners and pronouns
4 Krashen’s (1977) ‘natural order’ derived from the accuracy-based morpheme studies
places possessive ‘s in the same ‘box’ as 3rd PER SG –s in Vs. In PT, these two morphemes belong to quite different procedures.
2. The development of English as a second language
from the schedule in (2), because, unlike nominal morphemes such as possessive ‘s
or plural –s, they are not generalisable, and their lexical selection is not obligatory
but paradigmatic, that is, they are selected from a lexical list including other possible determiners such as article the or demonstrative this contributing to mark definiteness in N. Their lack of predictability and generalisability suggests that they do
not qualify for inclusion in the schedule of morphological development.
Nevertheless it is worth presenting the data in this (lexical) area of development
given their parallel semantic role in the NP.
(5) Possessive ’s, and possessive determiners and pronouns in Kumi’s corpus
As (5) demonstrates, Kumi’s possessive determiners are numerous, consistent, and
emerge at t2, that is, in parallel with possessive ’s. In (6a) Kumi shows that, simultaneously with possessive ’s, she is able to differentiate possessive determiners (my
big sister) from SUBJ and other pronominals (I go… me and my big sister… we
play). The utterance in (6b) shows that Kumi is well aware of the functional similarity between lexical possessive determiners and possessive ‘s, which she uses when
the lexical item is missing at t5. Then at t9, when she has the lexicon, she progresses to the specific person determiner, as in (6c).
(6) a. t2 I go . me and my big sister go to the Terry’s house and we play
b. t5 she hold she’s frog
c. t9 he call her name
The sequence in which possessive determiners emerge in Kumi’s data shows that
my, your, his, and her are the most numerous by far, followed by our and its far
behind (5 and 1 occurrences respectively), whereas the 3rd person plural form
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
(their) does not occur in the corpus. Possessive pronouns, on the other hand, also
occur, but they start much later than possessive determiners (at t4), are far fewer
(10 occurrences in total vs 190 for possessive determiners), and comprise a smaller range of forms: mostly mine, a couple of occurrences of yours, and a single
token of his.
Curiously enough, as mentioned above, possessive ’s does not always appear
in PT schedules and is rare in the published data. On the other hand, we have
seen that it emerges early and is confidently deployed by Kumi, as her seven
occurrences at t2 in (3a) and (5) testify. One may be tempted to say that Kumi’s
early acquisition and abundant use of possessive ‘s looks like L1 transfer from her
native Japanese, whose possessive phrase has a similar frame (e.g., head last, genitive case marker on possessor), whereas learners with other L1s – such as Italian
and Spanish (head first and no case markers on N), and Johnston’s informants –
seem to avoid it. With Pienemann, Di Biase, Kawaguchi & Håkansson (2005) we
assume that transfer is developmentally moderated. So it is quite plausible that
Japanese L1 learners are facilitated in learning that particular English structure
early, first because it requires an early processing procedure (i.e., category procedure) and second because their L1 has a similar structure.5 This particular morpheme then could provide an empirical testing ground for the Developmentally
Moderated Transfer Hypothesis on learner data from a variety of L1s with and
without genitive case marking.
The next marker to emerge at the categorial stage is plural –s for Ns. For
English this means that learners must be able to differentiate first whether the referent is considered as one entity or more (cat vs cats), and secondly whether the
entity referred to is semantically countable or not (apples vs milk). Learners further
need to learn that English requires this –s ending form also for referring to generic countable entities (I like apples), but not with generic uncountables (I love coffee)
– not as simple as the learner might hope for, after all. This is especially so for
Kumi, whose L1 does not mark plural morphologically. Kumi’s data shows these
difficulties. The plural –s marker emerges at t4, one month later than –ing and possessive ’s, and a pattern of nonsuppliance persists up to t146 as the table in (3a)
above shows. In terms of structures belonging to the category procedure stage, this
indicates the existence of a soft barrier. We note further that t4 is also the time
when the phrasal procedure begins to be activated.
5 Japanese is left-branching whereas English is generally right-branching, but in this particular structure (i.e., possessive case marking on N) English behaves in a left-branching way (cf. Radford 2004: 169), like Japanese.
6 Which Ns fail to be marked with plural –s, e.g., generic versus non-generic, is worthy
of further investigation (cf. Charters, Dao & Jansen 2011).
2. The development of English as a second language
The next emerging structure, the so-called regular past –ed marker, provides
a stronger case of a soft barrier, in so far as it shows up rather late in Kumi’s corpus (at t9, about 18 months after arrival in Australia). This is rather late compared to ‘irregular’ past forms such as came and went used at t4, which are learned
item by item. Once learned, the –ed marker is generalised to Vs that do not
require it, as in (7a), and to newly learned Vs, whether regular or irregular as in
(7b). Then at t14 (the100th week of exposure) the child seems to have the rule
under control. This developmental path is common for all learners of English L1,
L2, and bilingual L1.
(7) a. t9 he felled [/féld/] down
b. t13 the snail creeped up the jar
We defer further discussion of this apparent delay of the emergence of regular
past –ed to where we discuss higher morphological procedures, except for saying that, in some ways, her delay reflects the difficulties Jia & Fuse (2007)
found with this morpheme in Chinese L1 child learners in New York followed
longitudinally for 5 years. For a treatment of its emergence in bilingual children
with a variety of other languages, we refer to Nicoladis, Song & Marentette
Phrasal procedure stage. The next higher morphological procedure is the phrasal
procedure, which assembles in a grammatically compatible way the various components of this syntactic unit. Its activation allows for intraphrasal agreement or, in
any case, for exchange of grammatical information occurring within the phrase.
The structures in the phrasal procedure stage seem to emerge in two steps. In
Kumi’s data, and perhaps this is true for ESL more generally, the first phrasal structure to emerge is the VP composed by the AUX be and a V marked with –ing. The
activation of the –ing marker (recall that this is the earliest emerging morpheme) is
gradually associated with forms of AUX be, with the earliest attempts occurring
already in the earliest sessions, as (8a-d) show. These AUX be forms are soon produced in contrast with other AUXs (can and can’t) emerging at about the same time
and in a similar developmental curve, as the remaining examples (8e-k) show and
the two rows of phrasal VP in the (3a) table indicate. So there is now the need for
an exchange of information at the VP node between the AUX and the lexical V in
order to produce either V-ing or a bare infinitive according to whether AUX is a be
form or can/can’t respectively.
(8) a. t1 boy is is hold the rock
b. t2 boy is go walk
c. t2 bird egg is er . falling down t1
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
he is go up
everyone can read this
this is fly (showing with her arms an aeroplane flying)
girl is eating the long chocolate
you can’t open the door
dog is look the a hole
dog is look the bee nest
then they can thinking to play
Then Kumi’s series of modal AUXs expands to include will and couldn’t at t9-t10,
as illustrated in (9).
(9) a. t9 they will melt when they are hot
b. t10 he couldn’t see it
In sum, Kumi rarely misses unification in the VP, that is, in total 7 occurrences out
of 119 – exhaustively shown in (8a-b, d, f, i-k) – and only three times after emergence.
The kind of agreement (government) which characterises the VP is grammatically stipulated and conceptually opaque. The ESL learner must learn to select
AUX according to a range of aspectual, temporal or modal motivations, and unify
these features with the relevant ones in the lexical V (is going vs can go vs has gone,
etc.). Once this differentiation emerges it is practically categorical. Notice that, in
Kumi’s data, the structure have + V-en does emerge with some high frequency irregular Vs and contracted AUX forms, as shown in (10), but it does not occur for regular Vs. A proper PT treatment with an empirical investigation of the development
of this set of VP structures in English requires more space that can be afforded here.
Interested readers are referred to Yamaguchi & Kawaguchi (2014).
(10) a. t11 they had put A and B and C to know which is which
b. t13 no she’s got a book
c. t14 we’ve done this
The next step in the phrasal procedure stage is characterised by unification of features at the NP node when a plural context is created by numerical or nonnumerical quantifiers which must be unified with plural referents (two cats; many dogs).
For example, the NP in (11), as well as the head N cats, contains a DET, a numeral quantifier and an adjective. The NP procedure here ensures that the head N is
appropriately retrieved in its plural form cats compatibly with the numeral quantifier two, which is semantically plural, because English grammar requires such
exchange of information; that is, the PL(ural) value of the head N’s NUM(ber) feature must be compatible with any modifying element within the same phrase
2. The development of English as a second language
which also has a NUM feature in its lexical entry.7 This example shows that, unlike
Chinese or Japanese (which do not encode number morphologically), English
encodes plural morphologically in countable Ns such as cat, but unlike Italian or
Spanish it does not encode it in the adjective. Hence black does not display a number feature.
(11) a. the two black cats
b. lexical entries
black A
cats N
According to the PT acquisition criterion the child acquires plural –s without
quantifiers (category procedure) at t4, then with numeric quantifiers at t7, and
later with other quantifiers at t9 (both phrasal procedure). Agreement between N
and its DET is also clearly intraphrasal. In English, this affects only the demonstratives (this vs these, and that vs those) and some quantifiers in plural forms (e.g., lots
of). Kumi acquires phrasal agreement within NP at t7, and then gradually consolidates it with less frequent lexical items such as most at t9, as shown in (12), until
she masters it completely at t14.
(12) t9 most kids like it
Recall that VP agreement is well in place at t4, that is, 8 months earlier than NP
agreement, and soon becomes categorical. NP plural on the other hand is not only
later, but it also takes longer for accuracy to be established, like the plural –s at category procedure level, with inaccuracies persisting to the end of the data collection.
This is a clear case of a soft barrier, where various features seem to interact with
number itself, such as definiteness, specificity, mass versus count, generic versus
non-generic, as well as the possible influence of the L1. Since our database represents a single case study the patterns emerging here deserve further investigation
over a greater number of learners, from different L1s and also covering a range of
age groups.
7 This type of unification – motivated by a semantic feature in the NP environment, as
mentioned in ch. 1, § 4.1, this volume – is clearly different from the other two types
of unification, which are ‘concord’ and ‘government’; however, like the other two, it
must be constructed online.
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
S-procedure stage. The activation of the next procedure, the S-procedure, allows for
long distance, sentential (interphrasal) agreement between SUBJ and the lexical V.
At this stage, learners of English can produce the verbal –s morpheme for the 3rd
person singular of the present tense, once they manage to merge the SUBJ feature
information in the NPSUBJ (PERS = 3rd; NUM = SG) with the relevant V feature
specifications (TENSE = PRES; SUBJ PERS = 3rd; SUBJ NUM = SG), as the diagram and lexical entries in (13) illustrate.
(13) C-structure with lexical entries for Mary loves those cats
At this stage, English morphology interacts with syntax, in the sense that, because
SUBJ must agree with V (unlike OBJ), in order to produce this agreement, learners must identify one NP (the first) as SUBJ. This signals that functional assignment is in place. Kumi acquires the verbal 3rd person singular –s morpheme at t9,
as shown in (14).
(14) t9 all the time he wakes up
Interestingly, this is the same time as she learns the past –ed morpheme. It may well
be, then, that until the morphological marker –s on V (for tense, as well as person)
emerges, there is little call for a systematic difference between another tense marker, that is past –ed and nonpast verbal forms. This in fact does not mean that Kumi
never marks the past before she can activate the S-procedure, because she does use
several irregular forms, such as came and went, already mentioned above, or saw
2. The development of English as a second language
and put. After t9 more of the expected overgeneralisations of –ed (Pinker & Prince
1988) are produced but by the end of the data collection period Kumi realises it at
100%. This late emergence at t9 of the categorial-level tense marker –ed, well after
the emergence of phrasal unification (i.e., be V-ing) at t4, does not constitute a
threat to the general PT schedule for morphology because another lexical-level
morpheme (–ing) has already emerged. In other words, not all of the possible
markers belonging to a particular stage need to emerge in order to establish
whether a particular PT stage has been achieved (cf. § 5, ch. 1, this volume). This
is an often misunderstood point about PT’s architecture.
S-BAR procedure stage. Further up the developmental sequence, we include here the
interclausal level with subordination phenomena which in English affect morphological form, as the examples in the appropriate cell show (e.g., I suggest he eat less
and it’s time you left). These highly optional constructions showing an obligatory
bare or a marked form of V in the subordinate clause belong to the ‘educated’ register of English. They are rather rare even in native speaker production and quite
difficult to elicit in learners. Not only are they not produced by the child Kumi,
but to our knowledge they are not yet documented in learner data. Needless to say
not all subordinate clauses belong here. Neither, of course, is this the full story, in
more ways than one, and several important issues are not mentioned here. For
example, the activation of S-BAR procedure also includes tag questions and indirect questions, requiring interclausal agreement between V in the main clause and
the subordinate clause. In any case, questions – as we argue – require a further pragmatic motivation, a focus, which is marked (morpho)syntactically (and prosodically) in the sentence, and as such are best treated separately.
In sum, Kumi’s data shown in (3a) supports the developmental hypotheses in
(2) with a calculated scalability of 100% if we consider only the three procedures,
and 55% if we consider all the eight morphological structures shown in (3b). The
latter result is well below other PT results in this volume and elsewhere, and indeed
it is not valid, because the minimum coefficent of scalability is conventionally set
at 60% (Hatch & Lazaraton 1991). This failure is almost exclusively due to the soft
barrier imposed by the later emergence of –ed as a lexical morpheme marking past
tense. Unlike the other V morphemes (–ing and 3rd per SG –s), which are almost
completely regular forms, past –ed competes with a strong cohort of so called ‘irregular’ past V-forms, some of which are also far more frequent than any –ed form.
This delays the emergence of the ‘regular’ form presumably on account of higher
costs incurred by learners in the course of lexical selection of the appropriate form
because they need to ‘suppress’ the competing and more frequent irregular patterns.
This means that general morphological progression needs to take into account not
only unification and syntactic distance but also lexical selection costs, which con-
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
firms the need for positing something such as our proposed soft barriers.
Alternatively placing –ed at a higher stage, as Krashen (1977) does on account of
relative timing, is not an option for PT, which does not relay on timing to place
structures into stages but on whether or not the morpheme requires unification
and at what syntactic level. Having said that, if we exclude the –ed structure from
the calculation of scalability, we obtain a coefficient of 75% for the other seven
morphemes over the three stages. This means that it is higly probable that they will
be acquired in that order.
In concluding this part for morphological development, we may say that
Kumi’s data, as well as confirming both the universal PT schedules and the specific English ones, contributes towards refining and strengthening the general theoretical framework by providing additional evidence for internal hierarchies within
each procedure. Thus, for the reasons discussed above in our schedule in (2), differently from Pienemann’s schedules (e.g., 2005, 2011a, b), we place possessive ‘s
below plural –s within the category procedure stage, and we ‘populate’ the VP slot
within the phrasal procedure stage with a range of AUX V structures, and place it
ahead of the NP agreement.
3. Syntactic development
In line with the presentation of the universal syntactic schedules afforded by
the Prominence Hypothesis and the Lexical Mapping Hypothesis in chapter 1,
§ 4.2, this volume, we deal here with the learner’s progress in English, thus
testing the hypotheses on Kumi’s longitudinal corpus first from canonical to
noncanonical word order (§ 3.1), and then from default to nondefault mapping (§ 3.2).
3.1. The Prominence Hypothesis
Just as we keep the morphological hierarchy separate from the syntactic hierarchy on account of their different motivation, so, in dealing with the path the
learner follows in acquiring the English syntax, we keep declarative sentences
separate from questions. As Bettoni and Di Biase mention in chapter 1, § 4.2,
the reason is that questions are pragmatically marked sentences, in so far as
their universal defining characteristic is the focusing of the requested information, which must then be marked linguistically. This means that we will have
three schedules based on the Prominence Hypothesis: one for declarative sentences, dealing with the DF TOP, and two for interrogative sentences (Y/N
questions, and constituent questions), dealing with the DF FOC. Let us start
with declaratives.
2. The development of English as a second language
Declaratives. The syntactic hierarchy for English declarative sentences, based on the
Prominence Hypothesis, is hypothesised in (15), and the distribution of the relevant syntactic structures over Kumi’s data is presented in (16).
(15) Developmental stages hypothesised for L2 English syntax based on the Prominence
Hypothesis: declaratives (after Pienemann, Di Biase & Kawaguchi 2005)
(16) Kumi’s syntactic development based on the Prominence Hypothesis: declaratives
As we saw for the development of morphology, the very first stage, with practically no grammatical marking, requires only the activation of the lemma access procedure. Again, as with the morphology data in (3a), the single word stage in syntax is taken as acquired by Kumi before t1, because in the first recorded session she
already produces 19 canonical word order sentences.
Then, once learners can activate the category procedure and begin to differentiate verbal from nominal elements the stage is set for them to start organising
their utterances using the language-specific canonical word order found in the
input from their target language. Kroeger (2004: 141) offers criteria for deciding
what is the unmarked or most basic word order in a language, among which he
lists: highest frequency, widest distribution, neutral in terms of mood, polarity or
voice (statements rather than questions or commands), positive (not negated) and
active (not passive). This means that learners of English map concepts and mean-
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
ings by default onto a fixed NV sequence, that is, participant (mapped on N)
before action or state (mapped on V). In grammatical terms this is a Specifier/Topic
followed by Head/Comment order. When there are more than one participants in
the event, the second one – being typically inanimate or in any case lower in topicality (cf. Dalrymple & Nikolaeva 2011) or conceptual hierarchy – is placed after
the head, like any other predicative or circumstantial material such as time/place.
This establishes an NVN sequence which most often will correspond to sequences
of actor-action followed by patient/theme and/or location and/or time. Provided
they are affirmative, minimally presuppositional, and pragmatically neutral, these
sentences are syntactically target-like, as shown in (17). Here we report the very
beginning of a continuous stretch of Kumi’s production when she is telling a frog
story from the book Frog where are you? at t1, that is, 4 weeks after her arrival in
Australia. Notice that, unlike Pienemann (2011b: tab. 4.7, p. 63), we do not consider then in a narrative as an adverbial ADJ, but as a sentence connector (conjunction) similar to and. Notice also that actions or states may be expressed by lexical
items other than V, such as a movement preposition (down) in (17i), and that the
NV sequence is produced most of the time, and can be assumed to function as the
syntactic core of the expressed meanings.
(17) a.
and boy sleep er
and frog go out. of the hor house
and boy wake .
no frog in bottle
then boy look the under shoe
and dog er in
the dog look er in the bottle
and boy open the a window
and and shout
i. then dog down er window
[N V: Actor Action]
[N V PP: Actor Action Place]
[N V: Actor Action]
[no N PP: (Neg) Theme Place]
[N V PP: Actor-Action Place]
[N pause False start]
[N V PP: Actor Action Place
[N V N and V: Agent Action Theme & Action]
[N P/V N: Actor Action (movement)]
The next step forward in the development of syntax accounted for by the
Prominence Hypothesis is achieved in English when the learner’s lexicon and categorial differentiation are strengthened and expanded, and TOP is placed in first
position. This TOP constituent is not typically a core element of the sentence –
that is, it can be a time or place circumstantial ADJ in declarative sentences, to
which prominence is assigned by the speaker. This structure comes about early and
is target-like8 in a highly configurational language such as English, where canonical word order is undisrupted in declarative sentences that are pragmatically rea8 This same kind of operation would produce ungrammatical sentences in German,
which requires V in second position (cf. ch. 9, this volume).
2. The development of English as a second language
sonably neutral – we say ‘reasonably’ here, because even the topicalisation of ADJ
is pragmatically marked, albeit minimally compared to the topicalisation of a core
constituent other than SUBJ, which is also a DF and TOP by default when no
other GF is marked as TOP (cf. the notion of Stage Topic in Erteschik-Shir 2007,
Shibatani’s (1994: 277-8) ‘stylistic focus’, and the discussion on the nature of SUBJ
in Falk 2006). Kumi reaches this intermediate stage at t2, as (16) and (18) show.
(18) a. t2 today we need a hat
b. t2 in kindy we have a rain drop
At this stage, which we call the ‘TOPADJ canonical order’ stage, the canonical
string remains unaffected by topicalisation. However, speakers may still need to
attribute prominence to grammatically more central constituents, typically OBJ.
At first, when OBJ is topicalised, the Prominence Hypothesis proposes that learners keep the canonical string complete by duplicating the topicalised element in its
canonical position with a pronominal element. Kumi does exactly this twice at t10
and once more at t13 in a kind of structure-preserving operation, as (19) shows. In
fact the syntax of all three examples – two in (18) at t2, and one in (19) at t13 –
may be represented with the single phrase structure rule in (20), where S = canonical word order.
(19) t13 tomato thing I like it
(20) S’
(↑TOP) = ↓
At the last stage, native-like noncanonically aligned sentences are hypothesised,
allowing for highly marked constructions such as those rarely used in English (cf.
Bresnan’s example reproduced in ch. 1, (19b), this volume) – being highly configurational, English prefers other than syntactic means for assigning prominence to
a particular element in a declarative, e.g., prosodic means involving stress and intonational contours (cf. Selkirk 1984; Levelt 1989: 303-305), a domain which
remains unexplored within PT. This last syntactic stage emerges in Kumi only at
t14 with a single example shown in (21). So this seems to be a large step forward,
achieved late.
(21) t14 this you can click. on lots of buttons
Interrogatives. As remarked before, questions are not pragmatically neutral constructions, and universally require the focusing of the requested – new and unpredictable – information (Kroeger 2004: 138-139). With Y/N questions, the focus
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
broadly falls on the whole sentence by default (wide focus), as in (22), whereas with
constituent (or content, or wh-) questions it falls on a specific constituent (narrow
focus), as in (23).
(22) a. did you have pizza for lunch?
b. yes
(23) a. what did you have for lunch?
b. a pizza
Specific constituents may of course be brought into focus also with Y/N questions,
as the felicity of the answer in (24b) attests.9 In English, this sort of focusing may
be achieved by prosodic means (Lambrecht 1994; Van Valin 2005), as shown in
(24a), but we do not deal with this narrow focus here because, as we said, prosody
is beyond the proposed scope of this chapter (and PT), as well as being out of reach
for the early L2 learner, who would have to become highly familiar with default L2
prosodic patterns in the first place in order to use them in this flexible way.
(24) a. did Mary buy a CAR?
b. she bought a motorbike
Across languages, different means are used to tell interrogatives apart from declaratives, such as special intonation patterns, a particle or clitic, a modal affix on V or
a change in word order (Kroeger 2005: 203-205). In English, the wide scope of
Y/N questions has grammatical reflexes such as word order restrictions and other
lexical properties of AUXs. These ordering and selectional restrictions are represented in the lexical entry of AUXs, and learners, however gradually, will need to
acquire them in order to produce marked order questions, involving AUX-SUBJ
order among other things. Also, Vs are capable of being assigned information structure (i-structure) features [±prominent] and [±new] (cf. ch. 1, 4.2.1, and (36), this
volume). On the other hand, the narrow scope of constituent questions also has
the same grammatical reflexes, and additionally it requires the wh-element bearing
the pragmatic focus in initial position. In the following pages, with Kim & Sells
(2008: 191, ch. 10), we adopt the feature question – QUE(stion) – for marking
any of these exponents characterising questions, as well as Choi’s (1999, 2001) istructure features, without going into formalisation details. We will deal with polar
questions first because they are more numerous in our informant’s data (138 out
of 180), and then with constituent questions.
9 This possibility of narrowing the scope of Y/N questions skews them towards content
questions (cf. Givón 2001: 230-34).
2. The development of English as a second language
Polar questions. Our proposed schedule for the development of Y/N questions is
hypothesised in (25), and then tested on Kumi’s data in (26).
(25) Developmental stages hypothesised for L2 English syntax based on the Prominence
Hypothesis: Y/N questions
(26) Kumi’s syntactic development based on the Prominence Hypothesis: Yes/No questions (reanalysis of Yamaguchi 2010)
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
In single-constituent Y/N questions the referent, known to both interlocutors,
requires confirmation or further information. This requesting modality, formalised
as QUEP in (25), is marked by prosody exclusively (with rising intonation). Some
of Kumi’s examples are shown in (27).
(27) a.
light green?
blue coat?
long or short?
yes or no?
one man?
Polar questions at the next hypothesised stage (i.e., the canonical word order stage)
are also expressed by prosodic means only, but now the learner of L2 English is able
to form a full SVO string. So, in the very first two stages, the contrast between the
QUEP modality and the declarative modality is entrusted to prosodic means exclusively. And this is why, up to now, the modality marker QUEP in (25) is enclosed
in square brackets to indicate that it does not yet have a lexical exponent. Only
later, at the QUE canonical word order stage, is a small set of lexical material (e.g.,
do, do you, do you have, do you got) preposed to the canonical string and added to
the prosody to jointly express the question modality, and later still, at the noncanonical order stage, these lexical means are integrated, as AUX, within the sentential syntax. This lexically expressed QUE is not enclosed in square brackets and
is, informationally, [+prominent]. We will now show these developments in some
more detail as we look at Kumi’s data.
In Kumi’s data, structures belonging both to the [QUEPcanonical order] and
QUE [canonical order] stages appear at the same time, quite early at t2. This does
not contradict PT because there is a conflation, not a skipping of the stage. At the
canonical word order stage, interestingly, the order NVX with copula appears earlier than with lexical V, as illustrated in (28). Quite likely this is aided by the fact
that the copula is far more frequent in the input than any lexical V.
(28) a. t2 your light is off?
b. t2 bicycle is a red?
c. t4 dog eating the doughnut?
What is even more interesting is the way in which, at the QUE canonical order
stage, Kumi initially discovers that QUE may have lexical exponents expressed
with do you have and do you got. Despite their occasional apparent grammaticality, the do element here is not yet integrated as a proper AUX within the questioned sentence. Significantly, at this time, Kumi’s AUX system in the declara-
2. The development of English as a second language
tive grammar is not fully developed either. Nor, of course, could the whole
string do you have or do you got ever be an XP – as the general progress of the
Prominence Hypothesis in chapter 1, § 4.2.1, would have it. This is indeed the
very feature that sets polar questions apart from other types of sentences, and
consequently the very reason why in our English Y/N question schedule in (25),
the XP in prominent position is replaced by QUE in prominent position also.
Kumi’s examples in (29) support this proposition of a do QUE which at this
stage is a lexical step towards developing a fuller AUX system. The 37 QUE elements produced by the child from t2 to t5 are all do you have..., some of which
(e.g., 29a-c) look grammatical, whereas others (e.g., 29d-f ) are not. Here Kumi
is attempting to create ‘syntactic amalgams’ (Lambrecht 1988), main or subordinate, which are as yet beyond her grammar. A very similar progression
towards the formation of complex clauses is observed also in English L1 acquisition (Diessel 2004).
(29) a.
do you have a man on the picture?
do you have a flower on the leaf picture?
do you have a trash can in the picture?
do you have a man coat is a blue?
do you have a man wear the sun glasses?
do you have a flower is pink?
Later on, at t6, do you have enters in competition with do you got, as in (30a-c),
which however is soon abandoned altogether (from t7) in favour again of do you
have. Later still (at t9), a further development occurs when Kumi seems to understand the need to signal polar questions with a QUE AUX other than do by replacing it with an inflected be form, as in (30d).
(30) a.
do you got a newspaper?
do you got a man wearing the red hat with yellow ribbon?
do you got a bin in there is a rubbish in bin?
is your man have a red hat?
Already at t6, however, a more important development seems to be happening
when Kumi not only experiments with an alternative (ungrammatical) QUE
(do you got) form, but also starts producing two Y/N structures belonging to
the next marked alignment stage, which displays an integrated AUX, as shown
in (31).
(31) a. t6 can you see?
b. t6 is bicycle falling down?
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
We can thus assume that from now onwards the ungrammatical use of the do you
have QUE element followed by other V-like material declines, and that instead an
integrated fronted do AUX becomes productive whenever there is no COP or other
AUX element available in the sentence to mark the polar question.
After their initial production at t6, Kumi’s Y/N questions belonging to the
last noncanonical word order stage remain rare for some time (until t13); that
is, she produces one COP example at t7 and one MOD AUX example at t9, as
shown in (32).
(32) a. t7 are you here?
b. t9 can I have the that walkman?
Finally, at t13 and t14, Kumi is able not only to identify do AUX fully as governing lexical V and carrying 3rd person singular SUBJ information, as in (33b), but
also to include a new, and less common, AUX such as could in (33c).
(33) a. t13 do you have a boy listening to music?
b. t14 does your man have green coat?
c. t14 could you ask me a question?
In sum, for polar questions, as the development of COP and AUX continues, it is
possible to trace the learner’s progress from a more lexical question marker followed
by canonical order to a more legitimate AUX properly integrated in a sentence with
marked word order. Torregrossa & Bettoni (2013) report similar results for Y/N
questions in English EFL data by adult Italian learners.
Constituent questions. For a typological perspective on constituent question formation whithin the nonderivational framework of LFG, we refer to Mycock (2007).
Constituent questions are realised typologically by two distinct means: in-situ languages, such as Japanese or Chinese, signal focal prominence of question phrases
by prosodic means (and a specific lexical set), whereas fronting languages, such as
English and Italian, signal it by syntactic means (and their specific lexical sets) –
although this does not mean that prosody does not play a role (e.g., the wh-word
is usually stressed in English), but simply that syntax plays a crucial role. In native
speakers’ English, however, in-situ questions such as those in (34) are perfectly
acceptable given appropriate pragmatic context and intonational contour, although
they are rare and highly marked, expressing for instance amazement, surprise or
disbelief. In fact, although typologically languages are clearly either in-situ or fronting, neither of these means of forming questions is exclusive of the other in a single given language.
2. The development of English as a second language
(34) a. [Tom ate a fly]
b. [Anne is going to the South Pole]
Tom ate what?
she is going where?
We will now lay out the developmental hypotheses for English constituent questions in (35), and then present Kumi’s results in (36). Note that Kumi’s L1,
Japanese, is typologically an ‘in-situ’ language, and she is learning English, a
‘fronting’ language.
(35) Developmental stages hypothesised for L2 English syntax based on the Prominence
Hypothesis: constituent questions
(36) Kumi’s syntactic development based on the Prominence Hypothesis: constituent questions (re-analysis of Yamaguchi 2010)
Our informant Kumi produces 180 questions in total as we mentioned, of
which content questions are less than one third (52). This imbalance may be
due to the lack of specific tasks designed to elicit this sort of data. Also the range
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
of wh-words produced is restricted. For instance, among GFs, SUBJ is not questioned, and neither is any OBL.
At the lemma access stage Kumi produces one instance of what at t1 and a
peak of this wh-word at t4 (3 occurrences), when it alternates with the formulaic
single word (I beg your) pardon? (also 3 occurrences). From t7 onwards, Kumi uses
the five question phrases listed in (37).
(37) a.
how many flowers?
which one?
what the noise?
which grandma?
what flowers?
It may be argued that these are neither single words nor formulas, and thus perhaps requiring an intermediate stage before the WHQUE canonical order stage.
However, we do not propose this because Kumi’s data is too limited, and leave
them at the initial stage because they are without a V. The remaining 8 questions
at this stage are clearly formulas because they are all exactly the same (where are
you?) and coincide with the title of the book Frog where are you? which was well
known to the child and used several times over the two years of data elicitation.
In our hypotheses for L2 English in (35), we place in-situ wh-questions at the
next canonical word order stage. Also Pienemann (2007) places them within the
hypothesis space of early learners. Yet it is interesting to note that, at an early stage,
they are attested neither in the literature, nor in Kumi’s corpus. That is, learners from
a variety of language backgrounds appear to latch on quickly to the specific structural means prevalent in the target language, regardless of whether their own L1 may
use a different mechanism. A case in point is Kumi, whose L1 Japanese prefers insitu question. Yet, like other learners of English, she appears not to use them, and
soon learns the English characteristic of fronted focal prominence, as her results will
show. Similarly, children in bilingual L1 acquisition where English is the dominant
language outside the home appear not to display in-situ constituent questions in
English (Qi 2012; Qi & Di Biase 2005 for Chinese-English; Itani-Adams 2007,
2011 for Japanese-English). However, in other contact situations, such as that of
Cantonese-English bilingual L1 acquisition in Hong Kong, meticulously described
by Yip & Matthews (2007: 93-111), in-situ questions do show up strongly and
unequivocally. This may point towards a significant role of environmental influences
in language development (cf. the discussion for L2 Italian data in ch. 8, this volume).
At the next stage, our hypothesis is that having learned to disentangle the DFs
TOP and FOC from SUBJ, learners of English can now place the focal questioned
constituent, represented by the wh-word, in first position followed, steadily, by
canonical word order. Kumi has three such questions, all why questions as shown
2. The development of English as a second language
in (38). Here the XP is not an argument but an ADJ. Thus this constituent question structure with FOCADJ parallels the TOPADJ canonical order structure predicted at this same stage for declarative sentences by the Prominence Hypothesis –
cf. (15) and (18).
(38) a. t8 why you come and shout in my house?
b. t9 why you fall down?
c. t11 why I put the crab in the jar?
As the sentences in (38) show, the outcome with why (ADJ) questions is not target-like in English – that is, it would be if the questioned constituent were itself
SUBJ, but we have already remarked that there is no who question in Kumi’s data.
SUBJ also has a special status in content questions in other languages besides
English, according to Falk (2001). For a study on adult L2 English which includes
who questions, we refer to Kawaguchi (in press). Her cross-sectional data supports
Falk’s claim in so far as who questions appear to be highly problematic for all learners except for the most advanced ones.
The developmental path for content questions concludes at the marked alignment stage, that is, when, after establishing the question word or phrase in FOC
position, the learner is able to use functional assignment and place SUBJ in the
appropriate post-COP/MOD/AUX position. Kumi starts off producing constituent questions belonging to this stage with the wh-word followed by COP at
t8. In (39) we show Kumi’s four examples occurring at t8 and t9. Recall that polar
questions with COP were already established at t6. So the addition of the wh-FOC
does not alter the post-AUX position of SUBJ.
(39) a. t8
b. t9
c. t9
what’s the purse?
where is she?
where are you? [twice]
As the reader may have noticed, Kumi’s two questions in (39c), coinciding with the
title of her book, are exactly the same as those we have placed among the formulas
when she produced them earlier on. However, there is evidence that she can now
construct them online. The evidence for this is provided by her other two examples: (39a), where COP not only occurs before SUBJ but is also contracted; and
(39b), where the same 3rd person COP form is not contracted and the SUBJ pronoun is correctly in 3rd person in contrast with the are in (39c).
Constituent questions belonging to the marked alignment stage with the whword before AUX emerge in Kumi’s corpus at t9. Notice that this AUX is actually
a form of do (40a-c), which has a strong pattern of preceding the SUBJ in polar
questions already at t2. In the very last session of data collection the child concludes
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
her developmental path and produces two examples of the more sophisticated whword with modals (40d-e) which were also established earlier (at t6) as preceding
SUBJ in polar questions.
(40) a.
what do you means?
what did she do?
what did they do?
what would he do?
what should they do?
In concluding this section on the development of syntax according to the
Prominence Hypothesis, we should again warn the reader – as we have already
done with respect to the development of English morphology – that this is not the
full story for questions. For example, although usually included in presentations of
PT, indirect questions and tag questions are not dealt with here. Our reason for leaving them out is that they both involve two clauses, the former through subordination, and the latter through long distance unification of verbal features across
coordination; furthermore the latter also involve negation. Indeed, negation and its
complex interaction with questions is not treated at all here. These constructions
require specific investigations dealing with subordination and negation.
3.2. The Lexical Mapping Hypothesis
As we have seen in chapter 1, § 4.2.2, this hypothesis deals with the mapping of
a-structure to f-structure, accounting for the development from default to nondefault structures, as motivated mainly by lexical selection, which may in turn be
based on discourse-pragmatic motivations. The universal schedule shown there in
(43) is language-specifically applied to L2 English with illustrative examples provided here in (41).
This staged development for L2 English has been convincingly tested in several empirical cross-sectional studies, including one by Wang (2010). Another, also
by Wang (2006, 2011), uses a large body of data gathered by means of patient-cued
pragmatic contexts, and demonstrates that only the most advanced Mandarinspeaking learners of English are able to produce sentences using the pragmatically
appropriate nondefault syntactic mapping. A third study by Keatinge & Keßler
(2009) investigates passive constructions in classrooms. A fourth by Kawaguchi
(2011) involving Japanese L2 investigates the acquisition of nondefault mapping
(e.g., passive and causative as well as unaccusative and psychological Vs) which may
not be observed frequently in naturalistic, longitudinal studies. We further test this
staged development for L2 English on Kumi’s longitudinal data here in (42)
including the intermediate stage of default mapping with additional argument.
2. The development of English as a second language
(41) Developmental stages hypothesised for L2 English syntax based on the Lexical Mapping
(42) Kumi’s syntactic development based on the Lexical Mapping Hypothesis
In dealing with the development of declarative sentences in § 3.1, we have already
shown the early steps taken by Kumi in mapping arguments in her new L2. In the
examples in (17) it is clear that initially the actor/agent-like roles are the ones most
commonly mapped in TOP (or SUBJ) position with theme/patient mapped in
OBJ position.
Not before t6 (after about four months’ exposure) does Kumi reach the intermediate stage of ‘default mapping with an additional argument’ where the added
argument is marked as such, that is, distinguishing it from SUBJ and OBJ, which
are marked by position only. At t6, this additional element is marked both as OBL
by a preposition and as OBJ2 by position, as in (43a) and (43b) respectively.
(43) a. t6
b. t6
Bruno Di Biase, Satomi Kawaguchi and Yumiko Yamaguchi
owl came from a tree hole
we give the kangaroo a kangaroo food
Unfortunately Kumi’s longitudinal study stops when nondefault mapping is just
emerging in her interlanguage. This last stage is only represented in her data by the
two examples shown in (44), one produced with a passive structure at t13, and the
other with an unaccusative V at t14.
(44) a. t13 the cat was chased by the mother bird
b. t14 is your bike not broken and fall falled over?
Nevertheless we can clearly see that her data fully supports our schedule hypothesised in (41).
4. Conclusion
In this chapter we have offered a refreshed and in some areas more complete and
supported map of L2 English within a PT perspective. For morphology, we clarify
the status of possessive –s within the categorial stage and propose to exclude possessive determiners and pronouns, identifying them as lexical rather than morphological structures. In addition we propose to place the unification of AUXs with
lexical Vs firmly within the phrasal stage. We have also exemplified and, at least in
part, explained the scouting and trailing observed in intrastage sequencing by proposing the concept of soft barriers to account for those instances where feature unification and syntactic distance may be insufficient to account for previously unexplained progressions within stages.
For syntax, we coherently incorporated and further explored the intuitions
broached in Pienemann, Di Biase & Kawaguchi’s 2005 PT extension, leading to
our proposed Prominence Hypothesis, which embraces questions as well as declaratives. We believe that the longitudinal data examined here, despite some gaps,
supports our current vision. Given the limited data belonging to a single child learner it would be important to validate and complete our proposals with further data
with a broader set of learners in a wider age range. It will also be crucial to pursue
an appropriate formalisation incorporating i-structure in a wider Optimality-theoretic LFG (OT-LFG) launched by Bresnan (e.g., 2000) (cf. Asudeh & Toivonen
2010: 453).
Needless to say, there are still gaps in this story. Several important issues are
not discussed here. For example, the activation of S-BAR procedure yields a further stage in English, which includes, for instance, tag questions and indirect questions, requiring interclausal agreement between V in the main clause and V in the
2. The development of English as a second language
subordinate clause. In any case, questions – as we argued – require a further pragmatic motivation, a focus, which is marked (morpho)syntactically (and/or prosodically) in the sentence, and as such are best treated separately. Another area not discussed here includes the variation implicit in the Hypothesis Space (Pienemann
1998: ch. 6) and also the relationship between accuracy and development. As for
the Lexical Mapping Hypothesis we are barely scratching the surface here, but we
believe we are on the right track and would like to encourage other scholars to take
this interesting area further.
The development of Italian as a second language
Bruno Di Biase* and Camilla Bettoni**
*University of Western Sydney, **University of Verona
1. Introduction
Italian is a nonconfigurational, null-SUBJ headmarking language characterised by
a rich morphology and a flexible syntax which is highly sensitive to pragmatic and
discourse choices.1 From the point of view of the effect of pragmatics on syntactic
structure, Van Valin (2005: 77) locates Italian among languages with ‘flexible syntax and rigid focus’. English, on the other hand, is among languages with ‘rigid syntax and flexible focus’, which makes the contrast between the two languages
intriguing. These typological characteristics are of interest to PT in two fundamental ways. First, with regards to the notion of transfer of grammatical information
within and between phrases of a sentence (cf. ch. 1, § 4.1, this volume), Di Biase
& Kawaguchi (2002) show that, despite the basic contrast with English, Italian
interlanguage data fully validates the universal hypotheses about the development
of morphological structures and their interaction with syntax as hypothesised in
Pienemann (1998), who had not looked at any Romance languages. Secondly, and
perhaps even more importantly, with regards to the LFG architecture of correspondences among its three parallel levels of linguistic representation, the need to
account for the nonconfigurationality of Italian syntax contributed substantially to
the formulation of PT’s hypotheses about the development of syntactic structures
at the interface with discourse-pragmatics (cf. ch. 1, § 4.2, this volume). As a matter of fact, Di Biase & Kawaguchi (2002) pioneered the use of the newly formalised LFG DFs in PT, thus foreshadowing the extension later developed in
Pienemann, Di Biase & Kawaguchi (2005). In what follows, unlike any previous
treatments of Italian processability, we revisit and expand the morphosyntactic
framework for Italian L2 development and propose a theoretically motivated way
1 For a brief description of Italian grammar written in English, cf. Maiden (1995), and
Vincent (2011). For more comprehensive treatment in Italian, cf. Renzi, Salvi &
Cardinaletti (2001), Salvi & Vanelli (2004), and Schwarze (2009) – the latter within
an LFG framework.
Grammatical development in second languages, 117-148
Bruno Di Biase and Camilla Bettoni
forward for dealing with the so-called intrastage phenomena (cf. § 2). We also offer
a fairly comprehensive discussion of the interface between syntax and discoursepragmatics, with empirical support (cf. § 3).
2. Morphological development of L2 Italian
PT-derived hypotheses for the acquisition of Italian must deal with its rich morphology instantiating all-pervasive and obligatory agreement patterns.
In terms of morphological typology, Italian is located higher than English on
the index of fusion continuum. This is the index which measures the extent to
which morphemes are segmentable, with agglutination at one end, where segmentation is straightforward, and fusion at the other end, where there is no segmentability (Comrie 1989: 46). English morphemes are more easily segmentable
than Italian morphemes, which in most cases fuse a number of grammatical features in a single exponent. Segmentation of inflectional morphemes is often more
problematic in Italian even than in other Romance languages such as French,
Spanish or Portuguese, all of which, for instance, have adopted suffixation of –s to
mark plural in nominal inflection. Italian on the other hand has a system of vowel
alternation (Vincent 2011), which makes nominal number and gender hard to factor out, and more opaque for learners.
The other important characteristic of Italian morphology is that it is stembased, like Russian and Hebrew, rather than word-based, like English or German.
This is significant from a processing point of view, because – for the vast majority
of nouns and adjectives, and for all verbs – Italian stems do not amount to full legal
words, and must always bear some inflectional ending. The function of these
inflectional endings is to express grammatical categories such as number, gender,
mood and tense (Maiden 1995: 92). For example, the lexical item in (1a) cannot
be realised in its bare stem (1b), but it must have one of the four inflectional vowel
endings typical of Italian nominals, as in (1c). The inflectional endings in (1c)
mark the gender contrast (masculine vs. feminine: e.g., ragazzo, ‘boy’ vs ragazza,
‘girl’; ragazzi, ‘boys’ vs ragazze, ‘girls’) and the number contrast (singular vs. plural,
e.g., ragazzo, ‘boy’ vs ragazzi, ‘boys’; ragazza, ‘girl’ vs ragazze, ‘girls’) in nominals.
Learners appear to acquire the phonological part of the process very early, namely,
the fact that Italian words typically display a vocalic ending. But then it takes them
much longer to account for the grammatical information loaded in the vocalic variation they hear in the input at the end of words.
(1) a. {ragazzo} [boy]
b. */ragats-/
c. / –o ~ –a ~ –i ~ –e /
3. The development of Italian as a second language
Apart from the irregularities found in any system, nominal group marking in
Italian is made more complex than the paradigm presented in (1) by the existence
of several phonologically-based noun classes. In addition, from a semantic point of
view, Ns with features +human and/or +animate do not always match their ‘natural’ and grammatical genders. All other nouns are assigned by the grammar to one
or the other gender in an arbitrary way, often following phonologically based criteria: e.g., Ns ending in the unmarked singular citation form –o tend to be assigned
to masculine gender (libro, ‘book’), and those ending in –a to feminine gender
(casa, ‘house’). Yet Ns ending in –e are masculine (pane, ‘bread’; leone, ‘lion’) or
feminine (neve, ‘snow’; tigre, ‘tiger’) in an arbitrary way.
Also nominal modifiers, such as determiners, demonstratives and adjectives,
must express the same gender and number values as their head Ns. Nominal modifiers also fall into classes: those with the same four endings seen in (1c) for Ns (rosso
~ rossa ~ rossi ~ rosse, ‘red’), and those which neutralise the gender distinction by
having –e ending for singular, and –i ending for plural (verde ~ verdi, ‘green’), irrespectively of whether their head is masculine or feminine. An anchor point in this
extreme variation is offered by the stability of –i as appearing with plural referents
consistently. No wonder, as we shall see, that learners latch on to this morpheme
and use it as a kind of prototypical, or default, plural marker. This phenomenon of
identifying a certain form as bearing some particular function by default seems to
be pervasive across all levels of morphological as well as syntactic and pragmatic
The task faced by the learner in sorting out Italian nominal inflection is complex enough. Yet it is rivalled by that imposed by verbal morphology. We will not
deal with the latter here, except to mention briefly that Italian Vs fall into three
classes, each with a characteristic thematic vowel distinguishing three conjugations
(–a–; –e–; –i–); and that a typical V has 47 or so finite forms, marking tense, aspect
and mood, as well as person and number.
The complexity of the Italian inflection system offers a good example of the
way in which the primary PT notion of information exchange within and across
constituents needs to be complemented by other principles in order to explain
the acquisitional process. Among these, there is the form-to-function relationship (Pienemann 1998: § 4.3). That is, the actual learning of the morphological
form of the affix in relation to its function is a different task from that of managing information distribution in the affixation process, where diacritic features
have to be exchanged within different grammatical structures. The figure in (2)
illustrates how Italian Ns mark the plural value of their number feature through
a complex set of form-function relations. The many-to-one relationship, where
several morphemes mark one and the same feature is exemplified in (2a). Then
there is the one-to-many relationship, as shown in (2b), where a particular morpheme marks more than one function. On the other hand, the most consistent
Bruno Di Biase and Camilla Bettoni
form-to-function relationship is shown in (2c), where the final –i vowel at the
end of Italian nominal forms only marks the number feature [plural]. Similar
form-function mapping problems may also be expected with the acquisition of
Italian verbal paradigms, where the vowel ending of one form (e.g., mangia
“(s/he) eats/is eating”) carries information regarding several features at once, such
as subject person, subject number, tense, aspect and mood. A veritable labyrinth
for the learner.
(2) Form-function relations of Italian plural nominal markers
In essence, the relationship between morphological forms and their functions
exhibits different degrees of complexity. This adds another dimension to the learning task which is separate and different from the task on which PT is focused,
namely the exchange of grammatical information and the use of diacritic features.
So far PT has not made any predictions on how a fuller paradigm develops.
However, on the one hand, the more regular and simpler one-to-one form-function relationships may help to bootstrap the more complex ones (cf. Andersen’s
1984 “one-to-one principle in interlanguage construction”). On the other hand,
teasing out of different factors allowing to progress from emergence2 to full mastery of the whole system is one of the directions in which future research can go –
and not only with regards to Italian (cf., e.g., the brief mention of Russian morphology in ch. 5).
2 For a thoughtful assessment of the emergence criterion in PT as applied to Italian, cf.
Pallotti (2007).
3. The development of Italian as a second language
2.1. The hypotheses
Let us now consider some of the main language-specific Italian L2 structural outcomes of the morphological processing procedures universally predicted by PT (cf.
ch. 1, § 4.1, this volume). Our hypotheses are shown in (3), and then discussed
with interlanguage examples taken from Di Biase’s corpus analysed in Di Biase &
Kawaguchi (2002) and Di Biase (2007), and briefly described in § 2.2.
(3) Developmental stages hypothesised for L2 Italian morphology (extended from Di Biase &
Kawaguchi 2002)
After leaving behind the single-words and formulaic (lemma access) stage, learners
begin to incorporate language-specific procedures at the next (lexical-morphology)
Bruno Di Biase and Camilla Bettoni
stage, when categorial marking for Italian Ns is achieved through the use of articlelike forms, such as la or il as generalised markers (cf. also, in a non-PT framework,
Chini & Ferraris 2003: § 3.4.1), as la acqua ‘the water’ in (4), and l’oca ‘the goose’
in (5) further below.
(4) Researcher: sotto nell’acqua? [under the water?]
no no la acqua [no no water]
In PT the category procedure is characterised by the lack of information
exchange with any other element in the phrase or clause. Our interpretation
then would not consider the combination of noun and article as evidence for
phrasal procedure morphology, independently of whether or not the specific
combination turns out to be target-like (cf. also Di Biase 1998). In fact, bare Ns
are not often produced at early stages of Italian L2 development, and are in any
case highly restricted also in native Italian, aside from personal names.
Furthermore, the Italian article is syntactically incapable of appearing by itself
and is prosodically not independent of the content word that follows, and
groups under its stress field. This, in turn, would favour a sort of formulaic
(exemplar) learning of article-N combinations. For all these reasons, such combinations are hypothesised as belonging to the lexical level, in the sense that article forms are considered as part and parcel of categorial marking of Ns rather
than as agreeing determiners3 . A clearly different situation obtains with other
determiners or N modifiers such as demonstratives, which may appear (syntactically) independently or at the end of phrase boundaries, and have their own
independent stress pattern and prosody. When such N modifiers must agree
with their N head, a phrasal procedure needs to be called upon.
At this category procedure stage in Italian, the plural –i diacritic, being the
most consistent marker of plurality in the language among several other markers
(cf. (2) above), turns out to be the first to emerge in conjunction with plural referents, as in (5).
(5) Lois: l’oca
ritornato *caroti
the goose returned carrots
[the goose has returned the carrots]
Like Ns, Vs also show categorial marking in Italian, such as the infinitive –re ending used in various contexts, as in (6). At this stage, perfective past tense can also
3 Naturally, some defaults may be learned quickly (la casa), others (il problema) may take
a long time (cf., the discussion on soft barriers below).
3. The development of Italian as a second language
be marked by the –to past-participle ending,4 although not yet in analytical constructions with their AUX, as in (5) above. Some person marking on V is also
attested at this stage. See, for example in (6), the formal contrast between capire and
capisco that may mark first person with the characteristic null SUBJ. This contrast,
however limited, is sufficient to show that person marking appears at a much earlier stage in Italian compared to English – a fact that can be explained by the nullSUBJ nature of Italian (cf. our discussion of this point below).
(6) Lois: non... capire...
non capisco
not understand-INF not understand-1.SG
[(I) don’t... understand... (I) don’t understand]
The morphological processes that characterise the next, phrasal, stage in Italian
interlanguage include nominal and verbal agreements. Within the NP, learners
start producing the agreement of determiners (other than articles as we have seen),
and/or adjectives in attributive function, with the gender/number of the head N,
as in (7). Within the VP, unification of number value (singular or plural) produces
two types of agreement: one between the copula and a predicative adjective or a
nominal, as sono cugini in (8); the other between the person marker in the (essere)
AUX5 and the ending in the lexical V, as in (9).
(7) Anne: non ho
maschili [maschi]
not have-1.SG many-PL.MASC friends-PL.MASC male-PL.MASC
[I don’t have many male friends]
(8) Amy: sono
della mia mamma
are-3.PL cousins-PL.MASC of my mother
[they are my mother’s cousins]
(9) Toni:
ah sì, sono
oh yes are-3.PL gone-PL.MASC
[oh yes, they went to the bdroom]
alla camera di letto
to the bedroom
Let us now move on with the learners to the S-procedure stage. Notice that in
English, on account of obligatory SUBJ, person variation in the V-form is placed
4 The –to marker also marks perfective aspect but it is difficult to segment out tense and
aspect marking.
5 Analytical tenses require different AUXs (essere ‘be’, avere ‘have’) that carry the finiteness features (Schwarze 2009: 150). However only the essere AUX is relevant for phrasal
agreement because it requires GEN/NUM agreement between SUBJ and lexical V.
Bruno Di Biase and Camilla Bettoni
high in the processability hierarchy (cf. ch. 2, § 2). Italian, on the other hand, being
a null-SUBJ language, maps the person-number (singular or plural speaker,
addressee or third person) directly on V form without a necessary co-reference to a
nominal or pronominal SUBJ (cf. Di Biase & Kawaguchi 2002 for an LFG formalisation). Indeed SUBJ may not be expressed at all, or be generated after V (cf. §
3.1 below). Results from psycholinguistic experiments (e.g., Vigliocco,
Butterworth & Semenza 1995; Vigliocco, Butterworth & Garrett 1996) support
the hypothesis that SUBJ-V agreement in null-SUBJ languages is generated via an
independent retrieval of the features of V and those of SUBJ. If this is the case, then
– for Italian and other null-SUBJ languages – achievement of interphrasal morphology may be more clearly expressed by structures other than SUBJ-V agreement, so as to allow for the fact that at least some of the different person-number
forms of V are acquired, as we have just seen, at an earlier stage.
In Italian too, of course, interphrasal morphology requires the S-procedure,
that is, the procedure for unifying different categories of constituents at sentence- or clause-level. This means that, for the emergence of structures belonging to this stage, the learner must recognise the grammatical relations (e.g.,
SUBJ, OBJ) expressed by the various constituents of the clause, as well as identify the category of each constituent, and more generally the relationship
between predicates and their arguments, including predicates of an adjectival or
nominal nature, as we have seen in (8)-(9). So what are the candidate structures
for Italian at the sentence agreement stage? One structure that can be built on
those in (8)-(9) is the unification of SUBJ features (gender and number) with
nonverbal predicates, as in (10).
(10) Anne: i genitori
di mia mamma sono
the parents-PL.MASC of my mother are-3.PL Australian-PL.MASC
[my mother’s relatives are Australian]
Other good candidates include agreements in verbal analytic constructions (with
AUXs) that are likely to be unified online, provided they require nondefault (i.e.,
not singular and masculine) unification. By this we mean not the unification of the
person feature of SUBJ, which is carried by the AUX, but of the values for its number and gender features, which must be unified with the lexical V, as in (11).
(11) Amy:
siamo andate
da Napoli a Palermo
we-1.PL are-1.PL gone-PL.FEM from Napoli to Palermo
[we went from Napoli to Palermo]
Here, the number value of the lexical V form (andate, ‘gone’) is unified with
that of the pronominal SUBJ (noi, ‘we’): plural in both cases. On the other
3. The development of Italian as a second language
hand, the feminine gender value also marked on the lexical V (andate) is not
marked in the pronominal SUBJ (noi), which could indifferently refer to males,
females or mixed referents. So, we may ask, where does the feminine gender
information of the lexical V-form come from? The answer to this question must
be that – because both features (gender and number) are required by V – the
gender value is retrieved by the V lemma directly from the conceptual structure.
Pronominal SUBJ, on the other hand, requires only the number value. It is
these kinds of feature distributions and unification patterns that lead Di Biase
(2007: § 1.2) to support the ‘independent retrieval’ assumption of Vigliocco
and her co-workers, who carried out numerous experiments concerning SUBJV agreement in a range of typologically different European languages (e.g.,
Vigliocco, Butterworth & Garrett 1996; Vigliocco, Hartsuiker, Jarema & Kolk
1996; Vigliocco & Franck 1999, 2001). This line of research suggests that, at
least in null-SUBJ languages with SUBJ-V agreement, both SUBJ and V
retrieve features from the conceptual structure independently, and then merge
them at the S-node. Hence, the V-form is essentially a lexical-stage form, stored
in the mental lexicon of the speaker, whose features must match or be compatible with its agreeing counterpart.
Still at the S-procedure stage, and again referring to (3) above, another structure hypothesised for L2 Italian is the TOP-V agreement occurring in clauses that
topicalise OBJ by (dis)placing it to the left of V from its canonical postverbal position (cf. § 3.2 on syntactic development). In such cases OBJ is an accusative clitic
pronoun co-referential with the TOP placed at the beginning of the clause. These
(i.e., the NP TOP and the OBJ clitic) must agree in number and gender values, as
in (12). Furthermore, if the V is in an analytic construction with an AUX, its past
participle form will bear the same number and gender values as the clitic, as in (13).
This structure then requires that learners recognise a full nominal TOP as
nonSUBJ, and mark their discourse-pragmatic choice explicitly by the ACC clitic
pronoun exhibiting NUM and GEN values in agreement with the NP TOP.
Learners who can produce such long distance agreement must clearly be able to
assign SUBJ and OBJ functions, and manipulate their agreement and position patterns. More about this complex structure will be said when presenting the development of Italian syntax in § 3.1, and in chapter 8.
(12) Toni: i broccoli
compra il cane
the broccoli-PL.MASC them-PL.MASC buys
the dog
[the broccoli, the dog buys it]
(13) Amy: le patate
the potatoes-PL.FEM them-PL.FEM has
[the potatoes, the dog has bought them]
la il cane
the dog
Bruno Di Biase and Camilla Bettoni
Finally, at the S-BAR procedure stage, subordination phenomena at the interclausal
level can affect morphological form also in Italian, as they do in English (cf. § 2,
ch. 2 in this volume). In (14) and in the appropriate cell in (3) above there are a
couple of examples.
(14) speriamo
hope-INDICATIVE.1.PL come-SUBJUNCTIVE.3.SG tomorrow
[we hope she comes tomorrow]
These constructions, which show a marked form of V in the subordinate clause,
belong to an educated register of standard Italian; they are rather rare in native
speakers’ everyday production, and difficult to elicit in learners. As already noticed
with regard to L2 English in chapter 2, learner data has yet to support the late
emergence of this last stage for L2 Italian. Indeed, the very complex area of subordination is still a research gap in PT, and requires further theoretical elaboration
and focused empirical investigation.
2.2. Hard barriers and soft barriers in morphological development
Evidence in support of the learner’s journey outlined in § 2.1 is provided by
two studies: Di Biase & Kawaguchi’s (2002) cross-sectional study of six
English-speaking Australian university students learning L2 Italian, two each
attending beginner, intermediate and advanced courses; and Bettoni & Di
Biase (2005) longitudinal study of a seven-year old Romenian girl recorded for
eight times after her arrival in an Italian school. In (15) we extend the analysis
presented in Di Biase & Kawaguchi (2002) by adding further phrasal and
interphrasal structures, and focus on the morphosyntax involving the most frequent, consistent and reliable form-to-function marking and agreement phenomena. The oral production data set from the six learners (and one native
control interacting with the researcher, not shown in the table) was elicited
over two sessions totalling between 35 and 60 minutes for each learner. The
first session included free conversation, a picture description and a story telling
task. A shorter second session focused on a communicative task devised by Di
Biase to elicit structures with OBJ clitics, which he called the ‘animal dinner
task’. This task shows the learnrs two cards at a time, one with the picture of
animals, and the other with that of food items. When the two cards appear on
the computer screen, learners are encouraged first to plan who (the agent) is
buying what (the theme) for a forthcoming dinner in the (simple) present
tense, and then to retell the events by saying who has brought what in the (analytic) past tense. Learners have to do this starting from the card on the left.
Since this card randomly shows the animals and the food items, active SVO
3. The development of Italian as a second language
(15a) Cross-sectional study of morphological development in L2 Italian (extended and
updated from Di Biase & Kawaguchi 2002)
(15b) Scalability matrix of Italian morphological data shown in (15a)
Bruno Di Biase and Camilla Bettoni
sentences (e.g., l’oca compra le carote, ‘the goose buys the carrots’) should alternate with topicalised ones (e.g., le carote le ha portate l’oca, ‘the carrots, [them]
has brought the goose’) – although it is of course impossible to rule out the use
of passives, a more formal choice in Italian (e.g., le carote sono comprate dall’oca, ‘the carrots are bought by the goose’). The total data set thus gathered consists of about 30,000 words, of which half produced by the learners, with a
mean length of 10.3 words per turn.
The initial lemma access stage is omitted from the results because it is safely reached by all learners. The rest of the counting is exhaustive over the database. However, default agreements (e.g., the citation, and most frequent, masculine/singular form) are excluded because otherwise the acquisition criteria
would be clouded over by the bulk of the default itself, as we will explain where
it becomes relevant. For the organisation of the table, see the criteria in chapter
1, § 5, this volume.
Our empirical results for all six learners in (15a-b), with nine implicationally arranged structures generating a matrix of 54 cells, support the hypotheses formulated in (3), yielding a coefficient of scalability (Hatch & Lazaraton
1991: 204-16) of 100% if we consider only the three procedures. This means
that the principle of ‘information exchange’ and its operation across major linguistic units (categorial, phrasal and interphrasal) is robust. In chapter 1, § 5,
we called ‘hard barriers’ these boundaries that the learner needs to negotiate
from one stage to the next in order to be recognised as having achieved a particular developmental stage. Thus Trish has not yet crossed the hard barrier to the
category procedure stage. Lois has, but has not crossed into phrasal procedure
stage yet. Carrie successfully reaches the phrasal procedure stage but not completely. The other three learners can be said to be capable of handling S-procedure but, curiously, each of them seems to be at distinct points of the interphrasal stage. So, having said that the hypothesis is supported, if we consider all the
structures severally, our data yields a coefficient of scalability of 89%, which is
more than sufficient to reduce the possibility of chance in the sequence in which
these structures emerge.
How can one interpret this difference between a scalability of 100% for
structures collapsed within the stages and a lower scalability for them taken
individually? Notice that a similar pattern obtains also for English, as shown in
(3a), § 2, ch. 2, this volume. The answer is probably that, whereas there are categorical implicational relationships among the procedures described by PT, these
relationships are not as strong between specific structures belonging to each
stage. A refinement of the gross division by stages was foreshadowed by
Mansouri & Håkansson (2007) under the guise of ‘intrastage’ sequencing in
learner language. These scholars identify structures belonging within a single
stage as being acquired according to a specific pattern. More specifically, three
3. The development of Italian as a second language
different structures for marking definitness in Arabic and two in Swedish,
respectively, were found to emerge in a specific sequence. Mansouri &
Håkansson (2007: 115) then propose an expanded ‘Hypothesis Space’ model
for PT to account for “more complex linguistic and functional phenomena that
have implications for language processing”. Furthermore, from a methodological point of view, they indicate the desirability of identifying for any stage the
‘optimal structure’ for empirically testing PT-based predictions, because the
structures emerging later within a stage may not be good candidates for testing
theoretical predictions. We concur with Mansouri and Håkansson that it is desirable to find an ‘optimal’ structure to characterise the stage. A case in point is
the –i plural marking for Italian, the most frequent and reliable out of all the
possible markers for plurals. However, we propose that it is crucial to identify
the source(s) of complexity, so that the sequencing itself also becomes theoretically predictable (and falsifiable).
There are two main sources of language-specific variation that constrain
the learner’s progress over these barriers within a stage, which we called ‘soft barriers’ (cf. ch. 1, § 5, ch. 2, § 2, for English, and ch. 4, § 3.2, for Japanese). These
sources are the lexicon, namely, the bundling of lexical features and lexical mapping, on the one hand, and on the other hand, the discourse-pragmatic requirements expressed in the grammar, as originally identified in the PT extension
(Pienemann, Di Biase & Kawaguchi 2005). Both lexical features and discourse
requirements are hypothesised to absorb further processing cost according to
the number of syntactic nodes requiring unification and, consequently, requiring greater attention by L2 speakers (cf. Segalowitz 2003; Taube-Shiffnorman
& Segalowitz 2005). These two constraints can be seen at play in the results presented in (15), where the solid line represents the ‘hard barrier’ between one
stage and the next, and the zigzag line represents the ‘soft barrier’ of a hypothesised step within the stage. In other words, while the environment of feature
unification (i.e., the scope of information exchange) provides the basic constraint between one stage and the next, both the quality and quantity of the features requiring unification over two or more nodes and discourse requirements
(which require greater attentional resources) will further constrain progress
within a stage. Let us consider these two sources in turn.
With regard to the language-specificity requirement of particular lexical
items, focusing on our results in (15), we notice that the accuracy rate for some
morphological agreements at the higher stages is still poor with the more
advanced learners. Why is this so? On the one hand, some of the structures
identified as candidates for our hypotheses rely on analytical constructions
based on nontransitive Vs selecting AUX essere. In fact, along with Maiden
(1995: 150) and Schwarze (2009: 153), among others, we consider active transitive constructions which select AUX avere as default given their frequency,
Bruno Di Biase and Camilla Bettoni
consistency and high predictability.6 Because avere constructions do not require agreement with the lexical V, they are not very informative in terms of developmental stages in morphological agreement. On the other hand, there is a
small group of Italian Vs with unaccusative meaning (e.g., andare, partire,
nascere) that select essere AUX (‘to be’) in analytic forms (Maiden 1995: 150153). In these cases, as in copular constructions and with passives and reflexives, the participle form of V must agree in gender and number with SUBJ.
This type of concord is a laborious thing to learn and may motivate what we
call here a soft barrier within the stage for the learner to overcome, causing the
rather shaky accuracy we can see in the results of more advanced learners such
as Anne, Toni and Amy. Their lower rates in fact occur when attempting structures with essere AUX (unaccusatives, passives, reflexives) in an otherwise
strong agreement performance. The intermediate stage learner, Carrie, interestingly crosses over the hard boundary to achieve the interphrasal stage, with
agreements between copula and predicative adjective, but she does not produce any contexts for the essere AUX and V-to agreement. We agree with a reviewer of this chapter that the evidence here is weak because no production does
not necessarily imply no acquisition; however, all learners have performed
exactly the same tasks designed to elicit this structure and only the three more
advanced learners produced it successfully.
The effect of the second source of language-specific variation constraining the learner’s progress within a stage can be exemplified at the top of the
S-procedure stage. Because all of the three more advanced learners have crossed over the hard barrier between phrasal and interphrasal stages, the difference between them can be characterised by the soft barrier imposed by the
appearence of topicalised OBJ. That is, Anne does not topicalise, although she
attempts to use some passive constructions, as in (16), which, however, do not
involve clitics.
(16) Anne: mm la lattuga
the lettuce-SG.FEM is bought-SG.FEM
[the lettuce is bought by the dog]
dal cane
by the dog
Toni crosses the OBJ topicalisation boundary, as (15) shows, but does not
manage the long distance TOP-V agreement over the next soft boundary, which
Amy does, as (13) shows. It could be argued that the computation of the
discourse-pragmatic information together with the morphosyntactic informa-
6 Some intransitives (e.g., unergatives such as dormire, camminare) also select avere, and
do not require agreement. Like transitives, these Vs tend to have an agentive SUBJ.
3. The development of Italian as a second language
tion may indeed require the S-procedure but that TOP lay outside of it. In other
words, it could be argued that there should be a ‘hard’ boundary to separate
more clearly the discourse-pragmatic information from the rest of the S-procedure because, even though the linguistic unit is still the clause, in processing
terms the long distance agreement is computed over potentially discontinuous
constituents. For the moment it may be best to let the ‘linguistic unit’ define
the hard boundaries, and let the additional lexical or discourse information define the soft boundaries (some of these issues will be picked up again in § 3.1).
In any case, the coefficient of scalability of the same results for the same learners shown in (15) improves dramatically from 89 to 100% when soft barriers
are taken into account.
3. Syntactic development
Italian appears to assign a lesser role to syntax than to morphology in interpreting
GFs (e.g., SUBJ can be null, but always marked in verbal morphology). As Bresnan
(1998: 119) observes more generally, morphological forms will compete with and
pre-empt phrases that carry no additional information. If the syntactic structure
nodes do not bear additional GFs that distinguish them from the morphological
structures, they must be omitted. This explains why the numerous Italian word
order options are used more for mapping pragmatic and semantic information
than for conveying grammatical information.
3.1. The Prominence Hypothesis
Let us then look at how the Prominence Hypothesis applies to the development of
Italian syntax. We will present first the key features of Italian grammar on which
the predicted developmental trajectory is based, and then the actual trajectory for
L2 Italian declarative sentences, drawn from work by Di Biase and his collaborators (e.g., Di Biase 2005; Di Biase & Bettoni 2007; Bettoni, Di Biase & Ferraris
2008; Bettoni, Di Biase & Nuzzo 2009; Bettoni & Di Biase 2011). The development of Italian content questions is discussed and tested with empirical data by
Bettoni & Ginelli in chapter 8, this volume.
Like English, Italian canonical word order is SVO, as shown in (17). This
means that, in pragmatically neutral, simple, declarative sentences, SUBJ is the
default TOP, and OBJ is the default FOC.
However, since Italian – unlike English – is a nonconfigurational language, its
canonical word order can be freed up for mapping pragmatic and semantic information. In theory at least, all permutations of the core elements are possible, as
shown in (18), where the propositional meaning of all the sentences is ‘Pierino eats
Bruno Di Biase and Camilla Bettoni
(17) Canonical correspondences between a-structure, f-structure and c-structure for the sentence Pierino mangia i fichi (‘Pierino eats the figs’)
the figs’. In practice, however, those in (18c-f) are highly dependent on marked
prosody for interpretation and rarely used. OBJ topicalisation, as in (18b) and illustrated formally in (19), needs no special prosodic emphasis, because the “functional uncertainty” of the first NP (Bresnan 2001: § 4.8) is solved by a coreferential
OBJ clitic marker on V (both i fichi and li are MASC-PL).
(18) a.
Pierino mangia i fichi
i fichi li mangia Pierino
Pierino i fichi mangia
i fichi Pierino mangia
mangia Pierino i fichi
mangia i fichi Pierino
3. The development of Italian as a second language
(19) Noncanonical correspondences between f-structure and c-structure for the sentence i
fichi li mangia Pierino (‘the figs, Pierino eats them’)
If the inanimate nature of the first NP in this sentence will semantically rule out
the possibility of the figs doing the eating, confusion could easily arise if animacy
is shared by both the participants in the eventuality. This would happen in (20),
for example, if neither SUBJ nor OBJ is unequivocally marked. In configurational
languages such as English one can tell positionally which NP has which GF: in
declaratives, the one before V is SUBJ, and the one after V, if any, bears some other
GF. As we have just seen, Italian marks preverbal OBJ with a clitic agreeing with
TOP: in (18b) masculine and plural (li and i fichi), and in (20) masculine and singular (lo and il bambino).
Bruno Di Biase and Camilla Bettoni
(20) a. la mamma accarezza
il bambino
[mummy caresses the child]
b. il bambino
[the child, mummy caresses him]
la mamma
If, on the other hand, it were the child doing the caressing (as in 21), and for discourse-pragmatic reasons the speaker would wish to place this NP in postverbal
focal position and place mummy in topical position (as in 21b), then the clitic signalling all this would agree with mummy and be feminine and singular.
(21) a. il bambino accarezza
la mamma
[the child caresses mummy]
b. la mamma
[mummy, the boy caresses her]
il bambino
Besides being nonconfigurational, Italian is also a null-SUBJ language. In spoken
Italian, SUBJ is neither referential nor pronominal in almost 70% of sentences
(Bates 1976), and is expressed exclusively by verbal morphology. This means that,
when previously mentioned, implied or shared, SUBJ is left out, as in (22a).
Should pronominal SUBJ be used, it would indicate emphasis or contrast, and
therefore most often occupy the postverbal focal position in declarative sentences,7
as in (22b), rather than its canonical preverbal position.
(22) a. hai
sentito i ragazzi?
have-2.SG heard
the boys?
[have you heard from the boys?
have-3.PL phoned
they have phoned]
b. hai
chiamato tu i ragazzi?
have-2.SG called
you the boys?
[was it you who has called the boys?
telefonato loro
have-3.PL phoned
it was them who have called]
7 Because Italian allows for flexible word order, it is a non-plastic language, i.e., it does
not allow variable placement of nuclear stress (Vallduvi & Engdahl 1996), in contrast
to English, for example.
3. The development of Italian as a second language
Even from this brief presentation of Italian word order rules, it is easy to see that
learners will be able to acquire them all only gradually. Let us then illustrate
their path hypothesised for syntactic development with regard to word order (cf.
(23) Developmental stages hypothesised for L2 Italian syntax based on the Prominence
Hypothesis: declaratives (after Bettoni & Di Biase 2011)
After the initial stage, when single concepts are mapped to single words or formulas, learners will at the next stage organise words according to the order most typically and most frequently recurring in the L2 input. This is SVO for Italian, including the possibility of VO, or V, as we have seen. Notice however that, at this stage,
the canonical-order sentence remains underspecified regarding the grammatical
functions of its core referents. That is, learners will analyse the preverbal NP, if present, semantically as agent or pragmatically as TOP rather than purely grammatically as SUBJ.8 Recall also our comment on the categorial stage in morphology in
§ 2 above: formal variation on V may begin to emerge in Italian interlanguage at
this stage – that is, much earlier than in English interlanguage. So, whereas the
English –s marker on third person singular emerges when the learner is able to
unify the relevant features of SUBJ-V agreement at the interphrasal stage, a variety
8 As a matter of fact, given that SUBJ is claimed to be universal (cf. the Subject
Condition of LFG, Falk 2006: 98, 170), learners assume that one element has SUBJ
status. This would mean a mapping by default between SUBJ, TOP and agent.
However, at this stage SUBJ remains underspecified, with some of its Italian-specific
properties unrealised.
Bruno Di Biase and Camilla Bettoni
of morphological verbal markings begin to appear at this category stage, because –
thanks to the null-SUBJ character of Italian – there is no SUBJ for the V to unify
with. The sentences in (24)-(25) are typical of this stage.
(24) mia famiglia. fratello sorella
mangia dolce swedish
my family brother sister-SUBJ
sweet swedish-OBJ
[my family, brother sister eat swedish sweet]
(25) mangio sul lavoro
at work-ADJ
[I eat at work]
To move beyond the canonical word order stage, learners must assign grammatical
functions to sentence constituents independently of their fixed position within the
canonical string. PT’s Prominence Hypothesis predicts that they will begin to do
this by uncoupling TOP from SUBJ. The first step in this development happens
when learners begin to contextualise (in time, space, etc.) the core sentence by
adding an ADJ to canonical order. If for pragmatic reasons this addition is introduced as TOP occupying a position early in the sentence as in (26), there will now
be two preverbal constituents. This will force learners to distinguish between TOP
and SUBJ, change the relationship between the c-structure and f-structure, and
require additional processing procedures. With the addition of ADJ as TOP, however, the first position is occupied by a nonargument constituent, so the SVO
structure nudges intact to the right.
(26) mezz’ora dopo
mio padre
half an hour later-TOPADJ my father-SUBJ
[half an hour later my father catches the train]
il treno
the train-OBJ
Further in their development, learners will also be able to disrupt canonical word
order. Once again this will happen gradually: first when only one core argument is
in a noncanonical position, namely when SUBJ is postverbal, as in (27), and then
when both are assigned to a noncanonical position, namely when SUBJ is postverbal and OBJ preverbal, as in (28).
(27) compro io
il pane
the bred-OBJ
[I am going to /it’s me buying the bread]
(28) hindi
[Hindi, everybody knows it]
3. The development of Italian as a second language
Supporting evidence for this trajectory comes from several corpora, gathered in
Australia and in Italy. We summarise here the results of the analysis of a crosssectional data set of 15 learners working or studying in Verona or Milan (Bettoni
& Di Biase 2011). They are all in their twenties (9 females, 6 males), with a wide
range of competence levels in L2 Italian and L1 backgrounds, including Arabic,
Czech, English, French, German, Japanese, Mongolian, Portuguese, Romanian,
Russian, and Tigrinya. Three tasks were used to elicit the structures listed in (29).
The first two, targeting canonical word order and ADJ topicalisation, although
standard in SLA, were adjusted for Italian in order to allow for production of
both overt SUBJ and null-SUBJ. Thus, first, a short picture story retelling task
involving at least two actors encourages the use of canonical word order with
both full referential or pronominal SUBJ and pro-drop. Second, a spot-the-difference task gives speakers the chance to contextualise the different actions
depicted in the two drawings by prefacing such phrasal ADJ as nel disegno a (‘in
drawing a’) and nel secondo disegno b (‘in the second drawing b’). Thirdly, for eliciting OBJ topicalisations we adapted Di Biase’s ‘animal dinner’ described in §
2.2, which became a ‘birthday party’, and a ‘trip into the mountains’ where various protagonists must contribute a present in the first case and a piece of equipment in the second. From the learners’ oral production data set, 512 declaratives
sentences could be extracted, about 34 sentences per learner. They are all main
clauses with lexical Vs requiring default mapping of thematic roles onto GFs,
such as dare (‘give’, a three-argument V), comprare (‘buy’, a two-argument V) or
saltare (‘jump’, a one-argument V). In other words, sentences with so-called
exceptional Vs and nonbasic V forms such as passives (Pinker 1984) are not considered here, because all of these involve nondefault mapping between a-structure and f-structure, and consequently PT’s Lexical Mapping Hypothesis. Nor
do we consider copular and presentative sentences, because they are ‘nonverbal
predicates’ (Kroeger 2005: ch. 10).
In (29) we arrange structures vertically from the bottom up with the three
groupings coherently reflecting our hypotheses, and learners horizontally from
the left according to developmental progress. In general terms, results shows an
implicational developmental pattern: all learners who have achieved the XP
canonical word order stage have acquired various instantiations of the previous
canonical word order stage. Then, likewise, those who have achieved the noncanonical (marked) stage have clearly achieved both previous ones. The fact
that the L1 background of our 15 learners is typologically varied makes our
evidence even more convincing. This of course does not deny the influence of
L1. Transfer does occur, but it is constrained by the procedures the learner can
handle (Pienemann, Di Biase, Kawaguchi & Håkansson 2005). As such, it
does not affect the developmental sequence, although it may affect the hypothesis space of a given stage, the speed with which it is attained, or the structur-
Bruno Di Biase and Camilla Bettoni
al preferences it manifests, as shown for RYA. Within this PT-based implicational developmental pattern, a closer look at the figures reveals further interesting patterns.
(29) Cross-sectional study of syntactic development in L2 Italian based on the Prominence
Hypothesis: declaratives (after Bettoni & Di Biase 2011)
All our learners are beyond the lemma access stage. Even GHI, who is the weakest,
produces several sentences with canonical word order, as also found in other studies of L2 Italian using different theoretical frames (e.g., Andorno, Bernini,
Giacalone Ramat & Valentini 2003). In first position these sentences include most
commonly the agent with default mapping, as in (30), but also – under the pragmatic constraints of our specific task – the theme with nondefault mapping, as in
(31). Unable to produce an OBJTHEME topicalisation required by the task we have
described above, GHI (like the other less proficient learners) produces instead a
clear example of functional assignment specified solely by position. Without
labouring the point here, this is a case where the interfacing between the morphological resources and syntactic progress still needs to be worked out in more detail
along both the Prominence Hypothesis schedule and the Lexical Mapping
Hypothesis one (cf. ch. 1, § 4.3, this volume).
(30) GHI lui
scito [=lasciato]
carta identità
identity card-OBJ
per scrivere
to write
[he has left (his) identity card (on the counter) in order to write]
3. The development of Italian as a second language
(31) GHI la torta
the cake-SUBJTHEME bring-INF
[the cake the policemen take]
i carabinieri
the policemen-OBJAGENT
As we move from GHI rightwards in the table, there is significant variation
among the learners in the distribution of the different constructions belonging
to this canonical word order stage, concerning mainly three points: morphological accuracy of V inflection, the thematic roles in the SVO order, and use of
SUBJ vs null SUBJ. With regard to the first point, although a detailed analysis
on this set of data is not yet available, it is worth noting that learners to the left
of our interlanguage path have poor V inflection – marking mostly aspectual
but not grammatical person, unequivocally indicating SUBJ, as in (30)-(31).
On the other hand, inflection improves considerably as we move rightward in
the table. Secondly, as the figures for SVO show, the weaker learners produce
ungrammaticised topicalisations of themes in first position similar to GHI’s in
(31). On the other hand, learners to the right clearly avoid them. Thirdly, the
table shows how null SUBJ tends to be avoided by the weaker learners, who
have unreliable morphological means to assign SUBJ, and then progressively
increases rightwards. Here the only exception is RYA – an interesting case, most
likely due to the fact that his L1 is English, which requires overt SUBJ. In any
case, most learners overuse pronominal SUBJ, as he does in (32), where lui
(‘he’) is redundantly repeated three times.
(32) RYA lui non poteva sentirla bene
lui ha detto a lei.. ah.. a lei
lui detto un
indirizzo falso
he not could hear her well
he has said to her ah to her
he said an
address false
[he couldn’t hear her well, he told her, ah, her, he told a false address]
All our learners except GHI have also reached the XP canonical word order stage,
and are able to front ADJ to the left of the canonical sequence. The reader will
notice that in (29), as well as the structure with SUBJ, as in (33), we also count the
one with null SUBJ, as in (34), both by VAN. This needs an explanation because
if there is no SUBJ to compete with TOP, one could think that the evidence for
progress to this stage is missing. However, we claim that in a null-SUBJ language,
VO is also a canonical string.
(33) VAN nel disegno B
il ragazzo
in the drawing b-ADJ the boy-SUBJ goes-V
[in drawing B the boy rides a bicycle]
nella bicicletta
on the bicycle-OBL
Bruno Di Biase and Camilla Bettoni
(34) VAN nel disegno b
in the drawing b-ADJ look at-V
[in drawing B they look at the bags]
le borse
the bags-OBJ
Among our learners, three have progressed also to the highest noncanonical
word order stage. Being able to assign functions independently from position,
they now have the means to place OBJ and SUBJ in a position other than their
canonical one, and mark them with the correct morphological resources. In
this cross-sectional study we can now plot the developmental path learners traverse when faced with a task requiring them to begin the sentence with a role
other than the agent, and we can discern several phases. In a first phase, represented by GHI, MID, MAR, MAC, CHA, and VAN, learners most often
begin the sentence with the theme without grammaticising the topicalisation,
as we have seen in (31); at other times however they use canonical order but
choose to start with SUBJ as agent, thus ignoring the prompt from the task, as
in (35). In a second phase, represented by LEI, NAT, and RIC, learners start
to grammaticise the topicalisation of the theme by using a clitic, but the grammaticisation is incomplete in the sense that one, two or all three agreements
(SUBJ-V, TOP-clitic, and TOP-clitic-participle) are missing or not unified.
For example, in (36) LEI uses the clitic, but neither SUBJ-V nor TOP-clitic
agreement is in place; and in (37) RIC mixes features of passivation with those
of topicalisation.
(35) MAR lo stereo.. lo stereo... il cameriere porta il cameriere porta il stereo
[the stereo, the stereo, the waiter brings the waiter
brings the stereo]
(36) LEI
i fragoli
the stawberries-FM.PL it-FM.SG
[the strawberries [them] take the dancers]
(37) RIC la torta
the cake-FM.SG
it-FM.SG takes-3.SG
[the cake (it) takes the policemen]
le ballerine
the dancers-3.PL
dai poliziotti
by the policemen-3.PL
In a third phase, represented by EVA and HEL, learners produce some grammatical and some ungrammatical topicalisations. In the final phase, represented by
RYA, all topicalisations are targetlike; and even the third long distance, interphrasal
dependency is computed with the correct agreement between the clitic and the participle, as in (38).
3. The development of Italian as a second language
(38) RYA le calze
the socks-FM.PL them-FM.PL have-3.SG taken-FM.PL
[the socks (them) has taken the nurse]
the nurse-3.SG
Although a fine-grained analysis of the increasing morphological accuracy still
needs to be done, we consider this sequence an important confirmation that learners may shoot up syntactically in their word order development well before they
acquire the full morphological means to unequivocally mark all the GFs in their
3.2. The Lexical Mapping Hypothesis
Italian Vs lexically requiring nondefault mapping of thematic roles onto GFs are
not only numerous in type, but some of them also occur quite frequently as tokens
in everyday speech. Among this latter group, we find unaccusative Vs (e.g., arrivare
‘arrive’ and morire ‘die’), exceptional Vs (e.g., piacere ‘like’ and mancare ‘be missing’), and nonbasic V forms, such as reflexive Vs (e.g., lavarsi ‘wash [oneself ]’ and
arrabbiarsi ‘get angry’).
Although as formulas they can be used by L2 learners quite early – see, for
example, the ubiquitous mi piace (‘I like it’) –, all these Vs lexically requiring
nondefault mapping are acquired – that is, processed online with some grammatical accuracy – quite late, at the final nondefault mapping stage. Given the
great variety among them, it is not surprising that they may also be acquired at
different times within this last stage, as we have seen in the previous discussion
of morphological development. Factors affecting their sequencing include (a)
the number of arguments they subcategorise: see, for example, nascere (‘be
born’) vs piacere (‘like’), which require one and two arguments respectively; (b)
their information structure; that is, whether, in a pragmatically neutral context,
they require noncanonical word order in addition to nondefault mapping: see,
for example, passive V forms, which require topical (preverbal) SUBJ, as in (39),
vs unergatives, unaccusatives and most exceptional Vs, which prefer focal
(postverbal) SUBJ, as in (40), (41) and (42) respectively; (c) the nature and
function of their clitic particles in the case of the reflexive V family; that is,
whether the clitic particle is purely expletive, as in pentirsi (‘repent’), which does
not have a corresponding nonreflexive form *pentire, or semantically somewhat
more transparent, as in baciarsi (‘kiss [each other]’), which has a transitive form
baciare (‘kiss [someone]’) as well; furthermore, whether the reflexive construction is decausative, as in (43), entails OBJ reduction, as in (44), OBL reduction,
as in (45), or no grammatical function reduction but adds a pragmatic modality to the eventuality being described, as in (46).
Bruno Di Biase and Camilla Bettoni
(39) la pasta viene servita calda
[pasta is being served hot]
(40) ha telefonato la nonna
[the grandmother has phoned]
(41) è nata una bambina
[a baby girl is born]
(42) ai bambini piace la pizza
[pizza pleases the children = the children love pizza]
(43) il bus
si ferma
[the bus stops]
(44) si lava
[(he) washes (himself)]
(45) si leva la giacca
[(he) takes off the jacket]
(46) si mangia la pizza
[(he) eats (his) pizza (loving it)]
In (47) we propose our language specific developmental schedule for Italian based
on the Lexical Mapping Hypothesis discussed universally in chapter 1 (cf. § 4.2.2,
and ch. 2, § 3.2, for English).
At the default mapping stage, learners may soon lexically saturate all the roles
that are semantically relevant to their Vs, provided of course they store the approriate items in the lexicon. They will, however, be unable to map them onto the
appropriate GFs unequivocally, that is by morphological means. So, for Vs lexically requiring a further argument besides those mapped on SUBJ and OBJ, learners
will tend to leave it functionally unmarked, as in (48)-(49) by MAR and CHA
3. The development of Italian as a second language
(47) Developmental stages hypothesised for L2 Italian syntax based on the Lexical Mapping
respectively, two of the learners in the corpus mentioned above in § 3.1. Italian, in
fact – unlike configurational English – has no OBJθ, and requires both the beneficiary or source roles to be mapped onto OBL by means of a PP.
(48) MAR
(49) CHA
la ragazza
[the girl
la ragazza
è data mh
has given
ha dato
a Carlo
la chiave
the key]
la chiave
un uomo ahh dove San Pietro
a man
where Saint Peter (is)]
a un uomo
dove è San Pietro
Bruno Di Biase and Camilla Bettoni
Likewise, any V types may be used early by beginners. The crucial point, however,
is again that in certain cases the outcome will be ungrammatical or pragmatically
inappropriate because at this stage Vs are treated as mapping the highest thematic
role onto the highest GF in the first linear position. So, besides felicitous outcomes
with ‘regular’ Vs, as in (50), beginners will produce sentences such as that in (51),
where MID uses the exceptional V (piacere, ‘is pleasing’) ignoring that in Italian its
experiencer role requires to be mapped as TOPOBL.
(50) MAR la signora
[the lady
(51) MID
i pagliacci
target ai pagliacci
la:: scarpe
the shoes]
le caramelle
At the next stage up, learners begin to mark the GFs of arguments other than
SUBJ and OBJ more clearly, and produce sentences such as those in (52)-(53).
Here, although the prepositions are lexically inaccurate (per and di respectively
instead of a), the very fact of using a PP marks the function of the constituent as
OBL, that is as neither SUBJ nor OBJ (the two core GFs of the canonical order).
We assume, as explained in chapter 1, § 4.2.2, that this is a necessary intermediate stage before learners acquire the means to map arguments onto GFs in an
unequivocal way.
(52) RIC il marito
ha detto [= dato] i fiori
has given
[the husband has given the flower to his wife]
(53) TAN la receptionista
per [= a] sua moglie
ha chiesto di [= a] Carlo
the receptionist-SUBJ V
[the receptionist has asked Carlo for his identity card]
la mh sua carta
di identità
his identity cardOBJTHEME
At the final nondefault mapping stage learners will then be able to assign arguments to any thematic roles as required by their lexical entries, and hence assign
prominence to thematic roles low in the hierarchy by mapping them onto
SUBJ, as in (54), or demote those that are high up by mapping them, for exam-
3. The development of Italian as a second language
ple, onto OBL or ADJ, or even suppress them, as in (55). These latter two
examples are fictitious.
(54) al ministro
i fondi
[the minister is lacking the funds]
(55) il presidente
è stato fischiato
[the president has been booed]
Empirical work testing our Lexical Mapping Hypothesis for L2 Italian has just
begun. Needless to say parts of our schedule in (47) are not yet supported by actual data. We mention here three studies in chronological order, which are heterogeneous for the variety of Vs and constructions they deal with. Bettoni, Di Biase
& Nuzzo (2009) deal with nondefault mapping requirements from the point of
view of the acquisition of postverbal SUBJ, and show that, within the last stage of
development, unergative and unaccusative Vs are acquired with their required
mapping before exceptional Vs. Nuzzo (2012) deals with the acquisition of passives, and clearly shows how long and complex the path toward their full acquisition can be, despite the fact that structurally they require canonical word order.
The point Nuzzo makes is that even learners who can produce SVO sentences
with a high degree of grammatical accuracy – as long as default mapping is required – may fail to do so when mapping requirements are nondefault. The explanation, we assume, is that they fail to integrate pragmatic and structural information
in order to give their sentences a suitable perspective so as to guide the listeners’
attention according to their communicative intentions. The third study testing
the Lexical Mapping Hypothesis is by Bettoni & Fratter (2013), and deals with
Italian reflexive V forms.
4. Conclusion
As we have seen, Italian is an interesting language for PT in so far as it presents different challenges compared to English, chiefly posed by its rich obligatory morphology licensing more flexible options in word order. Since the seminal work on
L2 Italian by Di Biase & Kawaguchi in 2002, the analyses mentioned in this chapter have not only added to our knowledge of the development of this language but
they have also paved the way for other nonconfigurational languages. More discussion and results will be presented in Bettoni & Ginelli’s chapter 8 on the syntactic
development of content questions.
Bruno Di Biase and Camilla Bettoni
This chapter also takes up the challenge of identifying the sources of sequencing regularities noted within a particular stage, and indicates both how they can
be conceptualised by identifying soft (intrastage) barriers as compared to hard
(interstage) barriers. Furthermore, this chapter offers a key for predicting their
sequencing. The quality and quantity of features to be processed at any given
moment and how specific DFs are involved will offer a principled key for going
beyond the more traditional PT predictions and offering a more detailed map for
possible language intervention. We have also indicated how looking at the variable behaviour of learners, i.e., at their accuracy in a principled way, rather than
ignoring it as soon as one can establish emergence of one particular structure, can
be quite revealing of more fine-grained developmental patterns. In any case more
work needs to be done, foremost in two directions, in our opinion.
First, with regard to obligatory structures, in a language with such a syncretic morphology as Italian, the challenges posed by the emergence criteria
need more attention, as Pallotti (2007) lucidly states. But there are other crucial
problems to disentangle along the long path from emergence to acquisition. For
example, in strict LFG terms, the two sentences in (56) would both belong to
the category procedure stage (with no information exchange outside the VP),
because LFG sees clitic pronouns as verbal morphology rather than separate
(56) a. glieli
he-DAT they-ACC.PL.MASC give-1.SG
[I give them to him today]
b. glieli
he-DAT they-ACC.PL.MASC have-1.SG
[I have given them to him today]
given-PL.MASC yesterday
Yet in order to produce them, learners will need to identify the GFs of all three
argument roles of the V dare (‘give’), and formally distinguish between OBJ and
OBLDAT – not to mention the added complexity of marking number and gender on the clitic pronoun and on the past participle. Could we then not assume
that learners will proceed from sentences with referential nouns or full pronouns
to those with clitics, as in (57a), and from just one clitic, as in (57b), to both of
them, as in (56 above) only after they have reached the stage of default mapping
and additional arguments in the Lexical Mapping Hypothesis hierarchy?
(57) a. dò
i libri
a lui
give-1.SG the books to him
[I give the books to him today]
3. The development of Italian as a second language
b. li
a lui
they-ACC.PL.MASC give-1.SG to him
[I give them to him today]
Could we not also assume that learners will produce accurate sentences first with
SUBJ morphology, then with clitic OBJ markers, and finally clitic with OBLDAT
ones? See, in this regard, the stage of default mapping and additional arguments in
(47), § 3.2. The rich nature of categorial markings in Italian offers a good area for
testing hypotheses of morphological development across soft barriers within stages,
and its interfaces with syntactic development.
Secondly, in terms of the acquisition of optional or alternative syntactic constructions, PT can be quite precise not only on structural optionality dealt with by
the Prominence Hypothesis, but also on the lexical-semantic requirements dealt
with by the Lexical Mapping Hypothesis. In this direction, Italian has a lot of
ground to offer to both theory development and empirical testing.
The development of Japanese as a second language
Satomi Kawaguchi
University of Western Sydney
1. Introduction
Japanese has an important role to play for testing and developing PT’s hypotheses because of its typological characteristics. With regard to the configurationality spectrum which has languages like English and Warlpiri at opposite ends (cf.
§ 2.2, ch. 1, this volume), Japanese is located far from configurational English
and closer towards nonconfigurational Warlpiri. Morphologically, Japanese is
rich in verbal inflections, but unlike Italian its morphological organisation is
agglutinating rather than fusional, and also unlike Italian its grammatical relations are marked on nominal constituents rather than on the verb, which makes
Japanese a dependent-marking language. In this latter sense it is similar to
Russian and Serbian (cf. chh. 5 and 6 respectively). Syntactically, Japanese is an
SOV, head-last language which allows great freedom in the ordering of nominal
constituents, as long as V is sentence-final. For a fuller description of the
Japanese language, written in English, readers are referred to Kuno (1973) and
Shibatani (1990). For a shorter survey, see Shibatani (2011).
This chapter represents a new development in the studies on L2 Japanese in
so far as, first, it updates the schedules for morphological and syntactic development outlined in earlier work according to the changes to PT proposed by Bettoni
& Di Biase in chapter 1 of this volume, and, second, in doing so, it explores the
boundaries of PT-based hypotheses on the acquisition of Japanese syntax. Work
on the acquisition of Japanese nominal and verbal morphology in the PT framework began with Doi & Yoshioka (1987, 1990), Huter (1996), Kawaguchi (2000,
2002), and Di Biase & Kawaguchi (2002). Then, Japanese L2 studies contributed substantially to the extended scope of PT. For example, Kawaguchi’s (2005)
longitudinal study was the first to test the two new hypotheses known as
“Extended PT” (i.e., the Topic Hypothesis and the Lexical Mapping Hypothesis
proposed in Pienemann, Di Biase & Kawaguchi 2005). Furthermore, Kawaguchi
(2007) reports on the acquisition of structures involving nonbasic V forms such
as passives, causatives and benefactives, as well as exceptional Vs such as unaccusatives, all of which lexically require nondefault mapping – I use here Pinker’s
Grammatical development in second languages, 149-172
Satomi Kawaguchi
(1984) expressions ‘nonbasic’ for V forms and ‘exceptional’ for V types introduced in ch. 1, § 4.2.2, this volume. Further, Kawaguchi’s (2009a) cross-sectional
study involving 24 intermediate-advanced university learners presents a detailed
analysis of the acquisition of causative constructions in relation to the Lexical
Mapping Hypothesis. So far, this hypothesis has been applied also to other languages such as English (Kawaguchi 2013; Keatinge & Keßler 2009; Wang 2009)
and Italian (Bettoni & Fratter 2013; Nuzzo 2012), but Japanese remains the only
one for which it has been shown how to treat causative constructions within the
PT framework. Kawaguchi’s series of studies deal with Japanese as a foreign language in an instructional setting involving university students in Australia, but her
results find support in other contexts, such as those on child L2 acquisition of
Japanese in a naturalistic environment (Iwasaki 2008), adult language acquisition
of L2 Japanese in an intensive course (Iwasaki 2013), and significantly bilingual
L1 acquisition in Japanese-English (Itani-Adams 2007, 2008, 2009, 2011, 2013).
The current state of PT as applied to L2 Japanese involves in particular the development of syntax and a reanalysis of the data from Kawaguchi (2010) with regard
to the Prominence Hypothesis, and Kawaguchi (2005) with regards to the Lexical
Mapping Hypothesis. About the latter hypothesis, Japanese as an L2 is the only
language besides English (cf. § 3.2, ch. 2, this volume) that tests empirically the
intermediate stage of default mapping plus additional arguments, formally introduced by Bettoni & Di Biase in chapter 1, § 4.2.2.
In the following sections each developmental schedule is preceded by a sketch
of the main characteristics of its respective Japanese target structures, and followed
by the pertinent empirical data.
2. Morphological development
Japanese is an agglutinating language, characterised as predominantly suffixing
rather than prefixing in its inflectional morphology (Dryer 2013). With regard
to verbal morphology, various affixes added to the stem of a V provide meanings
such as tense, aspect, mood, politeness, and polarity. In English a notion similar to agglutinative morphology is the addition of a morpheme (e.g., past –ed,
progressive –ing) to the lexical V, of a separate AUX (e.g., can, may, must), or a
separate main V (e.g., cause, want and seem for causative, desiderative and evidential constructions respectively). In Japanese these morphemes are attached
immediately after the V stem. When more than two suffixes are sequenced,
their relative order is generally fixed. For example, the passive suffix must precede the tense suffix, as in (1a), which is grammatical, whereas (1b) is ungrammatical.
4. The development of Japanese as a second language
(1) a. tabe-rare-ta
[(it) was eaten]
b. *tabe-ta-rare
Because Japanese morphology is agglutinating it is relatively easier to segment
for the learner than Italian fusional morphology. This can be seen in the example (1a) above where the verb stem (tabe- ‘eat’), the passive auxiliary verb (–rare),
and the verbal morphological inflection expressing past tense (–ta) are all uniquely identifiable. On the other hand, complex morphophonological processes
are involved in affixation. The basic Japanese syllable structure is CV whereby
phonological rules intervene to avoid consonant clusters and vowel clusters in
the affixation process. For example, the initial consonant of the suffix is elided
when a consonant-initial verbal morpheme, such as the present tense suffix –ru,
is added to the consonant-final stem as in (2a). Similarly the initial vowel of the
suffix is elided when a vowel initial suffix, such as the negative suffix –anai, is
added to the vowel-final verb stem as in (2b) (Shibatani 1987). However, these
morphophonological processes lie outside the current scope of PT, so they will
not be mentioned further here.
(2) a. kak-u
b. mi-nai
[do/does not see]
With regard to nominal morphology, Japanese displays both structural and semantic case markers. Consistently with the characteristics of a head-last language, these
are all postpositional particles. There are four structural markers: –ga for nominative case, –o for accusative case, –ni for dative case, and –no for genitive case, as
shown in (3).
(3) Harumi-ga
imooto-no tegami-o Taro-ni
Harumi-NOM sister-GEN letter-ACC Taro-DAT
[Harumi gave (her) sister’s letter to Taro]
Then there are several semantic case markers, such as –de and –kara.
Satomi Kawaguchi
(4) a. Taro-ga
Taro-NOM pen-INSTR letter-ACC
[Taro wrote the letter with a pen]
b. Harumi-ga
Tokyo-kara ki-ta
Harumi-NOM Tokyo-ABL came-PAST
[Harumi came from Tokyo]
Other Japanese nominal morphemes include numeral classifiers, which are common in many East Asian languages such as Chinese and Korean. In Japanese, classifiers always follow numeral modifiers, and are chosen according to what type of
object is being counted. So, counting animals, or books, requires different classifiers, as (5a-b) show.1
(5) a. ni-hiki-no
two-CL-GEN dog
[two dogs]
b. ni-satsu-no hon
two-CL-GEN book
[two books]
A further, and highly characteristic, morpheme of Japanese nominals is the discourse marker –wa indicating primarily the sentence TOP, as in (6), but also
the sentence FOC. We will discuss this marker in § 3 below when dealing with
(6) sensei-wa
[teacher drinks coffee]
Finally, just as Japanese Vs do not mark SUBJ information regarding NUM, so Ns
do not mark NUM or GEN distinctions – although there are the plural N suffixes
–tachi and –ra, as in kodomo-tachi (child-PL, “children”) and boku-ra (1.PL, “we”),
but these are not productively used and can only be attached to a limited number
of lexical items. Likewise, just as there is no SUBJ V agreement, so there is no agreement between adjectives or numerals and nouns.
1 Japanese nominal classifiers have not been investigated in PT studies – a good area to
4. The development of Japanese as a second language
The developmental hypotheses for Japanese verbal morphology proposed
in Di Biase & Kawaguchi (2002) are shown in (7). Nominal morphology is not
dealt with in this chapter, except when it is deployed in syntax for marking GFs
(cf. § 3).
(7) Developmental stages hypothesised for L2 Japanese morphology (after Di Biase &
Kawaguchi 2002)
The developmental stages which can be hypothesised for Japanese verbal morphology using the framework of PT are described below. At the very beginning, like
learners of any other language, learners of Japanese also cannot activate any language-specific procedure, and are thus able to produce only invariant words such
as oishii (‘delicious’), or fixed expressions such as arigatoo (‘thank you’).
As soon as learners are able to activate the category procedure, lexical variation
results in some V inflection. In L2 Japanese the most common alternation is
between present tense and past tense, as in (8a-b). Furthermore – although, as
noted above, Japanese Vs do not inflect for person or number of SUBJ – other
Satomi Kawaguchi
Vstem–affix combinations can indicate the acquisition of lexical operations.
Among them, we find Vstem-NEG, as in (8c).
(8) a. tabe-masu
[(I/you/he/she, etc.) eat]
b. tabe-masi-ta
[(I/you/he/she, etc.) ate]
c. iki-mas-en
[(I/you/he/she, etc.) do/does not go]
Nominal morphology can also be activated at the stage of categorial procedure, and
post-nominal particles start to appear. However, the categorial procedure does not
permit the recognition of grammatical functions of NP in a sentence. Therefore,
the learner identifies nominal morpheme –ga (NOM) to mark agent-like NP, and –o
(ACC) to mark patient- or theme-like NP, which are likely associated with the initial and pre-verbal position respectively. These markers will be taken up again in §
3.2 when dealing with the Lexical Mapping Hypothesis.
Phrasal morphology emerges at the next stage. A Japanese V can combine
with another V. In such case, the first needs to be a gerund (–te) (Kageyama
1999), where –te functions as complementiser (COMP). V-te V is an example of
phrasal procedure because information exchange is required between two Vs in
terms of the ‘combinatoric TYPE’ feature whereby the main lexical V (head element in VP) takes gerundive form in order to combine with the auxiliary V (Sells
1995; 1999). Sells explains that the lexical feature TYPE holds crucial information for verbal projection (i.e., phrasal syntax). In V-te V construction, TYPE of
V-te is V-sis(ter). This means that V-te has to take another V as its sister. Thus,
the construction V-te V requires information unification between two Vs in
terms of TYPE, which means that its production requires the phrasal procedure.
The Vstem-complementiser V construction (V-te V) is one of the ways in which
two Vs can combine to add progressive aspect (V-te imasu, “be V-ing”), as in (9a),
trial (V-te mimasu, “try V-ing”), as in (9b), and request (V-te kudasai, “please V”),
as in (9c).
(9) a. hasit-te i-masu
[(I/you/he/she, etc.) is running]
4. The development of Japanese as a second language
b. hasit-te mi-masu
run-COMP try-POL
[(I/you/he/she, etc.) try running]
c. hasit-te kudasai
[please run]
At the next stage, Japanese requires the activation of the S-procedure for marking
GFs of NPs in sentences involving nondefault mapping between a-structure and fstructure, such as passive, causative and benefactive constructions. Thus, the morphological operations at the S-procedure stage will be discussed in § 3.2, which
illustrates syntactic development at the level of the sentence, where nondefault case
marking of Ns interacts with the predicate of the sentence.
At the S-BAR procedure stage, PT hypothesises one operation in Japanese
morphology, which is the marking of the GEN case on SUBJ in a N-modifying
clause. SUBJ in Japanese is usually marked as NOM, but in a relative clause it can
be marked as NOM or GEN, as exemplified in (10) – a phenomenon termed Ga/No
Conversion by Harada (1971). This requires the learner to be able to distinguish
between the subordinate clause and main clause.
(10) Kumiko-ga/no
Kumiko-NOM/GEN teach-PAST
[(I) saw a student who Kumiko taught]
Evidence in support of the morphological schedule outlined in (6) comes mainly
from two longitudinal studies and two cross-sectional studies involving 28 adult
learners of Japanese as an L2. Further details of these four studies can be found in
Kawaguchi (2010). The data in (11) illustrates the results of one of these studies –
a three year longitudinal study of a learner codenamed Lou – used here to exemplify the morphological development observed over the fuller data set. Lou is a
young woman, native speaker of English studying L2 Japanese at an Australian university. Her data was collected at 12 points in time over her three years as an undergraduate, that is, twice a semester over six semesters, yielding a total of about 3,200
token words and 1,345 type words.
Note that the data in (11) includes Lou’s development for verbal morphology
up to the VP procedure stage. In fact, as we have just mentioned, structures requiring the S-procedure involve nondefault mapping and they will be discussed in §
3.2. Further up, the hypothesised Japanese L2 morphology at the S-BAR stage has
yet to be empirically tested, just as is the case of L2 English and L2 Italian.
Satomi Kawaguchi
(11) Lou’s morphological development (after Kawaguchi 2010)
As (11) shows, at t1 Lou already produces several inflectional forms of V: the POL
form 7 times, the POL-Q form twice, and the PRES form once. In this session, she
produces the POL inflectional form with three different Vs. Furthermore, the Vs
hanasu (‘talk’) and iru (‘exist’) are realised in two different inflectional forms, namely, hanasi-masu (talk-POL) and hanas-u (talk-PRES), as well as i-masu (exist-POL) and
i-masu ka (exist-POL Q) respectively. Thus, at t1, with both lexical and formal variation, the emergence criterion is satisfied. Thereafter Lou continuously shows V
inflection throughout the 3-year longitudinal study.
The phrasal procedure stage is reached at t4, when Lou produces the V-te V
structure 8 times, as in (12), where –de is an allomorph of –te; thereafter she uses
this form consistently with a variety of lexical items.
(12) t4 futari kodomo-ga ason-de i-masu
two children-NOM play-COMP PROG-POL
[two children are playing]
4. The development of Japanese as a second language
Thus Lou’s developmental sequence for verbal morphology shows an implicational pattern, indicating that the hypothesised hierarchy is supported with 1.0 scalability. This same pattern is confirmed for child L2 Japanese JSL (Iwasaki 2008) and
for adult L2 Japanese (Iwasaki 2013), as well as for English-Japanese bilingual L1
acquisition (Itani-Adams 2013).
3. Syntactic development
As mentioned in § 1, Japanese is an SOV, head-last language allowing great freedom in the order of nominal constituents as long as V is sentence-final. Moreover,
Japanese is interesting to PT for its rich range of V structures (such as passives,
benefactives, causatives) requiring nondefault mapping of a-structure onto f-structure. Accordingly, after the very first stage of single words and formulas, PT organises the learning constraints for syntax around the two hypotheses sketched out first
universally in chapter 1, § 4.2 in this volume, and then for L2 English and L2
Italian in chapters 2 and 3 respectively. In fact, unlike Pienemann, Di Biase &
Kawaguchi in 2005, I now believe with Bettoni & Di Biase that it is crucial to separate word order issues from mapping issues at the initial stage of L2 syntactic
development. For this reason, I will now illustrate the Prominence Hypothesis in
§ 3.1 and then the Lexical Mapping Hypothesis in § 3.2 for L2 Japanese including their respective initial states, thus making Pienemann, Di Biase & Kawaguchi’s
(2005) Unmarked Alignment Hypothesis redundant.
3.1. The Prominence Hypothesis
As a dependent-marking language, Japanese deploys case markers on Ns to indicate their grammatical relations with V (i.e., GFs). The marker –wa indicates discourse relations (i.e., DFs) of topic, focus and contrast within the sentence (cf. §
2.2, ch. 1, in this volume).
With regard to GFs, by default the NOM case marker –ga is associated with the
GF SUBJ, the ACC marker –o with OBJ, and the DAT marker –ni with
OBLRECIPIENT/GOAL. Being thus marked by case particles, GFs may be placed
variably in c-structure – a phenomenon called ‘scrambling’ in Japanese. The examples below show scrambling in contrast to canonical order, where (13a) exhibits
canonical order – SOV in Japanese –and (13b) a scrambled order in which OBJ is
linked to the prominent initial position.
(13) a. Harumi-ga
yon-de age-ta
[Harumi read a/the letter to Taro]
Satomi Kawaguchi
b. tegami-o
yon-de age-ta
[Harumi read a/the letter to Taro]
In theory, more permutations in c-structure are possible in Japanese as long as V is
in final position. Scrambling brings a semantic effect of mild emphasis on the
fronted constituent among other effects such as aiding interpretation of
antecedent-pronoun relationship (Shibatani 1990: 261). When speakers wish to
add prominence to their GFs, they use the discourse marker –wa in preference to
position (Tsujimura 1996).
So, with regard to the DFs, Japanese (like Korean and a few other languages) has a special marker (–wa) for NPs to mark discourse prominence relations within the sentence. Most often –wa indicates TOP as it “singles out an
accompanying noun as the topic of the sentence” (Tsujimura 1996: 135), but
–wa marked phrases should not be defined just as ‘topic’ because this marker
may also indicate focus and contrast (Vermeulen 2009). Notice here, importantly, that such morphological marking of DFs does not require that the –wa
marked function be in sentence initial position, as shown in (18) and discussed
below. Topic, Focus and Contrast are notions of information structure that interact with syntax (e.g., Frey 2004), and it has been recently argued that contrastive focus should be analysed as a composition of the notions [focus] and [contrast] (Vermeulen 2009: 335). This issue is clearly beyond the scope of this
chapter, but see chapter 1, § 4.2.1, this volume where information structure features such as [±Prominent] and [±New] from Choi (2001) are used to constrain
the TOP-FOC and their relative prominence. In this connection it may be
worth noticing that, interestingly, –wa is called a ‘TOP marker’ in English,
whereas in Japanese it is called toritate jyoshi, which can be translated as ‘bringing up into prominence’ – surely a more suitable descriptor given the use PT
makes of it in this volume (cf. the Prominence Hypothesis proposed in (34)(35), ch. 1, § 4.2.1.).
Since every DF must be linked to a GF, –wa can overlay a range of GFs. For
example, –wa can bring into prominence the core GFs, namely SUBJ2 or OBJ, as
in (14) and (15) respectively, a nonargument GF, namely ADJ, as in (16), just as it
can mark prominence on any other argument function.
2 The reader may recall from § 2.2, ch. 1, this volume, that most languages do not distinguish explicitly TOP from SUBJ, which leads to an interpretation of SUBJ as TOP
by default.
4. The development of Japanese as a second language
(14) watasi-wa
kono tegami-o
I-TOPSUBJ yesterday-ADJ this letter-ACC.OBJ write-PAST
[I wrote this letter yesterday]
(15) kono tegami-wa
watasi-ga kai-ta
this letter-TOPOBJ yesterday-ADJ I-NOM
[this letter yesterday I wrote]
(16) kinoo-wa
kono tegami-o
yesterday-TOPADJ I-NOM.SUBJ this letter-ACC.OBJ write-PAST
[yesterday I wrote this letter]
Notice also that when TOP marks the core arguments SUBJ or OBJ, –wa replaces
the NOM or ACC case markers altogether, which supports LFG’s differentiation of
core from noncore GFs. In fact, in noncore GFs the prominence marker –wa is
added to the case-marked nominals, as the grammaticality of (17) can attest, in
which a noncore DAT OBL already marked with –ni is attributed prominence in
the sentence by the addition of the –wa marker.
(17) Harumi-ni-wa
kono tegami-o
Harumi-DAT.OBL-TOP yesterday-ADJ I-NOM.SUBJ this letter-ACC.OBJ write-PAST
[to Harumi I wrote this letter yesterday]
What is the difference, then, between the sentences in (15) through to (17) and
their possible equivalent without the –wa marker? That is, with kinoo-wa simply as
kinoo in (14) or watasi-wa replaced by watasi-ga in (15), tegami-wa by tegami-o in
(16), and Harumi-ni-wa simply by Harumi-ni in (17)? The answer involves prominence. Let us consider (14): the sentence with watasi-wa would answer the question
‘what did you do?’, where the topic-comment structure indicates that TOPSUBJ has
been established previously. On the other hand, the equivalent sentence with
watasi-ga would answer the question ‘what happened?’, where the SUBJ-predicate
structure is ‘event-reporting’ and the whole sentence is ‘new information’. Hence,
when speakers do express TOPSUBJ they mean to identify it unambiguously.
Likewise, the sentence with tegami-wa would answer a question about the letter, and
its equivalent with tegami-o a more general one about the whole event. With these
examples we can see clearly that, although there is a universal preference for TOP
to be encoded, as Levelt (1989: 260) says, in a “syntactically prominent position”,3
3 Levelt goes on to explain that “syntactically prominent […] can mean that the topic is
encoded as a grammatical subject […]. It can, alternatively, mean that the topic will be
encoded early in the sentence, whether or not in the role of subject.”
Satomi Kawaguchi
position in c-structure is not the only defining criterion in Japanese. Indeed, on the
one hand, morphology may support prominence assigned by sentence-initial position, as confirmed by the grammaticality of (17), where the OBJDAT Harumi-ni is
marked as TOP with –wa, so that we know that topical prominence is assigned to
this participant; such topicalised NPs usually does occupy a clause initial position
in Japanese. On the other hand, in the competition between syntactic position and
morphological marking of TOP, it is the morphology that wins out in establishing
topicality, as (18) shows. In this example, TOP is not the initial ADJ san-nen maeni (‘three years ago’), but, unambiguously in Japanese, the NP watasi-wa (I-TOP).
So the morphologically –wa marked GF, and not the initial GF, win out as the topic
of the sentence.
(18) san-nen mae-ni
three-year ago-at-ADJ I-TOP
[three years ago I went to Japan]
As mentioned above, –wa may mark FOC, as well as TOP, which explains why the
Japanese expression for it (toritate jyoshi, ‘bringing up into prominence’) is more
appropriate than the English gloss, which conflates the two into TOP. In fact in
Japanese it is quite possible to mark more than one NP with –wa. In this case, the
initial –wa is TOP and the other is commonly assumed to be FOC (Kuno 1973:
30ff), as in (19). In this example we can see again how morphology and syntax
interact in marking GFs, and how position becomes relevant (first TOP, then FOC
– as confirmed by Peter Sell, personal communication, 2005) because morphology is opaque, in so far as both TOP and FOC are marked with –wa.
(19) watasi-no e-wa
onnanoko-wa akai doresu-o ki-te i-masu
picture-TOPADJ girl-FOCSUBJ red dress-ACC wear-COMP PROG-POL
[in my picture it is the girl who is wearing a red dress]
Significantly, in discussing the Japanese –wa phrase, Shibatani (1990: 277) highlights a difference between forms that “do represent a true experiential judgement
structure”, such as the –wa marked SUBJ and OBJ in (14) and (15), and those that
do not, such as the –wa marked postpositional phrases and adverbials in (16)
through to (19). The latter are not true experiential judgement topics, but in his
view they exploit the highlighting function of the –wa marker of true topics to set
themselves apart from the rest of the sentence. Hence Shibatani (1990: 277) concludes that “(t)he adverbial topic, including a postpositional one, is basically a stylistic topic.” This view would have definitely caused problems with the original
Topic Hypothesis (Pienemann, Di Biase & Kawaguchi 2005), but it is compatible
with the current Prominence Hypothesis (cf. ch. 1 § 4.2.1, this volume), which
4. The development of Japanese as a second language
embraces both TOP and FOC with their information structure features
[±Prominent] and [±New] (Choi 2001) to ensure a clear demarcation between DFs
and prominence assignment.
This brief discussion of the –wa particle does not do justice to the complexities of its uses in Japanese native speakers’ speech. As a matter of fact, the whole
issue of the topicality vs focality of –wa is still controversial – involving as it does
such areas as topic continuity, nominal ellipsis and noncanonical case-marking. A
parallel case can be made for the function of ‘scrambling’ in word order to which,
controversially, is also attributed no obvious function (Hayashi, Tomlin & Yokota
2002), or in any case an “inconclusive” one (Kuno 1973), or, alternatively, one of
“mild emphasis” (Shibatani 1990).
Given these complexities, it is not surprising that learners will only gradually acquire the means of expressing GFs and assigning prominence to the NPs in
their Japanese sentences. Here I will first illustrate in (20) how learners of L2
Japanese are hypothesised to acquire the encoding of TOP in declaratives, and
then report in (21) the results with respect to declaratives using the same longitudinal study that provides the data for morphological development in § 2 (cf.
Kawaguchi 2010). In (21), all the sentences involving lexical verbs with at least
one argument are analysed. Exceptions are copula and presentational verbs such
as naru (‘become’) and aru/iru (‘there is~”) and verbs taking XCOMP such as
omou (‘think’) and iu (‘say’), which is consistent with the analysis in English and
Italian respectively in chapters 2 and 3, this volume. Further evidence supporting
the hypothesised schedule in (20) comes from two other studies: Kawaguchi
(2005) and Itani-Adams (2009).
(20) Developmental stages hypothesised for L2 Japanese syntax based on the Prominence
Hypothesis: declaratives (after Kawaguchi 2005)
Satomi Kawaguchi
(21) Lou’s syntactic development based on the Prominence Hypothesis: declaratives (re-analysis of Kawaguchi 2010)
The Prominence Hypothesis predicts that, after the initial stage of single words and
formulas, the learner can produce canonical word order, which cognitively speaking represents the most harmonious linking between c-structure and f-structure. In
Japanese, canonical word order yields the sequence SOV or TOPSUBJ OV as in
(22). Because Japanese allows nominal ellipsis, canonical word order also includes
V-last structures with at least one core argument, either SUBJ or OBJ, as in (22a)
and (22b) respectively.
(22) sensei-wa
teacher-TOPSUBJ coffee-ACC.OBJ drink-POL
[teacher drinks coffee]
(23) a. sensei-wa
teacher-TOPSUBJ drink-POL
[teacher drinks (it)]
b. koohii-o
[(she) drinks coffee]
In Lou’s data, SUBJ is initially elipted in t1 and t2. Thus she uses only one argument (i.e., OBJ) when required in these two sessions. SUBJ marking with –wa
emerges at t3 and is then consistently produced throughout the longitudinal study.
The next stage, the XPDF canonical word order stage, is characterised by the
learner’s ability to place ADJ in sentence-initial position, so as to express contextual information (time, place of the event, etc.). This operation, which does not disturb canonical order, triggers however a disengagement of SUBJ from its canoni-
4. The development of Japanese as a second language
cal first position in the clause. ADJ may then be TOP – conceding that it may be
a stylistic TOP, in Shibatani’s (1990: 277) term – and we can clearly see whether
learners are able to disentangle the TOP marker –wa from its initial exclusive association with SUBJ. So, in addition to being able to add a constituent to canonical
order, the learner can now assign to it the TOP function by marking it with –wa,
as in (19) above.
In Lou’s data, ADJTOP marked with –wa first appears twice in t3. However,
both sentences do not express SUBJ overtly, so that she starts producing the
ADJTOP S(O)V structure with an overt distinction between SUBJ and TOP only
at t4, as shown in (24). Marking thus either TOP or SUBJ by –wa, Lou now knows
that the sentence initial element can be ADJTOP, that is, a GF other than SUBJ.
(24) watasi-no e-wa
onnanoko-wa akai fuku-o
ki-te i-masu
I-GEN picture-TOPADJ girl-FOCSUBJ red dress-ACC.OBJ wear-COMP PROG-POL
[as for my picture, the girl is wearing a red dress]
At the next stage, the hypothesis predicts that learners can also assign prominence
to nonSUBJ constituents internal to SOV such as OBJ, as in (15) above. In order
to do this they need to disentangle the canonical association between the position
of OBJ and its case marker, which is ACC in pragmatically unmarked sentences. In
other words, OBJ marking with –wa as an alternative to –o, the ACC marker,
requires full functional assignment, namely, the ability to assign a GF or DF to
each constituent of the canonical word order independently of position.
In Lou’s data, –wa marked OBJ emerges at t4, with further uses at t10 and
t12. One example is shown in (25), where she first marks OBJ with –o (ACC) but
then corrects to –wa (TOP). This example clearly shows that Lou knows that –wa
may topicalise OBJ. Here we can see that, in assigning prominence to OBJ, the
canonical word order is not retained.
(25) um saakasu-no kippu-o um.. -wa ryoosin-ni
ticket-ACC-TOPOBJ parents-from receive-POL-PAST
[the circus ticket, I received (it) from my parents]
In sum, Lou’s data, however limited, supports the hypothesised schedule for the
Prominence Hypothesis in so far as she can assign prominence to constituents by
the use of –wa in an implicational sequence: first –wa marks SUBJ, the default
association between a DF and a nonDF, then ADJ, a nonargument GF, and finally OBJ, a core GF other than SUBJ. Here, clearly, only the surface has been
scratched, and much further work needs to be done, both in the description of
Japanese native speaker’s case and prominence assignment, as well as in that of
learners’ development.
Satomi Kawaguchi
3.2. The Lexical Mapping Hypothesis
The Lexical Mapping Hypothesis is particularly important in L2 Japanese, because many Japanese verbs and verbal constructions, including frequently used ones,
require nondefault mapping of thematic roles onto GFs – represented in LFG as
a-structure to f-structure mapping. Nondefault mapping may add a variety of
attributions to the event such as prominence, affectedness, causality, speaker perspective, or speaker affective participation. These attributions are expressed by the
speaker’s choice of V – which may itself require nondefault mapping of arguments
onto GFs (i.e., exceptional Vs) –, or of the V form that most closely conveys
his/her intention (i.e., nonbasic V forms, such as passives, causatives and benefactives). This (lexical) mapping grammatical mechanism must not be confused with
the mechanism involved in the Prominence Hypothesis, presented in the previous
section, which achieves prominence by alternative linear precedence relations on
the alignment of arguments – represented in LFG as c-structure to f-structure
As mentioned in chapter 1, § 2.2, this volume, default mapping, which is
assumed to require least processing effort, associates agent with SUBJ and patient
with OBJ. However, for discourse or pragmatic reasons, or given particular lexical choices, the speaker may highlight a thematic role other than the agent by
mapping it onto the highest-ranking GF, i.e., SUBJ (Keenan & Comrie 1977).
When nonagentive roles are mapped onto SUBJ, the default association of agent
with SUBJ is disrupted. According to Pinker (1984) such nondefault mapping
results from at least two sources: (i) intrinsically “exceptional” Vs such as receive in
English and morau (‘receive’) in Japanese; and (ii) nonbasic V forms such as passives and causatives.
Thus, exceptional Vs involve language-specific V features determined by
the specific lexical V selected by the speaker. Given their intrinsically languagespecificity and particular behaviour, there is no systematic way of identifying
them a priori, and learners must learn their argument specifications and mapping requirements one V at a time. Kawaguchi (2013) begins to examine the
acquisition of such Vs, but further investigation is required. On the other hand,
the study of the acquisition of nonbasic V forms is quite advanced for Japanese
(cf., e.g., Kawaguchi 2005, 2007, 2009a, 2009b, 2010) and may open the way
to their exploration in other L2s. As mentioned in chapter 1, § 4.2.2, this volume, the difference between Pinker’s intrinsically exceptional Vs and nonbasic V
forms is that, in the case of exceptional Vs, alternative forms, if available, are not
morphologically related (e.g., ageru/morau, ‘give/receive’), whereas in the case of
nonbasic forms, they may be morphologically derived from a more basic form
which exhibits default mapping (e.g., sikar/sikar-are, ‘scold/be scolded’). In any
case, whether or not alternative forms are lexically derived, current LFG consid-
4. The development of Japanese as a second language
ers them as separate lexical entries which select their own set of arguments.
Thus, in (26) the passive V sikar-are (‘be scolded’) only selects the SUBJ
kodomo-ga (‘the child’), as can be seen by the grammaticality of (26a). The
agent-specifying clause sensei-ni (‘by the teacher’) in (26b) turns out to complete
the information and is also grammatical, but it is the speaker’s choice to express
it or not, because grammatically it is not essential. On the other hand, in (26c)
the active V lexical entry sikaru (‘scold’) selects <SUBJ, OBJ> for its f-structure;
hence sensei-ga (‘the teacher’/agent/SUBJ) is required, as much as the other participant kodomo-o (‘the child’/patient/OBJ) is. Of course in Japanese either may
be dropped and represented by zero anaphora, but this depends on the discursive context and not on the V itself.
(26) a. kodomo-ga
child-NOM.SUBJ scold-PASS-PAST
[the child was scolded]
b. kodomo-ga
child-NOM.SUBJ teacher-ADJ scold-PASS-PAST
[the child was scolded by the teacher]
c. sensei-ga
teacher-NOM.SUBJ child-ACC.OBJ scold-HON-PAST
[the teacher scolded the child]
Passives can thus be counted among lexically productive alternative V forms, that
is, like other languages such as English and Italian (cf. ch. 2, § 3.2, and ch. 3, § 3.2,
this volume, respectively), Japanese typically exhibits passive constructions which,
among other things (such as revealing the speaker’s own stance on the event or
his/her affective engagement), may promote the patient role by mapping it to
SUBJ (cf. the formal illustration of the nondefault mapping of a-structure onto fstructure of an English passive sentence in (26), ch. 1, this volume) and demoting
the agent by mapping it onto an ADJ, as in (26b). However, the attribution of
prominence is not the only function of passives. In fact there is no prominence
involved at all in (26a): because there is only one argument (the theme, realised as
SUBJ), there is no competition. In such cases, then, the communicative function
of the passive construction is not prominence, but something else. The speaker, for
instance, may want to limit the scope of the eventuality exclusively to the theme,
because the agent is unknown or irrelevant. The choice of a passive construction,
then, is not at all parallel to, nor to be confused with, topicalization, which realises agents as SUBJ and theme/patients as OBJ regardless of the eventual order
adopted, and whose function is limited to the attribution of prominence for contrast or specification purposes.
Satomi Kawaguchi
To round up the LMH section here, I will present two other types of Japanese
nondefault mapping where prominence is not directly involved, that is, benefactives and causatives. Constructions involving the AUXs ‘give’ and ‘receive’, are collectively called ‘benefactives’. Backhouse (1993: 124-125) translates three types of
benefactive Vs as follows: (i) V-te ageru “(I/we, etc. do something for someone)”;
(ii) V-te kureru “(someone does something as a favour to me/us)” and; (iii) V-te
morau “(I/we, etc. receive the favour of someone doing something for us)”.
Examples of these three types are shown in (27). As we can see, in benefactive constructions the lexical V precedes the benefactive AUX that contributes the benefactive role, and we notice that all three constructions here exhibit the typical SUBJOBL-OBJ Japanese word order for three-argument Vs.
(27) a. give-schema (age-)
kodomo-ni hon-o
Mariko-NOM child-DAT book-ACC read-COMP give-PAST
[Mariko read a book to the child]
b. give-schema (kure-)
Mariko-NOM I-DAT
book-ACC read-COMP give-PAST
[Mariko read a book for me]
c. receive-schema (mora-)
kodomo-ga Mariko-ni hon-o
Mariko-DAT book-ACC read-COMP receive-PAST
[the child received the favour of Mariko reading a book]
In his LFG analysis, Ishikawa (1985) considers Japanese benefactive constructions
as functionally biclausal. Later work by Matsumoto (1996) and Shibatani (1994)
also support Ishikawa’s analysis. Following Ishikawa (1985: 152) and Matsumoto
(1996: 48) for the representation of the lexical entry and f-structure respectively, the
example of the benefactive construction in (27a) is represented in (28). As can be
seen, the benefactive predicate with ageru (‘give’) maps in a nondefult way the benefactor and beneficiary roles onto SUBJ and OBJ respectively, and additionally the
beneficial event role on XCOPM. In the f-structure SUBJ in XCOMP is linked to
the same GF (i.e., SUBJ) as in the matrix clause, as indicated by the joining line.
Benefactive constructions thus display nondefault mapping because they
involve a complex predicate where the f-structure of SUBJ in XCOMP can be
identified by reference to the value of an argument function in a higher matrix.
Without going into further detail here, also the benefactive constructions with
the morau (‘receive’) schema present nondefault mapping, but in the opposite
direction, that is, the beneficiary role is mapped onto SUBJ, the benefactive role
4. The development of Japanese as a second language
(28) lexical entry
ageru ‘give’: V
(↑PRED) = ‘ageru
< benefactor,
beneficial event >
onto OBL, and so on. Given such mapping arrangements, and the complexity
of structures generated by them, PT predicts that learners will produce them
only once they have reached the S-procedure stage, which gives them the morphological resources to mark GFs unequivocally.
Causative constructions, similarly to benefactive ones, display a complex
form-meaning relationship whereby the form expresses the meaning of ‘causing X
to do something or to be in some state’ (Shibatani 1990: 307; cf. also Noda,
Sakota, Shibuya & Kobayashi 2001). According to Matsumoto (1996), Japanese
causatives involve a causer controlled sub-event where the logical SUBJ of the
embedded clause is fused to the patient of the matrix clause. So, they involve nondefault mapping because one participant actually receives two thematic roles.
Alsina (1996: 86) assumes that “the causative verb and the base verb undergo predicate composition yielding one, single, complex, a-structure”. An example of a
Japanese causative construction is given in (29), more formally illustrated in (30).
Here Takashi plays a double role in the eventuality described in the sentence: he is
the patient of the causative V (Masako made Takashi…) and, at the same time, the
agent of the lexical V arau (…Takashi wash the car).
(29) Masako-ga
kuruma-o araw-ase-ta
Masako-NOM Takashi-DAT car-ACC wash-CAUSE-PAST
[Masako made Takashi wash the car]
(30) Mapping of a-structure onto f-structure for the transitive causative sentence Masako-ga
Takashi-ni kuruma-o araw-ase-masita (‘Masako made Takashi wash the car’)
Satomi Kawaguchi
The syntactic hierarchy based on PT’s Lexical Mapping Hypothesis for L2
Japanese is illustrated in (31), following the general proposal by Bettoni & Di Biase
in chapter 1, § 4.22, (42)-(43), and similar to the language-specific hierarchies for
L2 English in chapter 2, § 3.2, and for L2 Italian in chapter 3, § 3.2. In (32), Lou’s
data again provides the empirical evidence for the hypothesised hierarchy. Also
here, like for the Prominence Hypothesis in the previous section, copula and presentational Vs are excluded from the analysis.
(31) Developmental stages hypothesised for L2 Japanese syntax based on the Lexical Mapping
Hypothesis (after Kawaguchi 2005)
(32) Lou’s syntactic development based on the Lexical Mapping Hypothesis (updated analysis of Kawaguchi 2005, 2009b, 2010)
4. The development of Japanese as a second language
Lou reaches the default mapping stage with one occurrence at t1, and thereafter she
shows robust numbers of positive figures with rare errors in case marking, which
are indicated with a minus sign in (32). Examples of, respectively, positive and
negative occurrences at t8 are shown in (33) and (34).
(33) watasi-no
okaasan-wa uhm razania-o
lasagna-ACC cook-POL-PAST
[my mother cooked lasagna]
(34) haha-wa.... inu-o
mother-TOP dog-ACC eat-PAST
[literally: my mother ate the dog; intended: my mother (thinks) the dog ate (it)]
At the next stage (i.e., the default mapping and additional argument stage), the
propositional content is still expressed in a pragmatically neutral, default way.
This stage is reached when learners begin producing constructions consisting of
default mapping together with additional arguments which are differentiated
from SUBJ and OBJ, that is, when in Japanese they start using OBL arguments
with case markers other than NOM and ACC. Typically, as in Lou’s examples
shown in (35) and (36), they are able to map the nonSUBJ argument of an
intransitive V onto OBLLOC, and the third argument of a transitive V onto
(35) san-nen
watasi-wa nihon-ni
three-years before-when I-TOP
[three years ago I went to Japan]
(36) watasi-ni
Tomu-san ah hana-o
[to me Tom gave flowers]
Lou reaches this intermediate stage at t3 with four occurrences of the S OBL V
structure and soon thereafter, like at the previous stage, her figures are quite
consistenly positive. An example of her correct case marking is shown is (37).
(37) er fooku to naihu-o er (long pause) teeburu-ni...
fork and knife-ACC
[(I) arranged forks and knives on the table]
Finally, at the nondefault stage, learners will acquire the means to respond to
pragmatic motivations that require nondefault mapping choices. At this stage,
they may be able to produce benefactives, passives, and causatives. An example
Satomi Kawaguchi
of Lou’s successful deployment of each nondefault mapping construction is
shown in (38a-c), and one example of unsuccessful mapping of a passive construction is shown in (39).
(38) a. benefactive construction at t9
aisukuriimu-o kat-te ...
John-Mr-DAT ice cream-ACC buy-COMP give-POL-PAST
[(I) bought an ice cream for John]
b. passive construction at t9
densya-ni not-ta
doroboo-ni saifu-o
train-on ride-PAST when robber-DAT wallet-ACC steal-PASS-POL-PAST
[when (she) got on the train, (she) had her wallet stolen by a thief ]
c. causative construction at t10
uh bosu-wa watasi-ni
[my boss always asks me to make photocopy]
(39) passive construction at t9
inu-wa .. watasi-ni... kami-rare-masi-ta
[literally: a dog was bitten by me; intended: I was bitten by a dog]
In Lou’s data the benefactive construction is the first to emerge at this last stage.
The emergence of benefactives before passives is also confirmed by other Japanese
L2 researchers (e.g., Tanaka 2001). Benefactive constructions had been taught in
class just before their emergence at t5, so it is possible that teaching had an immediate if not lasting effect, because its use becomes more continuous only from t8
onwards. Lou’s benefactives are in any case all give-schema except one, when she
attempts, but fails, the V-te morau, receive-schema at t5, just after its introduction
in class. Why is the receive-schema produced less frequently compared to the giveschema? Is its processing more costly? The reason may lie in the higher degree of
nondefaultness with the receive-schema, which involves XCOMP’s SUBJ linked to
the OBJ source in the matrix clause, an operation that adds to the cost of online
speech production. So we have here an instance of a soft barrier in the sense mentioned in chapter 1, § 4.4, this volume, and chapters 2 and 3, §§ 2.2, for L2
English and L2 Italian respectively. Within this same nondefault mapping stage,
exceptional Vs emerge next in Lou’s data. However, being lexically idiosyncratic
and learned individually, there is no soft barrier marked in the table. On the other
hand, also passives and causatives are nondefault structurally, so a soft barrier in
marked in each case, respectively at t9 and t10.
4. The development of Japanese as a second language
In sum, naturally enough the dominant use of default mapping structures is
observed throughout Lou’s longitudinal study. In the first two interviews, that is in
the first three months of learning Japanese for her university degree, this learner
produces only default mapping structures. Then, after a further three months,
besides agent/experiencer and themes mapped respectively onto SUBJ and OBJ,
she is able to map additional arguments onto OBL. Finally, she attempts constructions with nondefault mapping at t5, in her second year, and uses two benefactive
constructions, one of which successfully. Passive and causative constructions
emerge at t9 and t10 respectively, that is, in her third year of studying Japanese for
her university degree.
4. Conclusion
Japanese is one of the languages that have been most explored in PT and whose
typological characteristics have contributed significantly to our understanding
of the cross-linguistic plausibility of the theory, and particularly the testing of
the early Topic Hypothesis and Lexical Mapping Hypothesis, as well as the current Prominence Hypothesis and the critical updating of the Lexical Mapping
Hypothesis. As a matter of fact, L2 Japanese was one of two languages to provide the earliest empirical evidence for a formal typological validation of PT’s
morphological development – the other language being L2 Italian (cf. Di Biase
& Kawaguchi 2002). Furthermore, Japanese was not only the first language to
test the Lexical Mapping Hypothesis (Kawaguchi 2005), but also to do so most
extensively (Kawaguchi 2010, 2013). Whereas other languages have so far provided evidence only for the development of passive structures, L2 Japanese has
provided it also for benefactives and causatives, further advancing their analysis
in this chapter.
This chapter presented the developmental stages in learning Japanese
incorporating the latest proposals in PT, with an updated analysis of a three year
longitudinal study. In particular, the Prominence Hypothesis is explored by
including morphological prominence, which is characteristic of Japanese. As for
the Lexical Mapping Hypothesis, it is further clarified and differentiated from
the Prominence Hypothesis, and an intermediate stage is added, with relevant
empirical analysis.
Broad areas within a PT framework, of course, still need attention in L2
Japanese, as well as in other L2s. These include the development of nominal morphology, interrogative sentences within the Prominence Hypothesis (a priority area
to be explored in L2 Japanese), and subordination, both in morphology and syntax, as well as the syntax-pragmatics interface. In this latter area, issues such as topic
continuity, nominal ellipsis and noncanonical case-marking are well worth investi-
Satomi Kawaguchi
gating from the learner’s point of view. Within the Lexical Mapping Hypothesis,
the acquisition of exceptional Vs requiring both nondefault mapping and noncanonical word order also needs to be addressed.
Exploring new issues
Camilla Bettoni* and Bruno Di Biase**
*University of Verona, **University of Western Sydney
Part III addresses some new theoretical and applied issues. The first five chapters
explore the scope of our newly formulated Prominence Hypothesis, and deal with
different languages and structures. They address the development of the case marking systems in Russian L2 and Serbian L2 (chh. 5 and 6 respectively), Differential
Object Marking (DOM) in Spanish (ch. 7), constituent questions in Italian L2
(ch. 8), and V2 in German L2 (ch. 9). The last two chapters widen the scope of
PT from an applied perspective and test whether PT schedules can hold in different conditions and situations, that is, the acquisition of Italian morphology by an
autistic child (ch. 10), and Japanese L2 structures in a CALL mode respectively (ch.
11). A brief summary of each chapter follows.
Chapters 5 and 6 present explorations within the PT framework of the development of the case system in Russian and Serbian, two nonconfigurational,
dependent-marking languages. For learners of these languages, case is a complex
feature to acquire for a variety of reasons: morphologically, there are many cases,
fusionally enmeshed with other nominal features such as number, gender and class.
Morphosyntactically, case must be computed on most nominal elements within
the NP. Then syntactically, at clause-level, case morphology itself constructs GFs
independently of phrase structure. Furthermore, if relations between case and function are default and predictable most of the time, at others the same case can construct alternative GFs, and the same GF can be constructed by different cases,
although with different lexical predicates. Given these complexities, the two chapters will suggest some hypotheses based on the Prominence Hypothesis, and test
them out on cross-sectional data. Chapter 5 on Russian analyses the interlanguage
of eight students learning their L2 in a foreign language context at the University
of Verona. Chapter 6 on Serbian deals with production data of three SerbianAustralian teenage bilinguals living in Sydney, a context of contact with a
(majority) language with a much-reduced expression of case, English. Results in
both chapters show that there is a direct relationship between the speakers’
availability of morphological case markers and their skills for deploying them to
exercise alternative pragmatically driven syntactic choices. More specifically, L2
speakers seem to progress from a first match between case-form and position to full
Grammatical development in second languages, 173-176
Camilla Bettoni and Bruno Di Biase
functional assignment by case independent of position. In other words, initial learners rely on more default case markers and fixed structures for their sentences; then
more advanced learners display both a fuller range of case markers, including nondefault ones, and the skills for deploying them according to their discourse-pragmatic needs.
Chapter 7 addresses the development of L2 Spanish, a relatively new language to PT, with the aim of investigating the acquisition of its DOM. DOM is arguably one of the most debated topics in Spanish grammar over the last 200 years, yet
little is understood in terms of its acquisition by L2 learners. There are good reasons for both these facts. Descriptive discussions must come to terms with such
diverse factors as animacy of OBJ (whether animate or inanimate), specificity of
OBJ (whether specific or nonspecific), form of OBJ (whether a proper N or a lexical N), and relative animacy, that is, the degree of animacy of OBJ relative to SUBJ
– a discourse related ‘global’ factor. Furthermore, descriptions often fail to take into
account sociolinguistic variation of the use of DOM over many countries where
Spanish is spoken in a variety of settings (monolingual, bilingual, heritage, etc.). In
the acquisition literature, DOM is often treated as a purely structural phenomenon
which is supposedly easy to learn. Chapter 7 instead places DOM high up among
the last stages of PT, and explains why that is so.
Constituent questions are dealt with in chapter 8. They are extremely interesting for PT in so far as they are sentences marked both pragmatically and grammatically. Pragmatically, they satisfy an important communicative need, as speakers use them to request new information. Hence constituent questions always
have an element ‘in focus’ (Lambrecht 1994; Mycock 2007), which is the interrogative phrase. Focality is not a prerogative of questions, but in this type of sentence the focus is obligatory, and responds to specific linguistic constraints.
Hence, constituent questions are also linguistically marked. Such constraints vary
cross-linguistically, and can be structurally complex to encode. Furthermore, in
comparison with declarative sentences, constituent questions occur less frequently in spontaneous conversation. For all these reasons, it is not surprising that they
are difficult to acquire by L2 learners. Chapter 8, then, describes how content
questions are realized syntactically in Italian by using the LFG framework, then
discusses PT’s hypotheses for their developmental hierarchy based on the
Prominence Hypothesis, and finally tests these hypotheses on empirical cross-sectional data from learners with a variety of L1 backgrounds.
Chapter 9 deals with German declarative sentences and constituent questions
in relation to its V2 rule, which is seen from the new PT perspective of the
Prominence Hypothesis. This chapter analyses learners’ development beyond
canonical word order, and compares progress in topicalised declarative sentences
and in constituent questions. Results show that V2 – that is, noncanonical word
PART III. Exploring new issues
order – emerges in questions before it does in declaratives, thus suggesting that question FOC is a more powerful trigger than TOP in learners’ progress beyond canonical order.
Chapter 10 deals with an autistic child acquiring Italian L2. Among the defining features of autism are delays and deficits in language and communication.
However, the exact nature of these problems is unclear because language outcomes
vary greatly. For example, some children never acquire speech, others acquire only
limited speech and yet others, despite early delays, acquire language within the normal range. Little is known about grammatical development in high functioning
children, yet understanding how language develops in the early stages in this population may provide valuable insight into how underlying processing difficulties
contribute to their speech delays. This chapter presents a case study designed to
assess whether a 6-year old child with high functioning autism learns an L2 by the
same developmental path as typically developing children. Results confirm that it
is the same, with some indication that progress is not slower. Rather, each stage
seems to develop at a rapid pace.
Chapter 11 addresses pedagogic and research methodological issues by connecting CALL and PT with a focus on Japanese L2. The CALL activity examined
is text-messaging exchange between learners of Japanese L2 in Australia and learners of English L2 in Japan. Participants text-chatted on various topics as an outof-class activity three times over a period of two months. Japanese L2 production
during these sessions is analysed in terms of lexicon, morphology and syntax.
Furthermore, language development is examined in order to check whether the trajectory of morphosyntactic structures defined by PT for oral production is confirmed also for the written production during text messaging. Results confirm
Pienemann’s (1998) Steadiness Hypothesis, but they also suggest that there are vast
individual differences in students’ language productions and learning outcomes as
measured by PT stages. This points to the need of not only monitoring learners
closely to promote overall linguistic development, but also using a reliable developmental measure such as PT’s schedules.
Acquiring case marking in Russian as
a second language: an exploratory study
on subject and object
Daniele Artoni and Marco Magnani
University of Verona
1. Introduction
This chapter deals with the development of case in Russian as a second language.
Case is an important morphological device for marking grammatical relations
among constituents, and as such is of great interest for testing the interface between
morphological and syntactic development proposed by Bettoni & Di Biase in the
introductory chapter of this volume (ch. 1, § 4.3). Yet so far, few studies have
specifically dealt with the L2 acquisition of case. Among the exceptions, some are
notable. Baten’s (2011, 2013) study on the acquisition of German case, and the
study on Serbian case in chapter 6 of this volume are within the PT framework.
Although they are different in design – the former being longitudinal on a variety
of Dutch L1 learners of German L2, and the latter cross-sectional on learners of
Serbian as a heritage language in Australia – they both show that learners begin to
mark case first positionally in the clause (i.e., they assign noM to preverbal SUBJ,
and ACC to postverbal oBJ), and then functionally (i.e., regardless of position in cstructure). Without the PT framework, studies on the L2 acquisition of case are
also scarse generally, and even more so with respect to Russian. one exception is
Kempe & MacWhinney (1998), who investigate the acquisition of case marking
in Russian in comparison to German. Their findings are interesting in so far as they
show that, although the Russian case-marking system is more complex than the
German one, it appears to be learned faster because its inflections are more reliable
for sentence interpretation. However, their study deals with comprehension, and
production remains unexplored.
In line with the PT framework proposed in this volume, in our chapter we
explore the development of Russian case and hypothesise a developmental schedule interfacing morphology with syntax. More specifically, we focus on how case
morphology is used to mark GFs, with particular attention to the two core GFs,
namely, SUBJ and oBJ. our hypotheses are then tested on cross-sectional data col3
Grammatical development in second languages, 177-194
Daniele Artoni and Marco Magnani
lected among 8 learners of Russian L2. Results will show that learners progress
from a first match between GFs and their default case markers in a fixed canonical
word order frame to expressing GFs also with nondefault case markers as lexically
required by V, and finally to marking full functional assignment by case independent of position. This allows learners to deploy more flexible word orders to express
their discourse-pragmatic needs unambiguously.
2. Case in Russian
From a typological point of view, Russian is a language with a low degree of configurationality and a rich case morphology that set it among the dependent-marking languages (for an overview of this language, cf. Comrie 2011). A case system is
a prominent characteristic of dependent-marking languages, traditionally defined,
in a general way, as a system marking dependent nominals to the type of relation
they bear to their heads in a phrase (Blake 1994). Case, then, is not a universal feature, as GFs can be identified by three different means: (a) case marking, which is
the main means used by dependent-marking languages such as Russian, Japanese
(cf. ch. 4, this volume), and Warlpiri (cf. (24), § 2.2, ch. 1); (b) agreement, which
is very productive in Romance languages like Italian (cf. ch. 3) and Spanish (cf. ch.
7), where SUBJ and V must agree in person, number, and sometimes also gender;
and (c) word order or position in phrase structure, used in configurational languages like english (cf. ch. 2). natural languages can then obviously exploit more
than one means to identify GFs. Russian and Latin, for instance, present both a
rich case morphology and SUBJ-V agreement. This allows speakers to resort to different word orders and organise sentences according to the pragmatic requirements
of the ToP-FoC structure of their sentences (cf. §§ 3.1, ch. 3 on Italian, and ch.
4 on Japanese).
Within the LFG framework, King (1995) discusses four types of case assignment in Russian: semantic, configurational, functional, and lexical. Semantic case
assignment, as its label suggests, occurs when a particular case is associated with a
particular semantic meaning at a-structure. Semantic cases are common across languages, but according to King (1995) the only candidate in Russian is the instrumental case for <instrument>, as shown in (1).
(1) ja
napisala pis’mo
I-noM wrote
letter-ACC pencil-InST
[I wrote a letter with a pencil]
Configurational case assignment occurs when a specific case is assigned to a n
appearing in a certain position in phrase structure. In King’s view, this occurs in
5. Acquiring case marking in Russian as a second language
Russian when genitive in n is daughter of nP, as exemplified and formalised in (2).
notice that, unlike with semantic case, genitive is assigned only by position in cstructure because the genitive sister of n can mark different semantic roles, such as
possessor, as well as agent as in (2).
(2) a. otvet
answer-noM pupil-Gen
[(the) answer by (the) pupil]
b. nP → n
((↓CASe) = Gen)
Crucially, to build up the sentence, case assignment can mark GFs. In Russian
three GFs require their own default case, namely, nominative (noM) for SUBJ,
accusative (ACC) for oBJ, and dative (DAT) for oBLGoAL, as exemplified and formalised in (3).
(3) a. mal’čik dal Inne
boy-noM gave Inna-DAT book-ACC
[the boy gave Inna a book]
b. (↑SUBJ CASe) = noM
(↑oBJ CASe) = ACC
Finally, lexical case assignment occurs when case is governed by a particular verb or
preposition, as exemplified and formalised for verbs in (4) and for prepositions in
(5) respectively. note that the rule in (4) is not contradicted by those in (3) because
specific lexical requirements can override defaultness.
(4) a. upravljaet
b. upravljat’ ‘manage’
(↑oBJ CASe) = InST
(5) a. u okon
by windows-Gen
b. u ‘at/near’
PReP <oBJ>
(↑oBJ CASe) = Gen
Daniele Artoni and Marco Magnani
In this study we will deal with GF case assignment and lexical case assignment
by V.
Russian canonical word order is SVo,1 so SUBJ marked as noM is the default
ToP in preverbal position, and oBJ marked as ACC is the default FoC in postverbal position. However, because of Russian nonconfigurationality, SVo occurs only
47% of the time in native oral production (Timberlake 2004). This means that
case-marked GFs are not positionally predictable. For discourse and pragmatic reasons, constituents can occur in different positions, allowing for all the six possible
combinations of the three core elements, as shown in (6) – even though it is important to note that word orders in (6d-f) rely heavily on prosodic features and, being
highly marked, are rarely used by Russian L1 speakers (Kallestinova 2007). In the
examples below these six combinations are identified thanks to the feminine nominative –a marking SUBJ, and the feminine accusative –u marking oBJ.
(6) a. Marija
Marija-noM.SUBJ eats-V porridge-ACC.oBJ
b. Marija
Marija-noM.SUBJ porridge ACC.oBJ eats
c. kašu
porridge-ACC.oBJ eats-V Marija-noM.SUBJ]
d. kašu
porridge-ACC.oBJ Marija-noM.SUBJ eats-V
e. est
Marija-noM.SUBJ porridge-ACC.oBJ
f. est
porridge-ACC.oBJ Marija-noM.SUBJ
What unites the sentences in (6) is that they all share the same f-structure. on the
other hand, the most relevant difference in c-structure is that oBJ occurs within
VP when it is postverbal, whereas it is set outside VP when it is displaced to ToP
position (King 1995: 206), as shown in (7). This implies that when oBJ is preverbal, the value of the oBJ CASe feature of V must be checked interphrasally with
the case value of nPoBJ.
1 For an alternative view on Russian word order, cf. King (1995), who suggests VSo as
the unmarked, pragmatically neuter word order.
5. Acquiring case marking in Russian as a second language
(7) Correspondences between a-structure, f-structure and c-structure for the sentence kašu est
Marija (‘the porridge Marija eats (it)’)
Readers will appreciate that for learners of Russian L2 case is a complex feature to
acquire. Yet the complexity of its functional deployment is well matched by the
complexity of its formal characteristics. There are six cases in Russian: nominative,
genitive, dative, accusative, instrumental, prepositional (locative and ablative),
which are fusionally enmeshed with other nominal features such as number (singular or plural), gender (masculine, feminine or neuter), animacy, and class. In (8) and
(9) we show the full case-marking paradigm for ns and pronouns respectively. As
we can see, the many-to-many relations between cases and markers are noteworthy.
In particular, with regard to SUBJ and oBJ, the two cases of interest in this chapter, neuter and masculine inanimate ns share the same marking strategies for noM
and ACC. For this reason, we will consider only feminine and masculine animate ns
for matching SUBJ to noM, and oBJ to ACC. on the other hand, with regard to
pronouns, no ambiguity between markers is found in the noM and ACC declension.
Daniele Artoni and Marco Magnani
(8) Russian case-marking paradigm: nouns (after Kempe & MacWhinney 1998)
(9) Russian case-marking paradigm: pronouns (after Kempe & MacWhinney 1998)
3. Developmental Hypotheses
In (10) we show our developmental hypotheses for Russian case relative to the GFs
SUBJ and oBJ. As we have noted in the introduction to this chapter, case is particularly interesting for PT because the two schedules for morphological and syntactic development interface in a crucial way. So, contrary to other chapters in this
volume (e.g., chh. 1-4), we include both morphological and syntactic development
5. Acquiring case marking in Russian as a second language
in a single table. We should also mention that our Prominence Hypothesis deals
with declarative sentences, leaving interrogatives to future work.
(10) Developmental stages hypothesised for Russian case
After the single-word and formulaic stage, as soon as the category procedure
becomes operative for morphology, learners are able to distinguish categorially
between ns and Vs. Formal marking of ns, then, begins to emerge. With regard
to the case feature specifically, at this stage learners begin to distinguish between
the noM form and a general non-noM form bearing any inflectional ending
other than noM (e.g., kaša ‘porridge-noM’ vs. kašu/kaše/kaši ‘porridge-nonnoM’). This initial opposition between noM and non-noM (cf. Jakobson 1971),
although minimal and restricted within the word, is sufficient to distinguish
formally between SUBJ and non-SUBJ GFs in syntax. However, because at this
stage GFs are identified positionally rather than functionally (cf. ch. 1, § 4.2.1),
learners will deploy this minimal form variation onto a minimally specified
SVo sequence – that is, they will produce canonical strings with a preverbal
nnoM and a postverbal nnon-noM, as shown in (11).
Daniele Artoni and Marco Magnani
(11) ochotniki
ubit volke
hunters-noM kills wolf-non-noM
[the hunters kill the wolf ]
At the next stage, with regard to morphological development, learners are able to
exchange information within a phrase. In Russian this phrasal procedure stage
involves a variety of structures, such as agreement between noun and adjective
within nP and government by the preposition within PP. The relevant structure
for our study is V-oBJ unification when learners are able to exchange information
within the VP between the value of the case feature of nPoBJ and the value of the
feature oBJ CASe required by V. If oBJ is marked by its default ACC case in its
default postverbal position, the evidence of intraphrasal exchange of information
remains equivocal. We prefer to consider unequivocal evidence of progress to this
stage oBJs marked with cases other than ACC (e.g., InST)2, as exemplified above in
(4), § 2, and further formalised in a full sentence in (12). needless to say, this does
not imply that all nondefault cases required by V will be learned soon. Being
required lexically, they will have to be learned individually V by V.
(12) Annotated c-structure for the sentence Oleg upravljaet biznesom (‘Oleg manages business’)
2 Also oBJInST could randomly appear as oBJnon-noM at the category procedure
stage. What provides unequivocal evidence for online information exchange within the
VP is a variety of cases with different lexical Vs.
5. Acquiring case marking in Russian as a second language
With regard to syntax, we hypothesise that at this stage learners will be able to
place an element other than SUBJ (typically ADJ) in the prominent first position
as in (13). This addition will bring about a differentiation between SUBJnoM and
the topical first constituent in the clause.
(13) na kartinke devuška
in picture
girl-noM eats
[in the picture the girl eats an apple]
At the last stage of their morpho-syntactic development, learners will be able to
assign GFs irrespectively of position. As we have seen, this requires two morphological resources in Russian: a head-marking strategy, namely SUBJ-V agreement
for the identification of SUBJ; and a dependent-marking strategy, namely casemarking for identifying the three main argument functions – that is, SUBJ, oBJ
and oBLGoAL. With regard to the former strategy, with the activation of the Sprocedure, learners can now produce the agreement between the SUBJ features
(number and gender) and the predicate, as shown in (14). Although SUBJ-V
agreement is not the focus of this study, in (14) we show the unification of SUBJ
with a nonverbal predicate rather than with a lexical V because in Russian, as well
as in other null-SUBJ languages,3 both SUBJ and V need to retrieve their features
from the conceptual structure independently (cf. ch. 3, § 2.1 on Italian L2 for a
discussion on this point).
(14) pogoda
weather-SG.FeM was-SG.FeM
[the weather was good]
Thanks again to the activation of the S-procedure, information exchange
between V and its complements can now occur across phrases. Hence learners
will be able to case-mark the topicalised oBJ with ACC as a result of the
exchange of information between VP and the external nP, and produce sentences like (15).
3 Russian, however, is not universally recognised as a null-SUBJ language. For instance,
Avrutin & Rohrbacher (1997) argue that Russian is not a null-SUBJ language, and
ascribe the rare instances of null SUBJ to contextually licensed ellipsis. on the other
hand, Müller (1988) and Pearlmutter & Moore (2002) support the null-SUBJ thesis,
and justify the contextual limits on the basis of discourse conditions that “make it much
less common than pro-drop in Italian or Spanish”.
Daniele Artoni and Marco Magnani
(15) knigu
čitaet mama
book-ACC reads
[the book mum reads (it)]
At this last stage, then, the interplay between morphology and syntax in the development of case is clear. on the one hand, morphology feeds syntax in the sense
that only when the S-procedure is firmly in place can learners case-mark constituents unambiguously regardless of word order constraints. on the other hand,
along the path for morphological development, when learners are able to free up
the rigidity of the canonical word order frame – crucially, by choosing to topicalise
the core function oBJ – they provide convincing evidence that case is assigned via
interphrasal information exchange.
4. The study
The evidence we bring to test our PT-based hypotheses discussed in § 3 comes
from a cross-sectional study of 8 learners of L2 Russian (2 males and 6 females), all
adults at different proficiency levels. As the table in (16) shows, they are all instructed learners, who have studied Russian at university for one to five years. Two of
them (AB and LI) have also acquired and practised the language in a Russian
speaking environment. Their L2 proficiency levels vary from A2 to C1 as measured
on the general CeFR (2001) scales. Their L1 backgrounds are also varied, including Italian for the majority of them, Serbian (Jo) and Georgian (LI). It is however important to note that, although some of the variables in (16) – particularly the
L1 background – can contribute in speeding up progress along the developmental
path, they will not affect the sequence in which stages are reached (cf. the
Developmentally Moderated Transfer Hypothesis in Pienemann, Di Biase,
Kawaguchi & Håkansson 2005).
(16) The learners
5. Acquiring case marking in Russian as a second language
We used five tasks to elicit the structures listed in (10). The first task, Znakomstvo
(‘Introduction’), is a conversation aimed to get to know and relax the learners,
and targeting general structures. The next two tasks, Krasnaja Šapočka (‘Red
Riding Hood’) and Najdite različija (‘Spot the differences’), are respectively a
story telling and spot-the-difference task targeting canonical word order and
ADJ topicalisation. In the fourth task, Eščë Krasnaja Šapočka (‘Red Riding Hood
Again’), two cards showing a person and an object were introduced together with
a V in its infinitive form, and the learner was asked to create a sentence with the
three given items. This task targets the use of oBJ with cases other than ACC
which are lexically required by V, such as upravljat’ (‘manage’) and zanimat’sja
(‘do/practice’), both requiring oBJ to be marked by InST. The last task, Večerinka
(‘The party’), targets oBJ topicalisation, and is an adaptation to Russian L2 of
the task used first by Di Biase & Kawaguchi (2002) and then by Bettoni & Di
Biase (2011) for eliciting Italian L2 structures (cf. § 2.2, ch. 3, this volume). Di
Biase’s ‘animal dinner’ task becomes a ‘party’ to which various participants must
contribute something. When two picture cards appear on the computer screen,
learners are encouraged to tell the researcher who is bringing what by starting
with the card on the left. Since this card shows sometimes the person/agent,
other times the object/theme, SVo sentences should alternate with oVS ones.
The communicative context of all the five tasks and the use of several distractors
scattered among the targeted structures aim to enhance learners’ online production, and exclude a focus on declarative knowledge.
our analysis comprises a total of 476 case-marked structures included in
333 main declarative sentences with lexical Vs – thus excluding copular and
presentative sentences, which are ‘nonverbal predicates’ (Kroeger 2005: ch. 10).
In (17) we illustrate the distributional analysis of the case markers on SUBJ and
oBJ among learners, structures and stages. The table is organised as set out in
this volume (cf. (48a-c), § 5, ch. 1), but four further points must be clarified.
First, most structures in the first column are repeated twice because when both
GFs are case-marked in a sentence, the figures for case refer only to the GF in
bold, and the brackets around the other GF mean either that its case is not considered in that row or that the GF is absent from the string (there being a null
SUBJ or an intransitive V). This then explains why in the table the total number of case-marked GFs (476) is higher than the number of analysed clauses
(333). Secondly, given the relatively high degree of case syncretism in the
Russian case system, in our analysis we consider only unambiguously casemarked ns. So, with regard to the marking of noM and ACC, as we have anticipated in § 2, for evidence of acquisition of ACC we only look at case marking
on feminine nouns (e.g., ručka ‘pen-noM’ vs. ručku ‘pen-ACC’) and masculine
animate nouns (e.g., student ‘student-noM’ vs. studenta ‘student-ACC’), and thus
exclude masculine inanimate and neuter nouns, where noM and ACC are equal-
Daniele Artoni and Marco Magnani
ly marked. With regard to the marking of InST for lexical case assignment, there
being no ambiguity with other case marks, we consider all nouns, provided that
the structure is produced online in a nonformulaic way. Thirdly, as evidence of
acquisition of V-oBJ unification through lexical case assignment, we do not
consider the occurrences of oBJs marked by the default ACC instead of the nondefault InST, because the use of the default ACC here may be due to the lack of
annotation of the oBJ CASe feature in the verbal lexicon. Thus, among the figures for oBJInST, the minus always indicates that oBJ is marked by noM – an
unequivocal example of lack of information exchange with V. Fourthly, following Pienemann (1998: 159), in this analysis we factor in the case feature only,
and consequently factor out number and gender. This means that, for example,
in the case of oBJ marked by InST, the use of the InST masculine ending –om
in *babuškom (‘grandmother-InST’) instead of the correct feminine one –oj is
considered as safe evidence of acquisition of the structure.
(17) Cross-sectional study of the development of Russian case
All our learners can produce sentences of preverbal ns marked as noM and postverbal ns marked as non-noM, as in (18). Their numbers are convincing, and we can
thus safely say that they have all reached the category procedure stage for morphology and the canonical word order stage for syntax.
(18) AL
videla volke
saw wolf-non-noM
However, a finer analysis reveals interesting differences in how oBJ is marked. on
the one hand, the overall average for the use of the incorrect noM marker is
31.48%. All learners but one (Jo) use it, even the more proficient ones, as shown
by the minus figures for AB (9 out of 17) and exemplified in (19). on the other
5. Acquiring case marking in Russian as a second language
hand, when oBJ is marked by a case other than noM, the overall average for the
correct ACC –u/–ju marker is 93.24%, as exemplified in (20). All learners but one
(AL) use ACC all the time, whereas AL, the least proficient, alternates between the
ACC marker and the PReP case –e marker, as in (18) above.
(19) AB ochotniki
hunters-noM found-PL wolf-noM
[the hunters found the wolf ]
(20) LI
[they saw the wolf ]
Four of our learners, CA, LI, AB and MT, show evidence of having reached the
next stage, which for morphology involves VP unification between V and oBJ. As
we have mentioned in § 3, uncontroversial evidence of this unification can be captured when V requires oBJ marked by a case other than ACC, such as InST in (21).
However, only one such instance by eL does not satisfy our emergence criterion
for morphology (cf. § 5, ch. 1, this volume).
(21) MT ona
she-noM does
[she practices music]
These four learners also use structures that topicalise ADJ, as in (22). notice that
evidence of progress at this stage is determined by the syntactic position of
SUBJnoM, which is no longer the default (initial) one, rather than by its case
marker, which is the default noM.
(22) LI
potom ona
posmotrela babušku
she-noM saw
[then she saw her grandmother]
Two learners, AB and MT, have also reached the last stage of development in so far
as they are able to use case to mark GFs irrespectively of their position in the clause.
Hence, they can produce sentences like (23), where oBJ marked as ACC is in preverbal position.
(23) MT vilku
prinesla balerina
brought dancer-noM
[the fork, the dancer brought it]
Daniele Artoni and Marco Magnani
As we have remarked earlier, oBJ topicalisation in these constructions was
deliberately prompted by our fifth task. What happens, then, when learners
who have not yet acquired the morphological resources of the last stage are
asked to perform this task? The most common solution is to use noM on both
ns in the oVS structure, as shown in (24). Although here postverbal noM is
correct, we cannot take it as evidence of progress to this noncanonical word
order stage because, as we have remarked above, noM is the default case – hence
the brackets in the table.
(24) AL vilka
fork-noM brought dancer-noM
[? the fork brought the dancer/the dancer brought the fork]
one could argue that lack of case marking in (24) may be motivated by the fact
that AL’s L1 is Italian, a language which never marks ns by case. However, we
can safely say that this is not an issue of L1 transfer, as ACC is never marked on
topicalised oBJs even by Jo and LI, whose L1s are Serbian and Georgian respectively – two heavily morphologised languages with a rich case system. In this
regard, we note that in two out of three of these topicalised declaratives, the
Serbian learner Jo marks preverbal oBJ by the default noM, and postverbal
SUBJ by ACC, as shown in (23).
(23) Jo vilka
fork-noM brought dancer-ACC
[the fork brought the dancer]
This is a common solution for learners who are at the earlier developmental
stage. MA, for example, marks 3 out of 4 postverbal SUBJs as ACC. So it would
seem that for 6 of our learners syntactic position still leads the way for marking
case, even in the context of an L1 background which is typologically very similar to Russian.
5. Discussion and conclusion
In this chapter we hypothesised a developmental hierarchy for case in Russian L2
considering the interface between PT’s schedules for morphology and syntax. We
have seen how case presents an intriguing challenge for testing such an interface,
and how Russian offers an ideal testing ground for several reasons – chiefly among
them its rich morphological system and its high nonconfigurationality, allowing for
permutations of the core elements in the clause.
5. Acquiring case marking in Russian as a second language
Results of our cross-sectional study confirm our PT-based hypotheses, in so
far as on both the morphological and the syntactic schedules there is no learner
who produces structures at a higher stage without producing also at least some at
the previous stage. Results also support the interface between morphological and
syntactic development suggested by Bettoni & Di Biase in chapter 1, § 4.3. In
particular, the interface is clear at the lowest and the highest grammatical stages.
That is, first, when the category procedure becomes operative in morphology,
learners begin to distinguish between noM and a general non-noM form, but can
only deploy this minimal form variation within the fixed frame of the SVo
canonical word order; and then last, only when learners can activate the morphological resources of the S-procedure can they free up the rigidity of canonical word
order constraints, and assign case to GFs irrespectively of position. on the other
hand, in the intermediate stages, that is, when the phrasal procedure stage is supposed to interface with the XPDF canonical word order stage, the interface
between the learners’ ability to produce V-oBJ unification in morphology and to
topicalise an ADJ in first position seems less clear, as also hypothesised by Bettoni
and Di Biase in chapter 1, § 4.3.
A closer look at learners’ production of pragmatically-driven oVS sentences can further contribute in explaining the path from case assignment based
on position towards case assignment irrespectively of position. Three main outcomes are used by learners in our data: (i) both nPs are marked by the default
noM; (ii) both nPs are marked positionally, that is, ToPoBJ by noM and
postverbal SUBJ by ACC; and (iii) like in target Russian, both nPs are casemarked functionally regardless of position, that is, ToPoBJ by ACC and postverbal SUBJ by noM. Such outcomes are not randomly distributed across learners,
and may be interpreted as subsequent stages along the developmental path.
More specifically, overextention of noM on both nPs tends to occur both at the
beginning of the interlanguage path (AL) as in (24), and at intermediate stages
(eL, CA and LI) as in (25).
(24) AL gitara
brought-SG.MASC dancer-noM.SG.FeM
[? the guitar brought the dancer / the dancer brought the guitar]
(25) LI
pear-noM.SG.FeM brought-SG.FeM teacher-noM.SG.FeM
[? the pear brought the teacher / the teacher brought the pear]
In (25), however, there is more accuracy in V morphology than in (24), in so
far as V is correctly inflected as SG and FeM – even though it is impossible to tell
whether it agrees with preverbal n or with postverbal n. Moreover, although
Daniele Artoni and Marco Magnani
both beginner and intermediate learners overextend noM in some sentences
with canonical word order as well, we have seen that intermediate learners are
more accurate in using ACC for oBJ than AL, who uses a variety of non-noM
markers for oBJ. Thus, whereas AL’s overextension of noM can be seen as a general default solution, for intermediate learners it seems to be an indicator for
their inability to use noM and ACC in ways other than positionally. A longitudinal study would confirm this whole sequence for case, as well as for SUBJ-V
agreement in all learners.
our study also shows that typological similarities between the L1 and the
target L2 do not allow learners to skip stages along the developmental sequence.
In fact, our Serbian learner (Jo) also marks constituents by case according to
their position and is thus at the earlier developmental stage. on the other hand,
our analysis suggests that such similarities can play an important role in increasing accuracy within a stage. In fact, Jo is the only one who marks postverbal
oBJs always accurately with ACC rather than with a non-noM case.
Finally, using King’s (1995) terms for case assignment, we have resorted to
evidence of lexical case assignment for proving V-oBJ unification at the phrasal
procedure stage, and to grammatical case assignment for proving ToPoBJ-V
unification at the S-procedure stage. This of course does not entail that lexical
case markers are acquired before structural ones generally. This study makes no
predictions as to the point of emergence of the different lexical case markers –
a thorny issue which needs to be worked out at the interface with the Lexical
Mapping Hypothesis. It only uses them when they are found in the data as evidence for the activation of the interphrasal precedure in order to rule out the
possibility that default ACC is assigned simply because of the n position.
In conclusion, this study has shown that marking SUBJ and oBJ by case
in Russian is no mean feat for L2 learners. Although minimal variation between
noM and non-noM is sufficient for distinguishing between SUBJ and oBJ in
the fixed canonical word order frame, trouble begins when V lexically requires
a nondefault match between GF and case, and when discourse-pragmatic
requirements place GFs other than SUBJ in the prominent first position.
Indeed, only intermediate learners will start marking oBJ by the nondefault
InST case, and only the most proficient ones can mark the GFs morphologically even when sentences display noncanonical word order. PT can explain why
this is so by tracing the learner’s developmental path from case marking based
on position towards case assignment independently of position. on the other
hand, our study only scratches the surface of the acquisition of case in a highly
nonconfigurational language such as Russian. Further investigation is needed in
several directions, as well as more substantial evidence on more diverse structures in a wider corpus. Future work, for example, should include the third case-
5. Acquiring case marking in Russian as a second language
marked argument GF, namely the dative oBLGoAL; interrogative sentences as
well as the interface between PT’s Prominence Hypothesis and Lexical Mapping
Hypothesis with regard to syntax; and King’s (1995) semantic and configurational case assignments with regard to morphology. These are unexplored areas
in Russian L2.
The authors wish to thank Camilla Bettoni, supervisor of their doctoral work, and
Bruno Di Biase, her co-editor of this volume, for their help in discussing and organising the issues dealt with in this chapter.
The development of case: a study of Serbian
in contact with Australian English
Bruno Di Biase*, Camilla Bettoni** and Lucija Medojević*
*University of Western Sydney, **University of Verona
1. Introduction
Within the PT framework, this chapter represents an exploration in three new
directions. First, we look at the development of a case system in learners. Despite
PT’s fairly extensive empirical application to a variety of languages and acquisitional circumstances and populations (cf. § 1, ch. 1, this volume), surprisingly
little attention has been given to the development of case systems, except for
Baten’s work on German L2 (2011, 2013) and Artoni & Magnani’s on Russian L2
(2013) and in chapter 5, this volume. The system we will be looking at is that of
Serbian, like Russian a heavily morphologised case-marking language (for an
overview of Serbian from a typological point of view, cf. Corbett & Browne
(2011). After a brief account of the case system in Serbian, we will present our PTbased hypotheses for its development.
Secondly, we look at how a heritage language develops in contact with a
dominant one. Many children of immigrants in Australia grow up with language exposure that is ‘situation-bound’ (Vihman & McLaughlin 1982; Clyne
2003; Qi, Di Biase & Campbell 2006), where their two languages, a heritage
language and English, develop separately (Meisel 1989; De Houwer 1990,
2005; Itani-Adams 2007). As far as Serbian is concerned, the Serbian input children receive is limited to home-based language use in communication with family members. Such a limitation in language input, coupled with the lack of
opportunities for output (cf. Swain’s 1995 output hypothesis) in a predominantly English-speaking environment, may affect the development and attainment (Doughty & Long 2003) of Serbian knowledge and skills. It is likely that
the case-marking system will be affected because Serbian, which is a heavily
morphologised case-marking language, is in contact with English, which is a
highly configurational language with a much-reduced morphological expression
of case. In testing our hypotheses on production data from three SerbianAustralian teenagers, all early bilinguals living in Sydney, we then expect to find
Grammatical development in second languages, 195-212
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
that their Serbian case system is not as fully developed as that of a comparable
native speaker.
Thirdly, we look at how the case system can interface with pragmatic choices,
in a similar way to Artoni & Magnani (cf. ch. 5, this volume). Because Serbian case
markers on nominals identify functional roles (Hammond 2005: 105), they help
give structural realisation to pragmatic factors such as topicality and prominence in
the sentence (Browne 1993). Whilst the default Serbian word order is SVo, constituents are largely ordered by topic-comment structure. So, a limited availability
of the case-marking system in their weaker language is important when discussing
bilinguals’ choices for structures at the interface between syntax and pragmatics.
Thus we also use PT to assess how a more developed case-system can allow for a
wider range of pragmatic-discourse driven syntactic choices.
2. The Serbian case system
A case system is a prominent characteristic of dependent marking languages (cf.
(1), part II, this volume), and – as we have seen in chapter 5 about Russian – is traditionally defined, in a general way, as a system marking dependent nominals to
the type of relation they bear to their heads in a phrase (Blake 1994). ns can therefore depend on heads belonging to various lexical categories: verb, noun, adjective
and preposition. Each of these lexical categories is associated with its typical cases
for the ns within their VP, nP, AP, and PP respectively. Among these lexical categories, V is crucial for the construction of the sentence because the cases it requires
for its thematic roles identify their GFs. A corollary of this definition of case is that,
when n is independent of any other words and fulfils the SUBJ function, it takes
the basic ‘unmarked’ form of the nominative case.
There are seven cases in Serbian: nominative, accusative, locative, genitive,
instrumental, dative and vocative. In this chapter, vocative will no longer be mentioned because it is independent of any other element in the sentence. Each lexical
category governs its typical case, such as ACC for V, and GEn for n, as shown respectively in (1a-b).
(1) a. mačka tera
chase mouse-ACC
[the cat is chasing a mouse]
b. interpretacija romana
je veoma interesantna
interpretation novel-GEn is very
[interpretation of the novel is very interesting]
6. The development of case: a study of Serbian in contact with Australian English
However, Serbian lexical categories may govern other cases, as shown in (2a-b),
where the oBJ required by V is expressed by GEn and InST respectively rather than
by ACC. needless to say, these less systematic associations (or more marked cases)
are harder to acquire.
(2) a. baka
je nakopala krompira
grandma be dug up
[grandma dug up the potatoes]
b. kralj vlada zemljom
king rule country-InST
[the king rules the country]
Like verbs, also nouns, adjectives and prepositions can assign a variety of cases. For
example, besides GEn, shown in (1b), n can assign DAT, as in (3).
(3) spomenik Puškinu
je jako visok
monument Pushkin-DAT is very high
[the monument to Pushkin is very high]
As case assigners, prepositions are especially taxing for the learner. Few of
them, if any, select only one case; and few cases, if any, depend on only one preposition. So, on the one hand, displaying high functional overlap with their
homonymy, polysemy, synonymy (Savić & Anđelković 2007), most prepositions
select their case according to their different meanings. For example, the preposition
u governs LoC when the meaning of PP is locational ‘in’, and ACC when it is directional ‘to’ as in (4a-b) respectively.
(4) a. ja sam u
I am in
[I am in the room]
b. idem
am going to
[I am going into the room]
on the other hand, ACC for example, can be governed by u (directional ‘to’) as we
have just seen in (4b), by u (temporal ‘in’) as in (5a), or za (‘for’) as in (5b).
(5) a. u moje slobodno vreme pravim
in my spare time-ACC draw
[in my spare time I draw cards]
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
b. prodajem kolače za svadbe
cakes for weddings-ACC and for
[I sell cakes for weddings and for christenings]
Further complexities can arise from the semantics of certain nPs themselves, rather
than from the element from which they depend. For example, Serbian nPs containing the numerals two, both, three or four obligatorily require an invariable paucal form (PAUC; Franks 1994: 606), and nPs containing a quantifier obligatorily
require GEn. Thus, constituents functioning as SUBJ or oBJ may occur in this
PAUC form as in (6a-b), or in GEn as in (6c-d). Likewise, nPs in PPs containing
a quantifier occur in GEn regardless of the preposition, as in (7a-b).
(6) a. troje dece
three children-PAUC.SUBJ
[three children are reading in the park]
u parku
in park
b. imam
troje dece
have-1.SG three children-PAUC.oBJ
[I have three children]
c. mnogo pasa
u parku
many dogs-GEn.SUBJ run-3.SG in park
[many dogs are running in the park]
d. imam
nekoliko hobija
have-1.SG several hobbies-GEn.oBJ
[I have several hobbies]
(7) a. Sergej popravlja sto sa
novim čekićem
Sergej fix table
with new hammer-InST
[Sergej is fixing the table with a new hammer]
b. Sergej radi sa
nekoliko čekića
Sergej work with several hammers-GEn
[Sergej is working with several hammers]
Yet another set of difficulties for the learner is created by the fact that Serbian morphology, like Russian, is highly fusional. on n, for example, a single inflectional
morpheme carries information about features such as gender (masculine, feminine
or neuter), number (singular or plural), and case (nominative, accusative, genitive,
dative, instrumental, or locative). Compounded with all this, there are three
phonologically-based n classes: the first class, arising from Proto-Slavonic o-stems,
includes most masculine and all neuter ns ending in –o, –a or a consonant in noM
6. The development of case: a study of Serbian in contact with Australian English
SG and –a in GEn SG; the second class, continuing proto-Slavonic a-stems, includes
most feminine ns and some masculine ns ending in –a in the noM SG and –e in
GEn SG; the third class, from Proto-Slavonic i-stems, includes all feminine ns apart
from a-stems, ending in a consonant in noM SG and –i in GEn SG (Browne 1993:
319-322). no wonder, then, that the cognitive load required in sorting out formto-function mapping (Pienemann 1998: 155) is heavy. Consider, for example, the
case declensions shown in (11). With the –o stem n grad (‘city’), noM and ACC suffixes coincide in the singular, and so do DAT and LoC with –u. Also, both with masculine and feminine ns, DAT, LoC and InSTR are practically identical in the plural:
what distinguishes them in spoken production, however, is prosody (Browne 1993:
(8) Declension by case of the masculine noun grad (‘city’) and the feminine noun žena
(‘woman, wife’)
A final difficulty for the learner should be mentioned. In Serbian, within the nP,
nominal modifiers that precede the head n must agree with the head in gender,
number and case (Hammond 2005). In other words, Serbian case must be computed formally not only on the head n or pronoun but also on adjectives and some
quantifiers (Hammond 2005), as illustrated in (9). This requirement allows further
flexibility in Serbian word order (not just the freeing up of SVo), in so far as elements within an nP can be discontinuous, as in (10) which may be used to express
discourse-pragmatic requirements of the speaker.
(9) peglam ovu
iron-1.SG this-ACC blue-ACC dress-ACC
[I iron this blue dress]
(10) ovu
this-ACC be-1.SG.AUx blue-ACC
[I have ironed this blue dress]
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
In our analysis here, we will deal mainly with case construction by V. Because of
the pivotal role V plays in building up the sentence (cf. § 2.1, ch. 1, this volume),
we are crucially interested in how case can mark the GFs of their arguments (i.e.,
SUBJ, oBJ, and oBLθ) and less so in the internal structure of nPs, which would
require a larger study and more space than is available here. Among GFs, however,
we ignore CoMP because, although argumental, it is unmarked by case as such
(although it may of course contain case in its internal structure). on the other
hand, because oBLθ are most often expressed by PPs, we take into consideration
also prepositions as case assigners. In other words, here we focus on how learners
first learn to associate a set of lemma specifications with a set of GFs and build up
their f-structure by means of case markers, and then exploit their functionally casemarked sentence constituents for discourse and pragmatic reasons. Let us then look
at cases as markers of GFs.
According to nordlinger’s (1998) quadripartite typological scheme reported in (1) in part II, this volume, Serbian is an example of a nonconfigurational
dependent-marking language. This means that, on the one hand, when linking
the lexicon to c-structure, along the configurational continuum, Serbian fstructure information is expressed by morphology (rather than by position, as
in English). on the other hand, along the dependency continuum, Serbian GFs
are marked inflectionally on the depending element (rather than on the head,
as in Italian, cf. ch. 3, this volume). Marking a dependent element inflectionally means using case. As a matter of fact, as remarked in chapter 5 about Russian,
case is one of three devices by which, typologically, languages can identify GFs,
especially the core ones SUBJ and oBJ – the other two devices being word
order as in English, and agreement as in Italian (Kroeger 2005: 102-ff ). In brief,
Serbian relies on obligatory n morphology for identifying GFs, regardless of
word order.
Case as function assigner (together with SUBJ-V agreement, which in
Serbian is a further device identifying core grammatical relations) allows for
great flexibility in the word order of the Serbian sentence. So, besides canonical
SVo order, all the other five permutations of the three core elements in a sentence are grammatically acceptable in Serbian: SoV, VSo, VoS, oSV, oVS.
These six orders do not exhaust all possible sequences because all of them are
possible with null SUBJ and null oBJ. And speakers exploit them all to organise sentences according to the pragmatic requirements of the topic-focus structure. once constituents are marked functionally by case, they can be positioned
varyingly in the sentence. For example, the same propositional content,
expressed with SVo in a pragmatically unmarked way in (11a), can be
expressed with different word order if the speaker wishes to topicalise oBJ in
6. The development of case: a study of Serbian in contact with Australian English
(11) a. mačka
cat-noM.SUBJ chase-3.SG mouse-ACC.oBJ
[the cat is chasing a mouse]
b. miša
mouse-ACC.oBJ chase-3.SG cat-noM.SUBJ
[the cat is chasing a mouse / it is a mouse that the cat is chasing]
In sum, for learners of Serbian, case is a complex feature to acquire for a variety of
reasons: morphologically, there are several cases, fusionally enmeshed with other
nominal features such as number, gender and class. Morphosyntactically, case must
be computed on most nominal elements within the nP. Syntactically, cases identify GFs in the sentence. Most of the time, relations between case and function are
default and predictable. However, less predictably, the same case can construct
alternative GFs, and the same GF can be constructed by different cases, although
lexically with different predicates. on the basis of these complexities, we now suggest some hypotheses for the development of the case system in learners of Serbian
as a heritage language.
3. The developmental hypotheses
Morphologically, we hypothesize that, among the Serbian cases, the first to be used
will be noM for three main reasons: noM is the citation form; it often coincides
with the ACC form in the singular; and it is the case learners pervasively find in the
prominent first position in the sentence. once they notice variation in form, early
contrasts may be set up as noM~nonnoM with the nonnoM form(s) overextended. Then the sequencing in the spread of case forms from the emergence of a first
contrast to the whole paradigm is an empirical matter, which may be subject to
variation, both contextual and individual. And it goes without saying that we are
talking here about the emergence of case-markers, without any consideration for
their formal accuracy in terms of gender, number, or class.
Morphosyntactically, we predict that case will be marked according to PT’s
well-tested hierarchy based on the activation of the processing procedures which
allow for the exchange of information about case first within the phrase, and then
within the sentence. As already mentioned, the focus in our analysis here is at sentence level, i.e., on cases marking GFs, so we will not take into account case agreement phenomena. on the other hand, the case marker of the n in a PP bearing
the oBL function will be considered, even if this is specified in the lexical entry of
the preposition, and hence phrasal in terms of morphological development and
independent of functional assignment in terms of syntactic development.
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
Syntactically, the sequence in (12) shows our developmental hypothesis for
Serbian declaratives based on PT’s Prominence Hypothesis (cf. § 4.2.1, ch. 1, this
volume, for the universal sequence, and chh. 2, 3 and 4 for English, Italian, and
Japanese respectively). At the canonical word order stage, the first match will occur
between form and position; thus in an SVo language like Serbian, noM is associated with preverbal position, and ACC with postverbal position. As a matter of fact,
to our English Serbian bilinguals, case marking will actually seem redundant (as it
may have happened historically for English). only later, when functional assignment is in place, will learners be able to match noM with SUBJ, ACC with oBJ,
and DAT with oBL independently from the positionally determined SVo blocked
sequence. So, at the noncanonical word order stage, they will be able to depart
from the rigidity of the fixed SVo sequence in order to express their own discourse
and pragmatic choices.
(12) Developmental stages hypohesised for Serbian L2 syntax based on the Prominence
Hypothesis: declaratives
The reader should bear in mind that, like the previous study on Russian, this is an
exploratory study of case – not only for Serbian but also within PT’s framework.
Thus not all hypotheses are testable on our current cross-sectional data. Assuming
that teenage Serbian-English bilinguals in Australia have positive attitudes towards
preserving their mother tongue and use it in the home environment to communi-
6. The development of case: a study of Serbian in contact with Australian English
cate with family and friends, it is expected that they would use native adult-like
skills in sentence processing, should they possess them. Using as a benchmark the
presumably full Serbian case system of our adult bilingual informant with schooling in Serbia, we therefore focus on the range of structures and case markers displayed by our three bilingual teenagers, and the way they are deployed to allow for
speaker perspective and expressiveness beyond canonical word order rigidity.
4. Method
As already mentioned in § 1, the informants in this study are three teenage SerbianAustralian bilinguals. All three acquired both languages as first languages, with
Serbian as the home and community language, and English as the dominant one.
Their families originate from Vojvodina, a northern region of Serbia, so all of them
speak the same Ekavic dialect. A Serbian native speaker who migrated to Australia
when she was 20, and thus learned English as an adult, also participated in this
study to ensure that the range of targeted structures is actually produced by native
speakers in similar situations of elicitation. She too originates from the same
Serbian region. Further details about the four informants are shown in (13).
(13) The informants
Data elicitation took place in dyadic conversations between the informant and the
Serbian researcher. Informants were asked to perform three communicative tasks.
During the short conversation, they responded to questions about their personal
experience for the purpose of gathering information on language maintenance
strategies. During the spot-the-difference task, they could report on at least seven
differences between the two pictures shown. In the story-telling task, they narrated the children’s story of ‘Goldilocks and the three bears’ following a sequence of
Among the whole corpus, the data set analysed for this study consists of 231
Serbian sentences produced by our four bilingual speakers. In accordance with the
current state of PT in general, and its Prominence Hypothesis in particular, these
are all main declarative sentences with finite lexical V. This means that we do not
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
consider constructions involving subordination, as well as reflexives, passives, and
causatives. Also copulas and presentatives are excluded because they are ‘non verbal
predicates’ (Kroeger 2005: ch. 10). of these 231 sentences, 40 are irrelevant to our
aims here because they are unmarked for case, that is, they bear only GFs which
are CoMP, oBLLoC or ADJ expressed with adverbs as in (14), or have all participants express by a codeswitch from English as in (15).
(14) a. ned sad
da spava
now-ADJ be-3.SG.AUx want-3.SG to sleep-CoMP
[now she wanted to sleep]
b. ned ušli
entered be-3.PL.AUx inside-oBLLoC
[they went inside]
(15) Tri Charmed
one witchy sort of stuff
Charmed-ø.SUBJ have-3.SG that witchy sort of stuff-ø.oBJ
[Charmed (television show) has that witchy sort of stuff ]
Thus our main analysis in § 5 considers 191 sentences, all of which mark at least
one GF by case.
5. The analysis
In (16) we show the distribution of syntactic structures among the four informants.
In order to avoid cluttering up a single table, we consider first only argument functions, and then deal separately with the various positions of the nonargument function ADJ. This explains why the xP canonical word order stage with its typical
ADJ SVo structure is missing in (16).
As expected, by far the most frequent way of organising syntax is by canonical word order with all the informants, although it is worth noticing that all three
early bilinguals dominant in English tend to avoid the more frequent null-SUBJ
sentences typical of Serbian monolingual speakers, and use overt pronominal or
referential SUBJ more often than the late bilingual nicole does – especially Trish,
who uses this structure more than twice as much. Pragmatically marked structures
are much rarer, both those that place GFs other than SUBJ in preverbal position
and those that place SUBJ in postverbal position for lexical requirements.
Regarding noncanonical word orders, we observe several notable differences
between our informants. Foremost is the fact that nicole’s range is much wider
than that of the other three bilingual speakers. For example, the sentences in (17)
6. The development of case: a study of Serbian in contact with Australian English
(16) Cross-sectional study: distribution of argument functions among structures and informants
illustrate clearly how nicole is the only informant producing postverbal SUBJ in
the context (‘the bear said/replied/asked’) of the Goldilocks story retelling task. In
this context even Don avoids postverbal SUBJ, although he is clearly able to use it
with unaccusative Vs, as in (18).
(17) a. nic odgovorio je
mali medved
small bear-noM.SUBJ
[the small bear has replied]
b. Don mali medved
small bear-noM.SUBJ
[the small bear says]
c. ned najmanji medved
smallest bear-noM.SUBJ be-3.SG.AUx said
[the smallest bear has said]
d. Tri
mali medveć1
little bear-noM.SUBJ
[the little bear has asked]
be-3.SG.AUx asked
1 Medveć is the informant’s’s version of the diminutive medvedić. Case, however, is used
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
(18) Don došo
muški medved
do svoje sup
come-3.SG male bear-noM.SUBJ next to his soup-oBL
[the male bear came towards his soup]
Furthermore, nicole is the only informant who topicalises oBL, as shown in (19).
(19) nic na mojoj stolici
on my chair-LoC.oBL be-3.SG.AUx someone-noM.SUBJ sat
[someone has sat on my chair]
Finally, ned and Trish use only one noncanonical structure each, neither of which
involves postverbal SUBJ, as shown in (20). on the other hand, nicole and Don
use four and five postverbal SUBJ structures respectively, as already exemplified in
(17a) and (18).
(20) a. ned istoriju
history-ACC.oBJ (you) know like-1.SG
[history, you know, I like]
b. Tri ona
to sve
she-noM.SUBJ be-3.SG.AUx that all-ACC ate
[she ate all that]
Let us now look at the position of ADJ in the production of our informants. In
Serbian there are no constraints on where it can be placed, and speakers are free to
place it according to their discourse or pragmatic need. However, the three positions
ADJ can occupy with reference to canonical word order (after, within, and before)
gain different significance in terms of developmental syntax. If ADJ follows canonical word order, it belongs to the early syntactic stage, which clearly also Trish has
safely reached. If ADJ occurs between the core GFs, namely between either SUBJ
and V, or V and oBJ, it is a sign that learners are starting to free up the SVo block,
and GFs are somehow no longer assigned by position alone. However, in this case
ADJ does not compete with SUBJ for the association with the initial DF ToP. Also
when ADJ is before V in a sentence with null SUBJ it is not, by itself, sufficient evidence that a learner has reached the xP canonical word order stage. Since this latter stage involves the separation of SUBJ and ToP, we assume that the passage of
stage is clear only when SUBJ is present and competes for prominence with ADJ
as ToP in sentence-initial position.2 So, if ADJ precedes SUBJ, there is a more sub2 In § 3.1, ch. 3, this volume, Di Biase & Bettoni place the ToPADJ Vo structure at
the higher xP canonical word order stage. Be as it may, unequivocal evidence for the
passage of stage is provided when SUBJ is present and competes with ADJ as ToP.
6. The development of case: a study of Serbian in contact with Australian English
stantial sign of having reached the xP canonical word order stage. Finally, any other
position of ADJ occurring with marked orders is no longer developmentally significant, because by then learners have freed up fully the more important core GFs. In
(21) we show where our informants stand with regard to the placement of ADJ in
the 64 sentences containing this GF among the total 191 ones.
(21) Cross-sectional study: distribution of ADJ among structures and informants
In the table we can see again how, compared to Trish and ned, nicole and Don
use a wider range of positions, including at least one example each of topicalised
ADJ with referential or pronominal SUBJ. ned too has two examples of this structure, as in (22), but Trish has none, although she can vary the ADJ position within canonical word order as in (23a-b), and can topicalise ADJ in a null-SUBJ sentence as in (23c). We explain Trish’s and ned’s smaller range of structures by the
fact that their functional assignment relies more on position than on case.
(22) ned prvo je
sela na najveću stolicu
first be-3.SG.AUx sat on biggest chair
[first (she) sat on the biggest chair]
(23) Tri a. moja
yellow suknju
my-noM only-ADJ have-3.SG yellow skirt-ACC
[my (girl) only has a yellow skirt]
b. ja
one lolies
I-noM like-1.SG a little bit-ADJ those lolies-ø
[I like those lollies a little bit]
c. samo
jednu glavu
only-ADJ see-2.SG one head-ACC
[you only see one head]
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
In (24) we show how our four informants can handle case marking of direct
(SUBJ and oBJ) and indirect (oBLCASE and oBLPP) argument functions. not
included in the analysis is case marking in ADJ, even when these are expressed
by PPs rather than adverbs because it would not provide more information
about the use of case than oBL already does when it too is expressed by a PP.
Furthermore, cases are counted with a + sign when appropriately used in a given
context, whereas the brackets around a case after the figure preceded by the sign indicate which case is used inappropriately. Appropriateness, however,
should not be confused here with accuracy (Pienemann 1998). That is, keeping
in mind that, as we have already mentioned in § 2, a single inflectional morpheme in Serbian may be used for several morphological contrasts (e.g., number and gender, as well as case), we accept as appropriate any target case ending
regardless of gender, number etc. When these are hard to factor out, we give the
speaker the benefit of the doubt.
(24) Cross-sectional study: distribution of case markers in relation to argument functions
among informants
Predictably, nicole’s case marking system is fully accurate. Furthermore, her
range of case use is wider than that of the other informants, including also InST
and DAT. Whereas InST occurs within a PP, she is the only informant who uses
DAT in order to mark an oblique argumental function (oBLCASE), as shown in
6. The development of case: a study of Serbian in contact with Australian English
(25), rather than marking it by a oBLPP. As a matter of fact, also other informants mark oBLLoC by means of DAT more than once, as in tata osto kući (‘father
stayed home’), but we consider this frequent use of kući formulaic, and ignore
it in our analysis.
(25) nic mama medvedica
rekla medvedu
mother bear-noM.SUBJ be-3.SG.AUx said bear-DAT.oBL
[mother bear has told the bear]
All other informants show some uncertainty in the use of case markers. With Don
and ned, errors in case marking are limited to PPs, and thus attributed to the lexical entry of individual prepositions, with little relevance for functional assignment,
and consequently minimal risk for misunderstanding, in so far as the presence of
the preposition facilitates comprehension of the intended meaning (Anđelković
2000); an example is shown in (26), where in oBLLoC ned uses ACC instead of
(26) ned neko
sedeo na njegovu *stolicu
someone-noM.SUBJ be-3.SG.AUx sat
on his chair-ACC.oBL
[someone has sat on his chair]
With Trish, however, inaccurate case marking affects also core arguments.
Although all her SUBJs are marked as noM, as (24) shows, and her range of
noM~ACC distinctions is quite large, including both nominal and pronominal elements, as shown in (27), she marks oBJ twice with a wrong case: once with noM
and the other with GEn, as shown in (28a-b respectively).
(27) Tri a. veći medveć
spasio je
onu malu
bigger bear-noM.SUBJ saved be-3.SG.AUx that little-ACC.oBJ
[the bigger bear has saved that little one (girl)]
b. ona
she-noM.SUBJ be-3.SG.AUx tried
[she has tried that third (soup)]
(28) Tri a. ona
samo drži
she-noM.SUBJ only hold-3.SG.AUx
[she is only holding a little dog]
b. onak
[then they saw beds]
onu treću
that third-ACC.oBJ
mali *ker
little dog-noM.oBJ
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
This uncertainty in case marking displayed by Trish is also evident in several PPs,
where she tends to oversupply ACC. As a matter of fact, it would seem that Trish is
still at the stage of noM vs nonnoM variation along the morphological schedule,
and that her nonnoM forms only happen to coincide with ACC, DAT and GEn forms
in the native language – as suggested by one of the anonimous referees of this chapter. notice also that Trish is the informant who most often leaves constituents
unmarked by case by using English, as shown in (29). The use of this type of
codeswitching from the dominant language into a heritage language is a well attested characteristic of first and second generation migrants’ speech in general, in
Australia (Bettoni 1991; Clyne 2003) as elsewhere (Auer 1998; Li Wei 2006). But
there can be no doubt that the more frequent the codeswitches, the weaker the case
system (cf. Schachter’s (1974) avoidance strategy).
(29) Tri ja
visual arts
I-noM.SUBJ like-1.SG visual arts- ø.oBJ
[I like visual arts]
Even if not altogether apparent in the tables and examples presented so far, the 191
sentences analysed for this study provide us with further evidence that the
Australian young bilinguals’ case marking system is less reliable than nicole’s, who
is a late bilingual with Serbian schooling. For example, compare nicole’s discontinuous oBJ constituent in her oSV sentence in (30a) with Don’s SVo canonical
sequence in uttering the very same referential content in (30b). Even though Don’s
sentence is grammatically correct, there can be no doubt that it lacks expressiveness
compared to nicole’s.
(30) a. nic moju je
pojeo celu
my-ACC be-3.SG.AUx soup-ACC.oBJ someone-noM.SUBJ eaten entire-ACC
[someone has eaten my entire soup]
b. Don neko
pojeo celu moju supu
someone-noM.SUBJ be-3.SG.AUx eaten whole-ACC my-ACC soup-ACC.oBJ
[someone has eaten my whole soup]
one could argue that English interference can explain both reliance on SVo and
caseless transfers. Yet, in our data there are other structures where all three
Australian bilinguals seem to have problems despite a similarity between their two
languages. A case in point is possessive GEn as a marker of dependency from n.
Don produces three sentences in which GEn correctly marks the possessor (as the
English morpheme –s does), but sometimes insecurity with the case lets him add
the preposition the preposition od, as in (31a), which turns out to be ungrammat-
6. The development of case: a study of Serbian in contact with Australian English
ical because an n+n dependency would be expected by native speakers in this context. ned seems to ignore the issue by marking all elements in the oBJ constituent
with ACC in (31b), thus missing the n+n dependency. And Trish seems to lose
track of her sentence by using a preposition and then marking little (girl) with ACC
instead of GEn in (31c).
(31) a. Don na kraju
išla od malog deteta na *stolici
in end-LoC.ADJ be-3.SG.AUx gone of little child-GEn on chair-LoC.oBL
target:na kraju
išla na stolicu
malog deteta
in end-LoC.ADJ be-3.SG.AUx gone on chair-ACC.oBL little-GEn child-GEn
[in the end she went to the little child’s chair]
b. ned jedna
*ljubičastu *boju odelo
one-noM.SUBJ have-3.SG.AUx purple-ACC colour-ACC outfit-ACC.oBJ
ljubičaste boje
one-noM.SUBJ have-3.SG.AUx outfit-ACC.oBJ purple-GEn colour-GEn
[one (woman) has a purple coloured outfit]
c. Tri
njena majka,
od te Goldilocks *malu,
her mother-noM.SUBJ of that Goldilocks little-ACC (girl) be-3.SG.AUx came
target:njena majka,
majka maleGoldilocks,
her mother-noM.SUBJ,mother-noM little-GEn Goldilocks, be-3.SG.AUx came
[her mother, little Goldilocks’ mother, came]
6. Conclusion
In this chapter we have dealt with a case marking system within the PT framework,
and considered both morphological and syntactic aspects of its development. our
example of the Serbian language is particularly challenging for the learner (and the
researcher), not only because, morphologically, there are several cases which mark
n dependency on several lexical categories and whose formal features are fusionally ‘mixed’ with those of gender, number and class features, but also because, besides
a more systematic, productive use of default case assigned syntactically, there are
numerous instances of lexically triggered nondefault case which override syntactic
assignment and must be learned individually, or at best bundled up in small sets.
We have then offered some hypotheses for the development of the Serbian case
marking system, both morphological and syntactic, and tested them on cross-sectional data pertaining to four bilingual informants whose knowledge of Serbian is
already fairly advanced.
our analysis of the data available cannot provide evidence for the full developmental path that lies ahead for learners acquiring Serbian as a heritage language.
Bruno Di Biase, Camilla Bettoni and Lucija Medojević
nevertheless, they confirm several of our PT-based hypotheses. First, syntactically,
no informant producing structures at the noncanonical word order stage has not
already safely in place those at the canonical word order stage, sequencing verbal
arguments in canonical order. Secondly, the position of the nonargument function
ADJ, freely placed according to discourse or pragmatic reasons, is the playground
for allowing learners to free up the default canonical word order. Thirdly, no
informant who displays a full range and accurate use of case markers does not
exploit the possibility of deploying them in order to produce noncanonical word
orders. on the other hand, the opposite seems to hold true. our informants whose
case system still shows some gaps or formal inaccuracies rely on the rigidity of position within canonical word order to identify GFs, or else appeal to the semantically more transparent PPs.
In sum, focusing on how V constructs the relationship with its arguments by
means of case, PT has allowed us to show how a wider range and a more reliable
deployment of formal cases correlates with a stronger possibility of exploiting them
to enhance discourse and pragmatic choices beyond canonical word order, thus
allowing expressiveness for the speaker without compromising comprehension by
the listener. This in turn has allowed us to discriminate between advanced and less
advanced heritage speakers of Serbian, and propose a new approach to the analysis
of a minority language. In our approach, contrary to Dimitrijević’s (2004a, 2004b;
Dimitrijević-Savić 2008), the contact between the two typologically different languages (i.e., nonconfigurational Serbian and configurational English) plays a lesser
role than immature development, which is constrained by a migration environment offering ‘situation-bound’ language exposure and arguably a much-reduced
input in the heritage language, both in quantity and quality. Further research focusing, for instance, on the comparative development of both the minority language
and the dominant language once the child begins his/her school experience may
clarify the role of the specific sort of input afforded by instruction (but cf.
Medojević 2014). We leave it to more robust data – gathered longitudinally or
cross-sectionally among more numerous learners with more varied competence levels – to test the full range of developmental hypotheses concerning the acquisition
of the case marking system of Serbian (and other languages), and thus confirm, or
falsify, our claims.
The authors wish to thank Rosanna Benacchio (University of Padua, Italy) and Lidija
Krebs-Lazendić (University of new South Wales, Australia) for their helpful suggestions
and proofreading of the Serbian language.
Exploring the acquisition of differential object
marking (DOM) in Spanish as a second language
Bruno Di Biase* and Barbara Hinger**
*University of Western Sydney, **University of Innsbruck
1. Introduction
The aim of this chapter is to explore the acquisition of differential object marking (DOM) in Spanish L2 and thus probe the higher level boundaries of the PT
framework. In coining the term DOM, Bossong (1983-84, 1991) presented
cross-linguistic data on more than 300 languages presenting this grammatical
characteristic, whereby direct objects (OBJs) of transitive Vs either remain
unmarked or are overtly marked by case or agreement on the basis of some
semantic or pragmatic feature. This marking has since attracted considerable
attention in linguistic theory (e.g., Aissen 2003; Dalrymple & Nikolaeva 2011;
Leonetti 2004; Torrego 1998, 1999, among many others). Unlike the many
purely structural approaches, Dalrymple & Nikolaeva’s (2011: 1-2) point out
that DOM, in the many languages where it manifests itself, “encompasses syntactic, semantic and informational-structural differences between marked and
unmarked objects”. So they propose that marked OBJs are associated, synchronically or historically, with the information-structure role of topic. Where
the connection between marked OBJs and topicality has been lost through
grammaticalisation, marked OBJs become associated with semantic features typical of topics, such as animacy, definiteness and specificity (Dalrymple &
Nikolaeva 2011:1-2). Spanish is one of the languages exhibiting DOM, whereby some OBJs are marked with the preposition a, also known as marked accusative (Torrego 1998), prepositional accusative, personal a, or accusative a
(Montrul & Bowles 2008; Tippets 2011).
Given our interest in acquisition, we note that Spanish DOM is highlighted in the
literature as difficult to acquire, not only for English L1 learners of Spanish L2
(Bowles & Montrul 2009; Guijarro-Fuentes 2011, 2012; VanPatten & Cadierno
1993), but also, perhaps more surprisingly, for early bilingual ‘heritage’ speakers of
Spanish in the USA (Montrul 2008; Montrul & Bowles 2009; Montrul &
Sanchez-Walker 2013; Silva-Corvalan 1994). For L2 Spanish, Farley & McCollam
Grammatical development in second languages, 213-242
Bruno Di Biase and Barbara Hinger
(2004) confirm this difficulty for a-marked OBJs. Their findings however show no
support for the PT-based schedule they derive from Johnston (1995) and
Pienemann (1998), who place a-marked OBJs well before subjunctives, which are
at the top stage in PT. To our mind this is not surprising because the earlier version
of PT locates structures on a single developmental axis, whether they are obligatory or optional, whether declaratives, negatives, interrogatives, or other pragmatically motivated constructions. Thus earlier PT is not equipped to deal with optional
and interface phenomena, including Spanish DOM. On the other hand, the current proposal for PT by Bettoni & Di Biase in chapter 1 of this volume offers principled explanations and more attuned predictions in the area of syntax-semantic
and syntax-discourse interfaces.
This chapter then sallies into an exploration of the rather controversial area
of case marking, itself a relatively new area in PT, but limited to DOM. More
specifically it will attempt to show that the current version of PT is better suited to account for the difficulties in learning differential case-marking, which we
see as a structure located at the interface of syntax-semantics and syntax-discourse. Similar observations about the difficulties advanced adult L2 learners
face when acquiring the interface of syntax with other cognitive domains led
Sorace & Filiaci (2006) to propose the Interface Hypothesis, attributing the difficulty, possibly, to computational limitations in integrating multiple sources of
information. These difficulties have been variously interpreted also in terms of
‘incomplete acquisition’ (e.g., Montrul 2008), from lack of access or partial
access to universal grammar, or some kind of ‘representational deficit’ (Clahsen
& Felser 2006), and in terms of the Feature (In)accessibility Hypothesis
(Guijarro-Fuentes 2011, 2012).
We interpret these difficulties in processing terms, in the sense that the computational complexity created by the requirement to integrate discourse information with syntactic information makes processing in real time harder for the learner (Hopp 2007), as opposed to the native speaker, who has already automatized the
necessary underlying processes. It is also plausible, as Wilson (2009) claims, that
the additional computational complexity created by the attempt to integrate different layers of information challenges the ability of the learner to allocate cognitive
resources appropriately. For example, competing constraints from the L1 may
interfere on how to interpret the semantic or discourse requirements. Indeed both
resource limitation and resource allocation may contribute simultaneously to the
learner’s difficulties. We will not delve any deeper into our processing option versus representational deficits as an explanation of the learners’ difficulty with interface structures. We will instead suggest to place Spanish DOM, a structure sitting
at the interface between syntax and semantics/discourse, within a current PT-based
schedule for L2 Spanish, and propose some initial empirical tests for our position.
In the remainder of this chapter we will first offer a quick sketch of Spanish and
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
the intriguing nature of its DOM, followed by a brief review of the acquisitional
literature (§ 2). Next, we will present our developmental schedules for L2 Spanish
over which DOM is distributed (§ 3), and empirically test them in a small-scale
cross-sectional study of oral production of six Austrian students of Spanish as a foreign language (§ 4).
2. Spanish and its Differential Object Marking
According to Tippets (2011: 107), Spanish uses the preposition a to mark “human
accusative (direct) objects. This transparent example of DOM relates to the animacy status of the accusative object”. The facts of Spanish, however, also support the
proposition that neither all animate OBJs are a-marked nor that all inanimate
OBJs are not a-marked. But before we zoom into this intriguing grammatical area
let us zoom out to a brief overview of the language.
A pluricentric language, Spanish is by far the most widely spoken of the
Romance languages, and the national language of 18 countries as well as Spain,
with large Spanish-speaking minorities in the USA, and significant minorities in a
number of other countries in the world (cf. Green 2011 for an accessible overview
of this language). Spanish is also, naturally, subject to regional and sociolinguistic
variation. It shares many characteristics with other Romance languages (cf. Italian,
ch. 3, this volume) including nonconfigurationality, null SUBJ phenomena, a rich
agreement morphology and pronominal clitics. Following Green’s (2011) description and modelling on his examples, Spanish word order is not fixed by grammatical requirements at a particular point in the sentence, which differentiates it from
French and even more from the Germanic languages. But it has strong constraints
within the main syntactic constituents, and its theoretical word order freedom is
subject to pragmatic conventions: themes precede rhemes and new information is
placed towards the end of the utterance.
In canonical word order, OBJs and complements follow V, as in (1a-b). Given
the tendency for SUBJ and TOP to coincide in spoken language, the SVO order
in (1a) is very frequent, whereas SOV in (1b) is register-dependent (e.g., in poetry
for rhyming reasons); VOS in (1c) would sound very odd, and VSO (1d), although
common in more formal registers, would signal contradiction or contrast in everyday language (‘it was Juana, not Carmen, who painted a car’).
(1) a. Juana
un coche
Juana-SUBJ painted-3.SG a car-OBJ
b. Juana
un coche
Juana-SUBJ a car-OBJ
Bruno Di Biase and Barbara Hinger
c. pintó
un coche
painted-3.SG a car-OBJ
d. pintó
un coche
painted-3.SG Juana-SUBJ a car-OBJ
Similarly to Italian (cf. § 3.1, ch. 3), OBJ (when definite NP or proper N) can be
topicalised by placing it at the beginning of the sentence, as in (2a), with an intonation break after TOP and an ACC clitic obligatorily marked on V.1 As Green
(2011) notes, the result of this ‘clitic copying’ is no longer a simple sentence but a
complete and perfectly grammatical structure in its own right. In other words, an
XP (los coches/the cars) is added to a complete S, which here comprises a transitive
V with its OBJ clitic marker coreferential with the external XP, followed by SUBJ.
Another important point is that, when OBJ is topicalised, SUBJ is postverbal as in
(2a), assuming a secondary2 TOP role. In any case the sentence is complete even
when the overt TOP is dropped (2b) because the obligatory OBJ is supplied by the
clitic (los) displaying anaphoric agreement with TOP.
(2) a. los coches,
the cars-PL.MASC they-ACC.PL.MASC painted-3.SG Joan
b. los
SUBJ is obligatorily postverbal also in presentationals (3a) and in content questions
(3b). However, again following Green (2011), interrogatives should not be
assumed to entail SV inversion because postverbal SUBJ frequently occurs in statements,3 and polar questions may show VS or SV order and rely on intonation to
differentiate from statements.
(3) a. hay
en Sydney
there are many-PL.MASC bridges-PL.MASC in Sydney
1 Like Italian, Spanish is a head-marking language, which therefore can mark OBJ morphologically on V (i.e., the head).
2 As Dalrymple & Nikolaeva (2011: 53-4) remind us, “(t)he topic role is not necessarily
unique”. Along with other scholars (e.g., Givón 1983; Lambrecht 1994; Polinsky
1995), they distinguish at least primary topic and secondary topic.
3 A class of unaccusatives Vs with SUBJ=FOC normally exhibit a VS order, similarly to
Italian (cf. § 3.2, ch. 1, this book): in the sentence llegó el jefe (‘arrived the boss’),
postverbal SUBJ is the unmarked position.
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
b. qué
Juana ?
Joan ?
As for morphology, Spanish explicitly and consistently marks number and gender
on all modifiers within NP, as well as number and person, and occasionally also
gender, between SUBJ and V, as (4a) shows. This provides optimal grounds for
testing classic PT. Interestingly for our concern in this chapter, Spanish – unlike
Italian (cf. § 2.1, ch. 3, this volume) – has no anaphoric agreement between lexical V and OBJ, not even in constructions topicalising OBJ, as is confirmed by the
grammaticality of (4b), where comido (eaten) does not carry the plural and feminine features of las manzanas (the apples).
(4) a. las manzanas
the apples-PL.FEM are 3.PL ripe-PL.FEM
b. las manzanas
han comido los monos
the apples-PL.FEM they-ACC.PL.FEM have eaten
the monkeys-PL.MASC
[the apples have been eaten by the monkeys]
Possible ambiguities in distinguishing between SUBJ and OBJ are resolved by syntactic differences in two important ways, both connected to specificity (Green
2011). The first syntactic difference is that the Spanish SUBJ NP – whether definite, indefinite or generic – requires a determiner, whereas the OBJ does not. This
is shown by the grammaticality of (5a), where the SUBJ el hombre (‘the man’)
appears with the determiner, and the OBJ manzanas (‘apples’) without it. The OBJ
without the article (manzanas, ‘apples’) responds to the question ‘what did the men
buy?’, and is hence a characteristic nonTOP, focused OBJ. On the other hand, (5b)
responds to the question ‘what happened?’. Hence the OBJ las manzanas (‘the
apples’) is again a nonTOP, but neither is it FOC because, as ‘event reporting’
(Lambrecht 1994), the whole sentence is in focus (new information). With regard
to SUBJ, (5c) is ungrammatical because the SUBJ hombres (‘men’) is without a
determiner, even though it is generic.
(5) a. el hombre compró
the man bought-3.SG apples
b. el hombre compró
las manzanas
the man bought-3.SG the apples
c. *hombres compraron manzanas
bought-3.PL apples
Bruno Di Biase and Barbara Hinger
The second syntactic difference distinguishing SUBJ and OBJ is at the core of
our concern in this chapter. We illustrate it in (6a-c), where OBJ has an inanimate
referent, in contrast to (7a-c), where the a-marking is traditionally attributed to
semantically human/animate OBJs. Inanimate OBJs such as coche (‘car’) in (6b),
on the other hand, would be ungrammatical with such mark. The completeness
and coherence of (6c) is provided by the ACC clitic lo (‘it’) ensuring that the transitive V pintó (‘painted’) does have an agreeing OBJ anaphora.
(6) a. Juana
un coche
painted a car]
b. *Juana pintó
a un coche
[Joan painted (a) a car]
c. Juana lo
Juana it-ACC.SG.MASC painted
[Joan painted it]
On the other hand, the human (and specific) OBJ a su hermano (‘her brother’) is
correctly marked, differently from the inanimate OBJ which requires no a mark.4
In any case, it is possible to confirm that un coche (‘a car’) in (6a) and a su hermano
(‘her brother’) in (7a) are both OBJ because both their equivalent sentences without the overt OBJ in (6c) and (7c) use the same ACC clitic lo (‘it’/‘him’).
(7) a. Juana pintó
a su hermano
[Juana painted (a) her brother]
b. *Juana pintó
su hermano
[Joan painted her bortehr]
c. Juana lo
Juana he-ACC.SG.MASC
[Juana painted him]
However, comparing the sentences in (7) with that in (8) we find the same a mark
4 Inanimate OBJs may be marked with a when the speaker wishes to give them a human
character, as in era como si abrazase a un árbol o a una roca [‘it was like embracing ‘a’
a tree or ‘a’ a rock’] (Real Academia Española 2010: 659).
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
(8) Juana le
dió el sombrero a Pancho
Juana he-DAT.MASC.SG gave the hat
to Pancho
[Juana gave the hat to Pancho]
Now, according to Green (2011: 213) “since a is also the preposition used to introduce datives there is no overt difference between the majority of direct and indirect
objects.5 Whether the categories have genuinely fused or are merely obscured by
surface syncretism is hard to say.” Green then goes on to point out that, whereas
most of Latin American varieties maintain separate third person pronominal ACC
and DAT clitics, much of Spain has lost this distinction. This last point finds some
confirmation in Hualde, Olarrea & Escobar (2001: 342-360), who describe a
range of variations in the use of ACC and DAT clitics in their overview of bilingual
and contact situations involving Spanish in Latin America, as well as Central
America and the USA.
To make DOM more of a puzzle, as we have already mentioned, animate
OBJs are not always a-marked, as (7) above may imply. The OBJ there is specific
enough (Juana’s brother), but animate OBJs need not be marked if they are not specific, as (9a) shows in contrast with (9b), where the speaker has in mind specifically a lawyer or a particular kind of lawyer, even if not explicitly declared. In (9c)
however we find evidence against the necessity of an animate OBJ having to be specific to be a-marked, as one could hardly say that the lawyer referred to is specific.
(9) a. necesito un abogado
[I need a lawyer] (any, not specific, not a-marked)
b. necesito a un abogado
[I need PREP a lawyer] (specifically a lawyer, not a doctor, a-marked)
c. no necesito a ningún abogado
[I don’t need PREP any lawyer] (animate, not specific, a-marked)
Thus, neither animacy nor specificity turn out to be a failproof guide to a-marking of OBJ. What may then be the function of the a-marker in (9b-c)? It would
seem that this set of contrastive examples actually supports Dalrymple &
Nikolaeva’s proposal that it should be treated as a marker of topicality.
Another somewhat related motivation for a-marking is what Tippets (2011)
calls ‘relative animacy’ of SUBJ and OBJ as in (10a), where the a-marked OBJ todos
5 In the LFG framework used in this book, direct (ACC) and indirect (DAT) objects are
OBJ and OBLDAT respectively.
Bruno Di Biase and Barbara Hinger
(everybody) is higher in the animacy scale (human) than the SUBJ la lluvia (inanimate) and is made prominent by means of the prepositional marker. This could be
covered under the ‘animacy’ feature of OBJ, disregarding the ‘inanimacy’ of the
SUBJ. However, [-animate] OBJs may also be a-marked, as in (10b), where both
SUBJ and OBJ are neither ‘personal’ nor animate, and neither could be said to be
specific (cf. also the anthropomorphizing example in note 4).
(10) a. la lluvia mojó a todos
[the rain drenched PREP everybody]
b. la noche sigue al dia
[the night follows PREP the day] (al = preposition a + article el)
c. quiero un perrito
[I want a puppy dog]
d. quiero a un perrito
[I love PREP a puppy dog]
This is an interesting cue to some of the language-internal reasons why a-marking
may be used: given its flexibility, Spanish word order may be an insufficient cue to
subjecthood, or objecthood, so morphological highlighting (or prominence) comes
to the rescue and disambiguates for the listener which of the two participants is not
the SUBJ – recall that SUBJ is an obligatory argument, unrestricted, and never
prepositionally marked. Again, we have a case where topicality seems the neater
functional explanation for a-marking of OBJ.
Finally, for our brief survey of a-marking, the lexical meaning of the V itself
may also guide the choice of DOM in subtle ways, as shown in (10c-d), where the
intended meaning of quiero (‘I want’ vs ‘I love’) guides the choice.
In sum, Spanish DOM is neither as straightforward nor as uniform as the
learner or the analyst might wish. Nevertheless, the intersection of semantic (animacy) and discourse-related factors (specificity, SUBJ-OBJ relationship, secondary
prominence) seem to lend support to Dalrymple & Nikolaeva’s (2011) claim that
marked OBJs are associated with the information structure role of topic.
Given such complexities it is interesting to look into what the actual DOM
behaviour of native speakers might be and how children learn it. Fortunately, following Laca’s (2006) lead and other researchers’ attempts to characterise factors that
are relevant to Spanish DOM beyond anecdotal and constructed examples, Tippets
(2011) conducted a quantitative investigation of spoken corpora from three different major Spanish-speaking metropolitan areas: Buenos Aires, Mexico City and
Madrid. He includes in his analysis only Vs that display nonuniform marking of
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
OBJ, and quantifies them over four basic factors: (i) animacy of OBJ, whether animate or inanimate; (ii) specificity of OBJ, whether specific or nonspecific; (iii)
form of OBJ, whether a proper N or a lexical N; and (iv) relative animacy, that is,
the degree of animacy of OBJ relative to SUBJ – the latter being a discourse related ‘global’ factor, compatible with Darlymple & Nicolaeva’s (2011) relating DOM
to secondary topics (cf. also Leonetti 2004). Tippets (2011) found a large measure
of agreement on animacy6 as the primary factor favouring DOM across all three
cities. He also found, significantly, that relative animacy is a stronger factor in
Buenos Aires and Madrid than ‘local’ (i.e., lexical) animacy. In Tippets’ own words,
relative animacy
compares the animacy status of the subject and the animacy status of the DO
using the animacy scale human > animate > inanimate (see Comrie 1979; Næss
2007). For example: El partido reune a los amigos. In this case the DO, amigos
(human), is higher than the S, el partido (inanimate), and is coded as a as such.”
(Tippets 2011: 109)
Furthermore this discourse-based factor, that is, relative animacy, ranks first in
both Buenos Aires and Madrid, and second in Mexico City. Animacy of OBJ, on
the other hand, ranks first in Mexico City and second in the other two cities.
Another significant surprise in this survey is the absence of specificity of OBJ as a
factor in the Madrid corpus. Finally, marked OBJs are between 33% and 39% of
all OBJs across corpora (bearing in mind that Tippets 2011 considers only Vs that
do present a-marking across corpora). Interestingly, between a minimum of 17%
and a maximum of 28% of animate OBJs are not a-marked across dialects. Thus
Tippets recognises the prescribed use (animate OBJs should be marked and inanimates should not), but he also laments that such prescription often fails to reflect
actual usage, and summarises the situation as follows:
It is not unusual to find human accusatives unmarked as well non-human
accusatives marked irrespective of animacy in spoken and written Spanish.
Additional “exceptional marking” may occur as a result of ambiguities that arise
from the relative flexibility of subject and object position and verbs with the personal a also serves a disambiguating function (2011:107).
Thus, Tippets’ survey confirms, on the one hand, the strong presence of DOM in
Spanish, but also its variation in usage across Spanish-speaking countries, with statistically significant differences between the three ‘dialects’ (Tippets’ term) considered, as well as a clear connection with discourse.
6 Inanimate OBJ, albeit infrequently, is confirmed to be a-marked between 5% and 8%
in the three corpora.
Bruno Di Biase and Barbara Hinger
It is not surprising, then, that the conditions under which OBJ must be
marked, must not be marked, or is optionally marked, have been vigorously discussed by many scholars, constituting arguably one of the most debated topics in
Spanish grammar over the last 200 years, as noted by Rodriguez-Mondoñedo
(2008). Recent work (e.g., Torrego 1998) attempts to tackle this issue by identifying a number of conditions that influence DOM in Spanish, such as animacy,
specificity, agentivity and the semantics of the predicate, that is, the type of V
involved. Each of these conditions, if valid in many or most uses of DOM, does
not appear to exhaust the full range of uses (cf. Rodriguez-Mondoñedo 2008).
On the other hand, Tippets’ work clearly points towards an account that may not
be exhaustively satisfactory by appealing simply to a number of structural conditions, however refined, without bringing to bear regional variation and discourse
We will shy away from reviewing further the huge literature on Spanish DOM
here, and note, much more modestly, that its reliance on a particular lexical feature
such as animacy, coupled with its relative indeterminacy – Aissen (2003) would say
‘fuzziness’ – and its relation to topichood, makes Spanish DOM quite complex to
learn in spite of its deceptively simple morphological form. Taken together, these
factors place DOM high in the Spanish acquisitional schedule because it presupposes, minimally, that functional assignment is in place. In fact, the learner must
be able to distinguish not only between SUBJ and OBJ, but also between different
types of OBJ (animate vs inanimate), their specificity, and ultimately their discourse status as a possible secondary topic, which we mark here as TOP2.
Furthermore, learners will need to disentangle a-marked ACC from the equally amarked DAT (OBLDAT is also typically human) in order to use the appropriate clitic
anaphora – a moving target itself, at least in Spain, as Green (2011) points out.
Now let us consider briefly the acquisition of DOM by various types of learners: L1, L2 and Heritage Language speakers (HS). Rodriguez-Mondoñedo (2008)
notes that, in spite of the vigorous debates on Spanish DOM, not much is known
about its acquisition by children. In looking at the available CHILDES corpora,
this scholar found sufficient longitudinal data in four Spanish L1 children, and
counted all their instances of OBJ that should or should not be a-marked to see
whether errors may be detected. For this reason he discounts, to our mind unfortunately, all cases where the marking is optional. From his careful analysis,
Rodriguez-Mondoñedo (2008) comes to the conclusion that “children master
Spanish DOM with a performance virtually errorless.” This conclusion, however,
is based on a conflation of results for both default and nondefault cases. Of course,
both must be accounted for, but one thing is to note that children do not mark
what is not marked in the input (the default case), and another is to find out how
they learn to deal with the marked cases. Averaging all out, the strength of the
default may well obliterate important facts. Wholly constructed from Rodriguez-
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
Mondoñedo’s (2008) data, although following the presentation commonly used in
this volume and elsewhere for SLA results, in (11) we present the pattern for his
marked cases only.
(11) DOM in L1 Spanish children (after Rodriguez-Mondoñedo 2008)
What seems to be clear from this perspective is, first, that the acquisitional pattern
is neither as strong nor as uniform across these four children as the author asserts,
and secondly, that the error rate for some of them is not insignificant, particularly
for Emilio and Koki. Emilio, as it turns out, marks DOM at less than chance.
Why? Rodriguez-Mondoñedo (2008) explains that Emilio is a Spanish speaking
boy living in a Catalan community, and that Catalan does not have DOM, then
he adds: “It has been observed already that Spanish-speaking children in bilingual
communities (when one of the languages does not have DOM) drop a occasionally (Silva-Corvalán 1994; Luján 1996; Montrul 2004 …)”. The other interesting
case is Koki, a Mexican child with Spanish-speaking mother and American
English-speaking father. Previous to the recordings in Mexico, the family lived for
the first six months in Poland, the next six in Argentina and two months in the US.
Koki’s accurate use of DOM is lower than that of the other children, except Emilio.
The bilingual home and the different environments of exposure may have contributed to her pattern of DOM acquisition. This is an interesting set of facts,
which together with Emilio’s may help understand the consistently low accuracy
with DOM – compared to native speakers within a single environment – that
Montrul and her colleagues (e.g., Montrul 2008; Montrul & Bowles 2009;
Montrul & Sanchez-Walker 2013), after Silva-Corvalan (1994), repeatedly find in
the US with young HS, whether simultaneous or sequential bilinguals, including
adult immigrants after a certain period of residence in the US.
Montrul and her colleagues attribute the difference in DOM to ‘incomplete
acquisition’ or later ‘attrition’ (in the case of adult immigrants), which in turn may
be due to the reduced quantity and quality of the input received by the HS bilinguals from adult HS whose spoken Spanish may itself be ‘attrited’ (except for recent
adult arrivals). Young HS bilinguals may also undergo attrition, not unlike adult
HS, under the influence of, and in convergence with, English, the dominant lan-
Bruno Di Biase and Barbara Hinger
guage, which has no DOM. But, as Montrul & Sanchez-Walker (2013: 127) candidly admit, DOM in Spanish “is conditioned by unclear semantic and pragmatic factors. We still do not know whether DOM is primarily regulated by type of
verb, type of object, other elements in the sentence or all of these”.
The nature of the Spanish-speaking environment should also be taken into
account, we believe. One would wish to ask which speakers do HS communicate
with in their Heritage Language? In the US there are Spanish speakers from all over
the Spanish-speaking world, which would surely not be a homogenous Spanishspeaking community, as Tippets (2011) demonstrates specifically regarding the
way DOM is used. On the other hand, Mexican native speakers living in Mexico
can be assumed, unlike their US counterparts, to be using their language in a
homogenous community, so a comparison between the two groups of speakers
needs to be moderated by this basic difference. Overall, then, it would be surprising if HS, adult or younger bilinguals, in the US used DOM according to the
structural factors (such as animacy and specificity) indicated by Montrul and her
colleagues. Variation, as found in their results, seems to be a reasonable expectation.
Furthermore, the type of task (grammaticality judgement) and the presentation of
results by group means rather than by individual speakers used in their research
paradigm would tend to obscure critical developmental or environmental (sociolinguistic) differences among speakers, resulting from their individual experience
and pattern of exposure, as we just saw with the four children looked at by
Rodriguez-Mondoñedo (2008).
Naturally enough those who have worked with Spanish L2 have found DOM
rather elusive and difficult for the learners. We will look here at two studies, that
is, one by Guijarro-Fuentes (2011, 2012) and the other by Farley & McCollam
(2004). Guijarro-Fuentes proposes an interesting approach which he calls the
Feature (In)accessibility Hypothesis, whereby DOM is the result of a number of
interacting features subsumed by OBJ, some of which are structurally given (such
as the grammatical relationship with the V, e.g., its ‘direct’ rather than ‘oblique’
nature of the relationship), and others are inherent in the particular OBJ (such as
animacy and specificity). The latter features, which Guijarro-Fuentes calls ‘interpretable’ features in his research paradigm (cf. Lardiere 2008), are not all equally
accessible to learners. He finds that ‘animacy’ is the most accessible (or learnable)
of these interpretable (or inherent) features.
Farley & McCollam (2004) report on an experiment involving 29 students.
They explicitly use DOM (which they call ‘personal a’) and subjunctive marking
to test PT’s Teachability Hypothesis (Pienemann 1984). Their conclusion is that
results do not support PT as presented by Pienemann (1998) and call into question the hierarchy for Spanish proposed by Johnston (1995). Following
Johnston’s develpomental hypotheses, Farley & McCollam reason that ‘personal
a’ is at a much lower stage (Johnston’s stage 4, corresponding to the phrasal pro-
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
cedure stage in this volume) than subjunctive marking (Johnston’s stage 7, corresponding to the S-BAR procedure stage). Then, following the logic of
Pienemann’s (1984) Teachability Hypothesis (i.e., learners cannot ‘skip’ a stage
even with instruction), they ensure that their learners are all ‘ready’ to learn
Johnston’s stage 4 (personal a), but that, conversely, none is ready for subjunctive
marking on V in subordinate clauses. The students are then divided into four
groups of 6 or 7 students each: three of the groups were instructed in each of the
two structures with different treatments (explicit instruction, structured input,
and processing instruction, VanPatten 1996, 2007), and the control group
received no instruction on those structures. Results are collated in (12), which
shows the number of learners who are deemed developmentally ready for the relevant structure, and the number of learners who have actually developed it after
the treatment. Results are clearly more favourable for subjunctive marking (c.
60%) than for ‘personal a’ (c. 40%), even though the learners were deemed to be
ready for DOM but not for subjunctive.
(12) Global results from Farley & McCollam (2004) teachability experiment
Given our discussion above regarding the nature of DOM, it seems that Farley &
McCollam (2004) place it too early in the developmental path. DOM is not just
morphological marking of OBJ when it is animate because, as pointed out above,
not all animate OBJs are a-marked – they must also be specific, among other
things. Furthermore, some of the tasks used in their study require a much higher
structure involving the primary topicalisation of a-marked OBJ plus the cliticOBJ,
as in (13), which introduces an extra complicating factor.
(13) a
los viejos
no los
los jóvenes
a-DOM the old-PL.MASC no they-Cl.ACC.PL.MASC respect-3.PL the young-PL.MASC
[(as for) the older people, young people don’t respect them]
This is definitely not a phrasal procedure stage structure because it calls for higher
level resources, in particular the S-procedure and full functional assignment (cf. Di
Biase & Kawaguchi 2002, and § 3.1, ch. 3 this volume, for a parallel structure in
Bruno Di Biase and Barbara Hinger
Italian). That is, the learner must distinguish not only SUBJ from OBJ, but also
differentiate among types of OBJ, one of which must be also distinguished from
OBLDAT . The initial XP a los viejos (‘[a] the elderly’) looks like a ‘garden path’ DAT.
If differentiation fails, the clitic may be wrongly case-assigned. On top of all that,
the construction in (13) also displays marked word order: TOPOBJ ClACC-V SUBJ.
Hence, it is definitely on the higher stages of L2 learning. This discussion naturally leads into our PT-based hypothesis, which takes into account the relevant facts
for a more precise location of DOM in the development of L2 Spanish.
3. The developmental hypotheses
The structure we are focusing on in this chapter – that is, the language-specific
DOM – is expressed by means of a morphological marker on OBJ. As such, we
begin by placing it in (14) within our morphological development hypothesis
for Spanish L2, in line with the current proposals for PT outlined in chapter 1
of this volume (cf. also the Italian schedule, ch. 3). That is, the morphological
schedule, whose structures are obligatory, is presented separately from the syntactic schedules based on the Prominence Hypothesis and the Lexical Mapping
Hypothesis which, on the other hand, represent speakers’ choices largely related to discourse-pragmatic factors. Hence, given that DOM sits at the intersection of the syntax-semantics and syntax-discourse interfaces, we need to discuss
it within the Prominence Hypothesis and the Lexical Mapping Hypothesis.
Like other morphological schedules in previous chapters of this volume,
the one for Spanish exhibits a developmental path beginning with single words
and formulas followed by lexical-level morphemes such as the plural –s in Ns.
Spanish has a largely regular masculine/feminine gender marking and its plural
morpheme –s is fairly easy to segment. For example, maestra/maestro, female and
male ‘teacher’ respectively, become maestras/maestros for their plurals. For these
procedures requiring lemma access and category marking no unification of features is required.
Learners who are able to process the next procedure, that is unification of
features within the phrase, are minimally able to unify the value of one feature,
such as [plural], at the phrasal node. Our schedule in (14) reflects this agreement at the NP node where the attributive plural form americanos (‘American’)
agrees with the head N coches (‘cars’), or the VP agreement between the predicative adjective grandes (‘big’ PL) with the plural form of the copula son (‘are’).
This is the stage at which Farley & McCollam’s (2004), following Johnston
(1995), placed DOM. In our view, it is possible that a-marked OBJs may
appear at this stage – as well as OBLs expressed by PPs (e.g., dative, locative) –
but only if DOM were understood simply as a structural morphological mark-
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
(14) Developmental stages hypothesised for L2 Spanish morphology, identifying the location of
er within the VP. However, we have seen that other syntactic considerations are
involved, which require full functional assignment, e.g., the differentiation of
OBJ from SUBJ via SUBJ-V agreement, which in turn relies on the S-procedure being in place. This is where the positional lock on SUBJ and OBJ is
opened and even a postverbal SUBJ may be still identified as such. Thus DOM
is best placed at the S-procedure stage, where unification happens at the node(s)
adjoining different phrases.
In our schedule then, at the S-procedure stage, as well as sentences exhibiting the familiar SUBJ-V agreement, we may find also DOM, both in its canonical postverbal position (but note that OBJ does not require any morphological
Bruno Di Biase and Barbara Hinger
agreement in such case) or as primary TOP in preverbal position, in a sentence
equivalent in structure to (13), an example borrowed from Farley & McCollam’s
(2004) DOM tasks. In this case, unification must occur between the gender and
number of the N in the NP a la maestra (‘PREP the teacher’) and the ACC clitic
of V. The full justification for placing DOM here rather than at the phrasal procedure stage will be clearer shortly when we deal with syntax.
The final stage on our morphological schedule is aligned with Farley &
McCollam (2004) and Bonilla (2015), and concerns the S-BAR procedure stage,
which in Spanish contains those constructions in which the relationship between
main and subordinate clauses must be marked through V inflection. Thus, in
(14), the subjunctive V form tenga (‘should have’) in the subordinate clause is
compatible with the epistemic me parece ridículo (‘to me, it seems ridiculous’) in
the main clause.
As we argued above, DOM is an optional structure sitting at the interface of
the syntax-discourse and syntax-semantics interface, hence it will be further discussed in the next two schedules for syntactic development. In (15) the developmental progression covered by the Prominence Hypothesis for declarative sentences is broadly laid out (cf. § 4.2, ch. 1, this volume, and also §§ 3.1, ch. 2 for
English, ch. 3 for Italian, and ch. 4 for Japanese).7 This hypothesis relates to word
order, canonical and otherwise, which respond to discourse-pragmatics constraints. These may also operate through other means, and prominence may be
achieved by means of prosody, for instance, or indeed morphology, as well as word
order. Thus we now focus on the location of DOM within the Prominence
Hypothesis schedule in (15).
In this schedule we hypothesize, as with other languages, that learners will
start from the canonical word order, which in Spanish is SVO. Furthermore,
because Spanish is a prodrop language, the majority of canonical utterances are
VO (cf. Bonilla 2015). In principle there is no reason why DOM should not be
located here. However, at this stage learners have not learned yet to differentiate
anything much inside the canonical order, and distinguish OBJ from SUBJ
merely by their position. So all OBJs are equally underspecified and their animacy or inanimacy would make no difference for early-to-high intermediate learners, especially if their L1 does not differentiate such OBJs (which is the case, e.g.,
for English, Japanese and Catalan). Likewise, at the next stage, a topicalised
adverbial phrase leaves the core elements of the canonical block unaltered. So, if
a potentially differentiable animate OBJ is produced, again the learner would not
7 Questions are extensively treated elsewhere in this book; cf. ch. 2 for English, and ch.
8 for Italian respectively, but also ch. 9 for German.
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
(15) Developmental stages hypothesised for L2 Spanish syntax, identifying the location of DOM
within the Prominence Hypothesis
be able to differentiate it inside the blocked string. It is important to note, however, that at both stages we will find animate arguments which are not a-marked
but nevertheless grammatical, such as in (16).
(16) a. Juan tiene amigos
[John has friends]
b. ahora Juan tiene dos hermanos
[now John has two brothers]
As we have seen in § 2, the feature [+animate], however central, is insufficient
by itself for OBJ to be a-marked and needs to be accompanied by other features
such as [+specific], which is linked to discourse. So we place DOM at the next
stage, the noncanonical word order stage, even though its descriptor may not be
quite adequate given that the actual word order for sentences like Juan vio a la
maestra (‘Juan saw PREP the teacher’) looks perfectly canonical. Recall that passives also look pretty canonical in this regard (e.g., in English as well as in
Spanish), but this does not mean that we place them at the default stage. In any
case, this is the stage at which full (i.e., independent of position) functional
assignment must be assumed to free up word order, because GFs are marked
morphologically or by other means. The learner at this stage must also be able
Bruno Di Biase and Barbara Hinger
to differentiate core arguments from each other and from non-core arguments
as well as match prominence features, that is, [±prominent, ±new], with the
appropriate constituent, whichever the order. Notice in particular that this amarking has little or no role to play in the grammatical relations of the OBJ
(which remains OBJ whether or not it is a-marked). What the preposition
marks here, by morphological means, is the ‘secondary prominence’ of an argument mapped onto an OBJ that shares in the ‘aboutness’ of the sentence, and
often, but not always, is animate and agentive-like (cf. morphologically marked
prominence in Japanese, ch. 4, this volume). This a-marking is maintained also
when the speaker promotes such a differentially marked OBJ as primary topic,
as in (13). So our representation of constructions such as these shows up in both
the morphological and the syntactic schedules, because the former, in (14),
reflects the required feature unification ensuring that the TOP constituent is
correctly interpreted as NON-SUBJ, and the latter, in (15), reflects discoursegenerated word order differences. In the end, because we characterise the differentiated OBJ as bearing a TOP2 function in the spirit of Dalrymple &
Nikolaeva (2011), it seems reasonable to place it within the Prominence
Hypothesis schedule and at the noncanonical stage, in contrast to earlier interpretations of it as an exclusively morphological structure.
Let us now turn, briefly, to the Lexical Mapping Hypothesis for Spanish in
(17), because DOM straddles across the syntax-semantic interface as well. We
noted for instance that the largest measure of agreement on the actual usage of
DOM reported by Tippets (2011) across three metropolitan varieties of Spanish
relates to a lexical semantic feature, that is, [animacy], with the highest ranking
for the discourse-based ‘relative animacy’ in two of the metropolitan areas considered (Buenos Aires and Madrid) and second rank in the remaining area
(Mexico City).
After the universal single word and formula stage, the development of lexical
mapping begins with V representing actions/states and with default mapping of the
higher ranking agent/actor on N preceding the theme/patient represented on the
other N. The prevailing VO default is unlikely to pick a differentially marked OBJ
since the majority (61% to 67%, according to Tippetts 2011) of OBJs in the input
are not marked. Hence learners are hypothesized to treat all OBJs in the same way,
whether animate or inanimate, specific or not specific, definite or indefinite. In
other words, the links to critical semantic features and discourse are, at this stage,
absent. Proceeding upward, learners begin to increase their stock of complements
with arguments additional to their default string. Crucially, additional arguments
must be differentiated from the default core arguments, particularly from OBJ
given the prodrop nature of the Spanish SUBJ. This is exactly the stage where the
semantics of the additional arguments, through their prepositions, makes more
transparent their relationship with V. So OBLDAT for goal or beneficiary will com-
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
(17) Developmental stages hypothesised for L2 Spanish syntax, identifying the location of DOM
within the Lexical Mapping Hypothesis
monly be animate, as in (18a), whereas OBL for location, instrument, and so on,
will not, as in (18b).
(18) a. José da pan a los niños
[José gives bread to the children]
b. ahora vivo en Barcelona
[now I live in Barcelona]
These OBLs are usually processed after the core argument(s) and are differentiated
from OBJ on account of position, as well as by the presence of a preposition. Given
that this stage is still characterised by default mapping (plus additional argument),
it is unlikely that DOM could be acquired at this point, because the semantic role
mapped on OBJ (theme or patient) is still assigned by default and the additional
arguments need to be differentiated from it. Nevertheless, the marking of different
OBLs with a variety of prepositions is another step towards distinguishing animate
arguments (other than SUBJ) from inanimate ones. In other words, the learner is
moving towards building the semantic interface with syntax, a resource for becoming able to mark OBJs differentially.
Bruno Di Biase and Barbara Hinger
The next and highest stage on this schedule includes all nondefault mappings
which account for alternative, pragmatically motivated constructions such as passives and complex constructions such as causatives. This higher stage is where hierarchical correspondences (such as the agent role mapping on the SUBJ function)
are subverted if the speaker’s communicative intention or lexical choice requires it.
Hence we can expect at this stage to find scope for a more confident differentiation of arguments, including differentiation between OBJs, mostly involving animacy and specificity among other constraints. It seems to us then a fairly safe stage
at which to place the emergence of DOM on this developmental schedule,
although its marginal optionality may still elude or confuse the learner for a long
time, leading to persistent errors.
In sum, our presentation of the three developmental schedules for morphological development and syntactic development along the Prominence Hypothesis
and the Lexical Mapping Hypothesis allows us to hypothesize, in (19), an implicational hierarchy for acquiring DOM in L2 Spanish.
(19) no marking of OBJ > marking of [±animate] OBLs > DOM > topicalised DOM
Overall one must bear in mind, however, that achieving either the S-procedure
stage or the distinction between animate and inanimate arguments does not guarantee that DOM will be fully deployed by the learner, given, among other things,
its interface with discourse prominence. Positive evidence will be the morphological a-marking of OBJ with transitive Vs. A higher level of evidence is obtained
when the learner also produces a TOP which is a-marked (even if inanimate),
which is coreferential (i.e., it shares number and gender features) with the ACC clitic
marker of OBJ on the V, as in (13) above.
4. The study
Our cross-sectional study of six learners of Spanish L2 is guided by a single research
question: do learners of Spanish L2 acquire DOM at the hypothesised developmental points in the PT schedules proposed in § 3?
4.1. Informants and tasks
The six learners in this study are Austrian German-speaking students. Their participation in the research project was voluntary and they all signed the relevant consent form. All of them are 19-year old females who have been studying Spanish in
a formal school setting for three years, totalling approximately 260 hours of
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
instruction, attending the same Spanish class in the same school given by the same
teacher and thus having been exposed to the same classroom material. In addition
to formal instruction, half of the students had gained exposure to natural input in
Spain: two of them, BEN and JAN (all names are code names) spent three months
in Spain, and one (VES) spent one month there, as can be gleaned from a questionnaire administered to participants to gather information about their language background. Instruction sessions over the two semesters of observation were digitally
audio-recorded and later transcribed and analysed. From this analysis it transpires
that explicit instruction on the subjunctive and DOM, the structures on focus in
this article, did not receive the same time allocation. Subjunctive constructions
were dealt with extensively from the end of the second year of instruction onwards,
whereas DOM was only dealt with in the second year and to a much lesser extent
(Hinger 2011).
In order to elicit spontaneous oral production data, the learners were administered two tasks in a single session outside of their classroom environment for 15
minutes on average. These task sessions also were digitally audio-recorded and later
transcribed and analysed. The first task is a classical communicative information
gap taken in pairs, the second a monologic task. Students picked their pair themselves with no linguistic criteria applied in reference to pairing. The two tasks consist of written and visual prompts “providing information about the context, the
content and the purpose” (Tankó 2005: 42) of the oral production. The topics chosen referred to contents of the ongoing Spanish classroom teaching, though neither
the task types nor the task goals and requirements had in any way been rehearsed.
The monologic task aims at eliciting oral production relating to the environmental problems caused by pollution, and the learners are given a visual and written
prompt (in Spanish) on the consequences of pollution for humanity, and are
encouraged to provide ideas to help stopping climate change. The interactive task
focuses on the situation of renting a room in the target country, namely in the capital of Spain, Madrid. The two partners were given different roles described on
cards written in Spanish. Partner A was provided with the information of living on
her own in a flat which she wants to share with a Spanish native speaker because
she spends some time in Madrid in order to improve her language knowledge. A
picture of the flat was also shown to partner A for her to describe the flat to partner B, who phoned after reading an announcement in the paper. So, partner B has
the role of a Spanish native speaker hunting for a room in Madrid. She is especially attracted by an announcement in the paper made by an Austrian because she had
spent some time in Austria and was keen to practice her German. In the end, she
decides not to hire the room and needs to find some explanations. In both tasks,
the researcher role was that of facilitator and occasional interlocutor in order to
help overcome situational problems such as anxiety by reassuring and motivating
learners, nonverbally and verbally, to continue their talk.
Bruno Di Biase and Barbara Hinger
The tasks were successful in eliciting a corpus of a total of 4,564 tokens and
1,529 types of Spanish lexical items distributed in 486 clauses, an average of about
94 clauses per informant, as shown in (20). The minimum number of clauses
uttered is 35 by GRA, the maximum 151 by JAN. These differences cannot be
attributed to the three month experience in Spain by JAN because informant BEN,
for instance, also spent the same amount of time in the target country and produced 80 clauses and 738 tokens, both being less than those of informant ROS,
who had never visited a Spanish-speaking country.
(20) The corpus for the six learners
4.2. Results and analysis
In this section we present the results concerning first morphological development
and then syntactic development.
With regard to their current morphological development in terms of classic
PT (Pienemann 1998), all of the informants have achieved the S-procedure stage,
and three of them also the S-BAR procedure stage, as shown in (21).
(21) Cross-sectional study of morphological development in L2 Spanish
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
Looking at the figures within the S-procedure stage, we can see that three learners
(GRA, VES and ROS) are still uncertain with the agreement between SUBJ and
adjective, but they do, like all the others, manage unification on the plural agreement between SUBJ and lexical V without fail.8 It may be the case that the copular clause absorbs greater processing resources because COP itself, placed between
the SUBJ and the predicative, also has to unify its [number] feature with both the
SUBJ and the predicative adjective. A possible case of a ‘soft barrier’ (cf. § 5, ch.1
and further ch. 2, § 2 for English, and ch. 3, § 2.2 for Italian, this volume).
Nevertheless, despite the low return on this structure, some good indicators of
functional assignment are in place, as exemplified in (22a-b).
(22) a. TAN los profesores
no son idoles [target: idolos]
the professors-PL no are-PL idols-PL
[the teachers are not idols]
b. VES
los jóvenes
pueden cambiar algo
the young-PL can-PL
change something
[young people can change something]
What seems to be rather harder to produce at this stage is topicalisation involving unification between an XPTOP with the clitic marking the OBJ (23a) or
OBL (23b) function on V, because the clitic must have both the grammatical
function (ACC or DAT respectively) that the displaced XP would have in its
default position in the sentence, and the same number and gender features as
the N in the XP.
(23) a. TAN Schwaz ah
Schwaz-SG.MASC ah it-Cl.ACC.SG.MASC know-1.SG
[Schwaz (a town in Austria), I know]
b. JAN
a ella
también le
gustan las plantas
to her-3.SG.FEM also
she-Cl.DAT.3.SG.FEM like-3.PL the plants-PL
[she also likes plants]
Topicalisations involving OBJ or OBL are quite rare: only 6 were found in the
whole database, and notice also that none of the learners produced a topicalised differentially marked OBJ. These constructions are other prime candidates for ‘soft
8 Note that SUBJ & V agreement is only counted when SUBJ is expressed lexically or
pronominally. Getting the right V-form with null SUBJ is part of categorial learning
(cf. § 2.1, ch. 3 on Italian, this book).
Bruno Di Biase and Barbara Hinger
barriers’ because, additionally to feature unification, they involve word order choices triggered by discourse-pragmatic constraints.
At the next, S-BAR procedure stage, we find three learners. One is ROS, who
satisfies the emergence criterion for subjunctive morphology because she has three
positive occurrences with lexical and structural variation (cf. § 5, ch. 1, this volume). Her accuracy, however, does not go beyond 50%. The one thing she misses
in three of her otherwise well constructed sentences is precisely the morphological
shape of V in the subordinate clause, as in (24), where she provides the default
indicative rather than the subjunctive V-form.
(24) ROS no creo
que necesito
lámparas todo el día
no believe-1.SG that need-1.SG.INDICATIVE lights
all day
[I don’t believe that I need lights all day]
The other two learners also produce several contexts of this higher level structure, and the task canvassing opinions about a controversial issue which young
people care about was clearly successful in pushing informants to attempt it.
However, again, like ROS in (24), BEN’s morphological V-form is the default
indicative in one case and the required subjunctive mood in two cases, as in
(25a), whereas JAN consistently marks V in the subordinate clause with the subjunctive mood, as in (25b).
(25) a. BEN no hace falta que cocinen
una cosa exótica
not has need that cook-3.SG. SUBJUNC a thing exotic
[there is no need (that she) cooks an exotic thing]
b. JAN
es muy importante que la gente
más respecto
is very important that the people-SG have-3.SG.SUBJUNC more respect
[it is very important that people show more respect]
Let us now turn, to our learners’ syntactic development with results shown in (26)
focusing on DOM, that is, to the TOP2 features of some animate OBJs, as identified in the Prominence Hypothesis in (15).
As we have mentioned, the overall frequencies of DOM structures are not
abundant in our corpus. This is due, at least in part, to the fact that, as in other
chapters of this volume, the counting for valid syntactic structures includes lexical V to the exclusion of copular and presentative sentences, whose word order
responds to specific constraints and are preferably analysed separately.
All our learners produce canonical word order, which is the foundational syntactic stage, and at least one structure from the next stage, the XP canonical word
order stage. Four of them also produce at least one OBJ or OBL topicalisation, as
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
(26) Cross-sectional study of syntactic development in L2 Spanish identifying the position of
DOM within the Prominence Hypothesis
exemplified in (23), which characterises them as higher intermediate learners. As
for DOM, all learners produce at least one potential context for it, that is, a sentence with an animate OBJ with a sufficient specificity load, but not all of them amark it. GRA and VES for example never do – regardless of animacy or specificity
they seem to regard all OBJs the same, as exemplified in (27a-d).
(27) a. GRA tengo
otras ofertas
have-1.SG other offers (not animate)
[I have other offers]
una amiga
b. GRA tengo
have-1.SG a friend (animate, not specific)
[I have a friend]
c. GRA voy
a llamar
mi amiga
to call ø-DOM my friend (animate and specific)
[I am going to call my friend]
d. VES pero ahora no
otra persona
but now no
have-1.SG ø-DOM other person (animate and specific)
[but now I don’t have anyone else]
It is crucial to note here that (27a-d) provide a contrastive key to our analysis: as
against (27a) with a [–animate] OBJ, both (27b and c) have animate OBJs.
However, una amiga (‘a friend’) in (27b) is not specific, and is therefore counted
among the 6 canonical word order structures produced by GRA in (26). By con-
Bruno Di Biase and Barbara Hinger
trast, mi amiga (‘my friend’) in (27c) by virtue of the possessive determiner is quite
specific, and is therefore counted as –1 for OBJTOP2 [= DOM] because it should
have been a-marked. Similarly, otra persona (‘another person’) in (27d) is counted
as –1 because it appears to have sufficient specificity in the discourse to be amarked as OBJTOP2, but VES does not a-mark it. An interesting case is provided
by ROS in (28), where la tierra (‘the earth’) is a [–animate] OBJ, but the intended
meaning (‘I love the earth’ rather than ‘I want the earth’) would likely trigger DOM
(cf. the contrast in (10c-d) and note 4). Given that this learner has achieved the top
stage of morphological development (subjunctive marking), it is plausible to
assume that she could mark DOM when discourse conditions may require it, but
she does not.
(28) ROS yo quiero
la tierra
love-1.SG ?ø-DOM the earth (not animate, specific, prominent)
[I love the earth]
Four out of our 6 learners behave within this pattern of not differentiating OBJs.
The exceptions are TAN and JAN, with one and three differentially marked animate
OBJs respectively, as in (29a-c), where DOM is produced in what appears to be
the appropriate environment for it.
(29) a. TAN conozco a
algunas personas
know-1.SG a-DOM some people
[I know certain people]
b. JAN
yo conozco a
un médico
I met-1.SG a-DOM a doctor
[I met many people from all over the country]
c. JAN
mucha gente de todo el país
met-1.SG a-DOM many people from all the country
[I met many people from all over the country]
Notice that the two learners who produce DOM are the only ones who produce
more than one topicalised OBJ (TAN) or OBL (JAN). Notice also that JAN, the
only informant with sufficient evidence for marking TOP2, does not produce a
DOM structure as primary TOP, and neither does anybody else. Such production would be the best evidence that the learner may not be conflating DOM
with DAT because the clitic anaphora in such cases would have to be ACC even
though the coreferential DOM TOP is expressed by a PP. We nevertheless
choose to leave the structure in our implicational hierarchy, to be tested on a
richer database.
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
Out of the four learners who do not produce DOM, two of them (ROS and
BEN) appear to be in a state of readiness for DOM, both producing at least one
topicalised structure with a DAT clitic, as in (30a-b). Though both expressions
sound fairly idiomatic, not all learners use them.
(30) a. ROS a ti
to you you-Cl.DAT.2.SG interest-3.SG much
[to you it is very interesting]
b. BEN a mí me
to me I-Cl.DAT.1.SG
[to me it is the same ]
To round up the result section we now present in (31) the analysis for the Lexical
Mapping Hypothesis as shown in (17).
(31) Cross-sectional study of syntactic development in L2 Spanish identifying the position of
DOM within the Lexical Mapping Hypothesis
We cannot help but notice again the paucity of production in the nondefault mapping stage. Given the highly marked or inherently lexical nature of constructions
such as passives, causatives, impersonals, and exceptional Vs, this is not surprising
in spontaneous speech. As a matter of fact, only JAN produces a passive construction, and all other figures in the top row of (31) refer to three Vs: gustar (‘like’),
interesar (‘interest’) and parecer (‘seem’), which map the lower ranking theme role
on postverbal SUBJ. DOM constructions are also placed in this schedule because
of their semantics/syntax interface character, as discussed at the bottom of § 3.
Their figures, naturally, are the same as in (24), which looks at them in terms of
TOP2. Here it is worth noting that JAN, the strongest DOM performer, is also the
one with the greater range of nondefault constructions.
Bruno Di Biase and Barbara Hinger
4.3. Discussion
Our developmental results for morphology in (21) and syntax in (26) and (31) are
compatible with the PT hypotheses in (14)-(15) and, essentially, also (17). We can
see from (21) that morphological resources are available to all learners up to the Sprocedure stage, so they can all handle functional assignment unequivocally. Three
learners have also achieved the higher S-BAR procedure stage, which means they
can handle at least some subjunctive constructions. This state of affairs appears to
be corroborated for syntax in (26), in so far as four learners are able to use some
marked word orders, which also assume unequivocal functional assignment, and in
(31), which shows they can handle nondefault mapping operations.
Despite the high stages achieved, only two of the learners apply DOM. The
other four do produce some contexts for DOM, but do not a-mark OBJ. In both
the learners who use DOM, the particular feature that seems to be relevant is animacy, but also specificity. This result partly coincides with Guijarro-Fuentes’
(2011, 2012), as well as with Tippets’ (2011) indication that animacy is the primary factor favouring DOM in his corpus-based study. The other features, such as
specificity, may be tied up more with the current discourse model of the speaker.
So it seems appropriate to place DOM high in the syntactic sequence within the
Prominence Hypothesis space, where full functional assignment is assumed, and
learners become capable of assigning prominence to any arguments. We have
argued that DOM is to be placed also at the top stage of the Lexical Mapping
Hypothesis, and this also has some support in the results. If we treated DOM as a
purely morphological marker we would place it, with Johnston (1995) and Farley
& McCollam (2004), at the phrasal stage because it is part of PP within VP. As we
have seen this is not the whole story, and our results confirm the multifaceted
nature of DOM. At bottom its use or nonuse is at a crossroad, an interface constrained by information which is semantic (animacy), syntactic (unambiguous OBJ
identification) and discourse-related (specificity, definiteness, relative animacy,
prominence), as well as socio-dialectal, as Tippets (2011) amply demonstrates.
Focusing on discourse now, discourse-related processing instructions (e.g.,
[make argument X prominent]) are part of the preverbal message, which instigates
the lexical selection process (cf. Levelt 1989: 98-99; cf. § 2.1, ch. 1, this volume).
These instructions about the assignment of prominence, itself an option exercised
by the speaker, will then be carried out by the lexical choice. The chosen lemma in
turn will include its morphological instructions – in the case of DOM whether the
a marker is required or not.9 In other words, the lemma receiving the prominence
9 This reflects our theoretical position, which should be subjected to research into the
processing of prominence – as yet an under-represented area in speech processing
7. Exploring the acquisition of differential object marking (DOM) in Spanish as a second language
instruction checks its grammatical function (i.e., OBJ) and, if SUBJ is already syntactically prominent (i.e., default TOP), then OBJ assumes its prominence (i.e.,
TOP2) by other (e.g., prosodic and morphological) means, which in Spanish is the
so-called personal a.
Indeed the morphology results indicate clearly that it is not enough for learners to be at the interphrasal stage – or even the interclausal S-BAR stage (stage 7 in
Johnston’s (1995) stages) – to be able to mark OBJ differentially. Both ROS and
BEN, for instance, are at the top of the morphological schedule, and yet, having
produced clear contexts for DOM, they are not using it. On the other hand JAN,
like BEN, is able to mark S-BAR procedure as required for many subjunctive clauses in Spanish, but, unlike her, she does mark DOM. More surprisingly TAN, who
is not at the S-BAR stage for morphology, is nevertheless able to mark DOM when
the occasion arises. Her single occasion is valid evidence because she is also capable
of topicalising OBJ.
So, in light of the similarity with Farley & McCollam’s (2004) results considered in § 2, we may ask whether DOM is a necessary resource for processing subjunctive clause marking. The answer appears to be negative, because DOM is to
some extent optional. Furthermore, DOM is just one of the structures requiring
the S-procedure, and PT predicts that S-BAR procedure structures will appear after
the emergence of any S-procedure structures (all learners in this study have
acquired one or more S-procedure structures). Hence there is no contradiction
within PT if a learner does not exhibit DOM and achieves subjunctive marking at
the same time. Our results appear to support the specific implicational progression
proposed in (19) within an overall developmental framework. The last step
hypothesised in this progression awaits validation from a richer database.
On the overall evidence, it may be said that the S-procedure is a necessary
resource for DOM to be marked, but not a sufficient condition, since discourse
factors are also involved. This means that there must be a sufficient degree of control, or automatisation, of more basic (e.g., lexical, phonological, and morphological) components for the learner to be able to integrate specific discourse-pragmatic information (cf. the discussion on the interface between morphological and syntactic development in § 4.3, ch. 1, this volume).
5. Conclusion
In this chapter we have engaged in an exploration of the acquisition of DOM in
L2 Spanish and shown that current PT is capable of accounting for the optionality of its use in processing terms – that is, discourse-pragmatic information needs to
be integrated online with semantic and syntactic information, which in turn
requires morphological resources.
Bruno Di Biase and Barbara Hinger
In terms of our research question – whether learners of L2 Spanish acquire
DOM at the hypothesised developmental points in the PT schedules proposed in
§ 3 – the answer is positive. We have thus placed DOM within a developmental
framework and indicated a specific implicational trajectory. Despite an important
limitation in terms of data robustness, it is also clear that our results are compatible with those found by Guijarro-Fuentes (2011, 2012), in the sense of pointing
to animacy as the default feature for marking DOM, as supported also by Tippets’
(2011) corpus work. They are also similar to Farley & McCollam’s (2004), but our
approach offers a detailed and multifaceted explanation for the phenomenon. With
more robust, possibly longitudinal, data sets other acqusitional issues may be
addressed, such as those surrounding DAT marking and clitics, as pointed out
repeatedly in Montrul’s work, the sociolinguistics of DOM in expatriate communities and its development in L1, as initiated by Rodriguez-Mondoñedo. Fully
developmental studies are then desirable, and may clarify complex issues better
than highly focused work on single phenomena.
Acquiring constituent questions in Italian
as a second language
Camilla Bettoni and Giorgia Ginelli
University of Verona
1. Constituent questions
In Levelt’s Model, the speaker must decide early whether the message to be
conveyed is declarative, imperative, or interrogative (Levelt 1989: § 3.5), so
indicators of mood are already present when the message generated in the conceptualizer reaches the formulator, and activates functional processing and
positional processing (cf. § 2.1, ch. 1, this volume). In constituent questions
the information the speaker wants to receive from the addressee is expressed by
one (or more) question phrase (QP from now on) or word (QW), which fills
the gap left by the focal constituent (or constituents) of the corresponding
declarative sentence (Van Valin 2001: 185). We can tell that the interrogative
phrase fills this gap by the fact that we cannot put another phrase of the same
type in the same sentence. For example, the sentence in (1) is ungrammatical,
because V can govern only one GF per type, with the exception of the ADJ
function (Darlymple 2001: 11). Here the V kiss governs the OBJ function, and
the presence of two OBJ arguments in the same sentence (whom and the girl)
is ungrammatical. On the other hand, (2) is grammatical because each QP
(what and where) fills the two required GFs (OBJ and OBL respectively) governed by the V put.
(1) [John kissed the girl]
*whom did John kiss the girl?
(2) [John put the book onto the table]
what did John put where?
Constituent questions are used to request some piece of new information, the
focus of the sentence. Focality is not a prerogative of interrogatives, but in interrogatives FOC is obligatory (Choi 1999), and undergoes specific constraints.
These vary cross-linguistically: all languages can satisfy the communicative need
of requesting information, but different languages do it using different linguis-
Grammatical development in second languages, 243-258
eUrOsLA MOnOGrAPHs serIes
Camilla Bettoni and Giorgia Ginelli
tic strategies. A derivational approach describes these strategies in terms of syntactic movement of QPs (cf. Chomsky 1977, as well as Cheng & Corver 2006;
rizzi 1997), and assumes a typological division between languages that formulate constituent questions by placing interrogative phrases in initial position
(simple- or multiple-fronting languages), and those that leave them in the same
position of their non-interrogative equivalents (in-situ languages). In the nonderivational framework of LFG, constituent questions are described as a particular kind of ‘filler-gap structures’ (Falk 2001; Kroeger 2004). These structures
imply the existence of a missing element at the ‘gap’ position, which is the actual specific piece of information required, and of a ‘filler’ bearing two functions:
(i) the DF FOC, associated with the interrogative phrase; and (ii) the GF, associated with the gap. In (2) above, for example, we can see the double status of
the filler (what), bearing both the DF FOC, associated with the interrogative
phrase, and the GF, in this case OBJ, associated with the gap. In LFG’s f-structure, abstract GFs and features try to capture universal syntactic principles that
vary cross-linguistically at other levels of representation. Hence, the f-structure
of constituent questions is independent from language-specific connotations
like, for instance, word order: “since the pragmatic function of constituent questions is much the same in all languages, the functional structure of constituent
questions in various languages is likely to be quite similar even when phrase
structure configuration is very different” (Kroeger 2004: 171; cf. also Bresnan
2001: 45).
According to Mycock (2007), both these approaches – the derivational
approach and the functional one – as previously formulated, fail to explain crucial aspects of constituent questions such as prosody. That is, they can capture
only those principles that underlie the formation of constituent questions in
languages like english that realize the focusing of QPs syntactically. However,
in languages like Japanese, the QP appears in situ, that is, in the same position
it occupies in the equivalent declarative sentence, and the focusing of the interrogative phrase is realized only prosodically. Because the DF FOC is not indicated at f-structure, or indeed at c-structure, Mycock (2007) maintains that
constituent questions can be reduced neither to their word order nor to their fstructure, but must be described in terms of both the informational distribution
at i-structure level, and the ways in which their i-structural status (i.e., FOC) is
marked in syntax and/or in prosody. she demonstrates convincingly that the
only universal feature of constituent questions is the focusing of the QP, and
that what varies cross-linguistically is how it is focused: syntactically, and/or
prosodically. For the theoretical development of incorporating within LFG the
possibility of at least three other structures, besides a-structure, c-structure and
f-structure, see Choi (2001), Falk (2001: 22ff ), and more recently Dalrymple
& nikolaeva (2011).
8. Acquiring constituent questions in Italian as a second language
2. Constituent questions in Italian
Most grammatical descriptions of Italian interrogatives indicate syntax as the key
feature in the formation of constituent questions (e.g., Lepschy & Lepschy 1981;
Fava 1995; serianni 1996; Dardano & Trifone 1997; salvi & Vanelli 2004;
schwarze 2009). This does not mean that prosody does not play a role in Italian,
but only that the most important aspect of Italian constituent interrogatives is
syntax. Leaving prosody for future investigations, we will deal here with the syntactic strategies that in Italian allow for the focusing of the new information
enclosed in QPs.1
(3) F-structure and c-structure of cosa beve Luigi? (what does Luigi drink?)
1 In his taxonomy of Focus, Dik’s (1997: 331ff.) distinction between Questioning Focus
and other types – notably Completive (new) Focus – is insightful.
Camilla Bettoni and Giorgia Ginelli
Typologically, Italian is a headmarking language located towards the less configurational end of the typological spectrum (cf. fig. (1), part II, this volume).
Although its canonical order is sVO, word order is relatively free, regulated
by discourse and pragmatic choices (cf. § 3.1, ch. 3). In constituent questions,
word order is marked, because the QP appears initially in the sentence rather
than in the position typical of its GF. The initial position then, in interrogatives, is associated with the DF FOC. Because Italian is a language with syntactic focusing, FOC is grammaticalized and indicated at f-structure, where it
bears also the GF specified by the argument list of the verbal predicate.
Consider the sentence in (3), for example, with the illustration of its f- and cstructures. Here word order is marked in two ways. First, the interrogative
phrase bearing the FOC function – co-referential with the OBJ gap after the
V beve – appears clause-initially, that is, not in the position associated with its
GF. secondly, sUBJ appears postverbally, that is, not in its canonical preverbal position.
The only exception to the markedness of word order in interrogatives is when
the focal QP itself bears the GF sUBJ, as in (4). In this case, sUBJ is questioned
in situ, clearly showing its special status as the only core GF which, in LFG, is also
a DF – FOC, in this case. Also in english sUBJ is the only constituent that is questioned in situ and does not require AUX, as in who drunk the wine? vs what did he
drink? (cf. Falk 2006 for a different approach to the notion of sUBJ in LFG which
reassesses this apparent anomaly and discusses thematic roles, a-structure, GFs, and
the mappings between them).
(4) chi
il vino?
who.sUBJ drink-3.sG the wine?
[who drinks the wine?]
Word order of Italian constituent questions can be marked in yet a third way if
the interrogative sentence is itself pragmatically marked, and there is a topicalised constituent in first position. Consider the sentence in (5), for example,
where also TOP=OBJ appears in a noncanonical position, besides
FOC=OBLGOAL and sUBJ. The complexity of this structure is further
increased by the fact that in Italian the topicalization of OBJ requires the use of
a clitic as morphological marker of OBJ onto V, as shown in §§ 3.1-3.2, chapter 3, this volume.
(5) il vino
a chi
offre Luigi ?
the wine-TOP.sG.MAsC to whom-FOCOBL it-Cl.ACC.sG.MAsC offers Luigi-sUBJ
[the wine to whom does Luigi offer it?]
8. Acquiring constituent questions in Italian as a second language
3. A developmental hypothesis
For the development of Italian constituent questions by L2 learners, we take as
our starting point PT’s hierarchy presented as the Prominence Hypothesis in
(34)-(35), § 4.2.1, chapter 1, this volume, and implemented for Italian declarative sentences in (23), § 3.1 in chapter 3. Our hypothesis is illustrated by the
hierarchy in (6).
(6) Developmental stages hypothesised for Italian L2 syntax based on the Prominence
Hypothesis: interrogatives (after Bettoni & Di Biase 2011)
studies within the PT framework on the acquisition of Italian simple, pragmatically neutral, minimally presuppositional declarative sentences (e.g., Di Biase &
Kawaguchi 2002; Di Biase & Bettoni 2007; Bettoni, Di Biase & Ferraris 2008;
Bettoni, Di Biase & nuzzo 2009; cf. also § 3.1, ch. 3, this volume) report that initially learners can produce only sentences with canonical word order whose structure has a fixed correspondence between arguments (such as agent, patient) and cstructure. Also questions go through this canonical word order stage. However,
soon after learners have the lexical means to convey an interrogative message and
the morphological means to distinguish between ns and Vs, the canonical word
order stage is quickly overtaken. In fact, constituent questions, as we have seen, are
sentences marked pragmatically, whose very essence is the focalisation of QP.
Camilla Bettoni and Giorgia Ginelli
Thus, at this early stage, fronted QP FOC is followed by canonical word
order. Having yet to learn to assign GFs to verbal arguments independently of position, learners do not master the means to free up the canonical order sequence, as
required by Italian constituent questions. The outcome is then ungrammatical, as
in (7)-(8), when sUBJ (expressed referentially or pronominally) is in preverbal
position without being questioned (that is, when FOC ≠ sUBJ).
(7) *che sport
Alberto fa?
which sport Alberto do-3.sG?
[which sport does Alberto do?]
(8) *dove tu
questi fiori?
where you find-2.sG these flowers?
[where do you find these flowers?]
On the other hand, the outcome is grammatical in either of two cases: when the
sentence is null sUBJ, as in (9), or the QP in initial position bears itself the sUBJ
function, as in (10) or (4) above.
(9) cosa
[what does (she) study?]
(10) chi
questa borsa?
who have-3.sG this bag?
[who has this bag?]
At this same XP canonical order stage, as well as fronted focal QWs or QPs, learners can also produce topical words or phrases. The resulting sentences can then vary
in terms of grammaticality, depending on the GF associated with TOP. When it
bears the ADJ function, which is an ungovernable noncore function, the outcome
is ungrammatical, as in (11)-(12), or grammatical, as in (13)-(14), not because of
the presence of ADJ (which in the target language does not require any adjustment
in the following sequence), but for the same reasons we have just seen regarding the
examples in (11)-(14).
(11) *adesso che sport Alberto fa?
now which sport Alberto do-3.sG?
[now which sport does Alberto do?]
(12) *in giardino dove
tu trovi
questi fiori?
in the garden where you find-2.sG these flowers?
[in the garden where do you find these flowers?]
8. Acquiring constituent questions in Italian as a second language
(13) all’università
cosa studia?
at the university what study-3.sG?
[at the university what does (she) study?]
(14) questa mattina chi
this morning who have-3.sG bought
[this morning who has bought the bread?]
il pane?
the bread?
On the other hand, if TOP bears the OBJ function and its coreferential OBJ clitic
marker on V is missing, the TOP function is left unmarked and the outcome is
ungrammatical, as in (15) – cf. Bresnan’s (2001: § 4.8) notion of ‘functional uncertainty’.
(15) *la cartolina dove hai
the postcard where have-2.sG bought?
[where have you bought the postcard?]
In this example, even if the sentence is ungrammatical, the meaning is quite
clear lexically, just as it is clear with topicalised ADJ. This is because the listener
is unlikely to take the inanimate OBJ postcard as sUBJ doing the action of
buying. Likewise, when TOP is ADJ, the fact that this GF is usually expressed
by a circumstantial adverb or a PP rules out the possibility for the listener to
take it as sUBJ, which by default occupies the first position in canonical order.
Problems of comprehension, however, do arise when both OBJ and sUBJ have
animate referents. In this case, if the coreferential clitic marking the first constituent as OBJ is missing on V, the reader will indeed take it as sUBJ (cf. (19)(21), ch. 3, this volume).
One stage further up in the developmental path, at the noncanonical word
order stage, having learned the necessary morphology to assign GFs to constituents (cf. § 4.2.1, ch. 1, this volume), learners can now mark them by means
other than position, and thus free up canonical word order. By placing sUBJ in
postverbal position, as in (16), they will codify grammatical constituent questions
also when there is a pronominal or referential sUBJ and the fronted QP bears GFs
other than sUBJ.
(16) che cosa compra
buy-3.sG Paolo?
[what does Paolo buy?]
Also at this stage, like at the previous stage, besides focalising QP, learners may wish
to topicalise constituents with a variety of GFs. When TOP is ADJ, as in (17), the
Camilla Bettoni and Giorgia Ginelli
sentence is grammatical because no further adjustment is required. On the other
hand, when TOP is OBJ, should the missing coreferential OBJ clitic leave the GF
unspecified, the sentence would be ungrammatical, as in (18). The agreement
between TOP and the clitic obviously increases the cognitive load of the production of a topicalised constituent question. Thus, although technically still within
the noncanonical word order stage, our hypothesis is that the sentence in (19) will
appear after that in (18).
(17) ieri
cosa ha
yesterday what have-3.sG done
[yesterday what did Alberto do?]
(18) *i libri
dove compra
the books where buy-3.sG
[where does Pino buy the books?]
(19) i libri
compra Pino?
the books-MAsC.PL where them-MAsC.PL buy-3.sG Pino?
[where does Pino buy the books?]
In this type of sentence, not only FOC QP, but also both the TOPOBJ and sUBJ
core functions appear in noncanonical position. The progress now occurs not
only because learners manage, first, to identify the DF and GF of each nP, but
also because they manage, secondly, to mark OBJ morphologically onto V with
the clitic coreferential with TOP – cf. (5) above. At this same stage there is one
further step to go in the learning process. When V is inflected analytically,
Italian requires that the past participle agree in number and gender with the
clitic. This adds a further burden to the online production of the sentence. Our
hypothesis, then, is that there will be yet another step within the stage when
learners first produce ungrammatical sentences with a default masculine singular past participle, as in (20), and finally fully grammatical topicalised interrogative sentences, as in (21).
(20) i libri
dove li
the books-MAsC.PL where them-MAsC.PL have-3.sG
[where has Pino bought the books?]
bought-MAsC.sG Pino?
(21) i libri
dove li
the books-MAsC.PL where them-MAsC.PL have-3.sG
[where has Pino bought the books?]
bought-MAsC.PL Pino?
8. Acquiring constituent questions in Italian as a second language
4. The study
In order to verify the developmental hierarchy hypothesized in § 3, we analyse
cross-sectional data pertaining to 12 learners with different levels of competence in Italian. All learners are european students in their early twenties
attending Italian L2 courses at the University of Verona: five of them (Ve, Pe,
Jh el, Me) are Czech, two (Cr, La) are German, and one each among Lu, ni,
Ju, Ma, and ev is respectively Dutch, english, russian, spanish and French –
their names of course having all been changed. The study also includes one
native speaker control (Ga). Because most of the structures tested are optional, this inclusion allows for comparison between the learners’ and the native
speaker’s production in the same situations. All subjects were recorded in
March and April 2008.
The data elicitation tasks for this study are partly inspired by those used for
english questions (cf., e.g., Pienemann 1998: 280; Keatinge 2008), and partly
specifically devised for Italian, which, unlike english, is a null-sUBJ language.
Learners are prompted to produce interrogative sentences in order to gather
information on two different items or events (e.g., two mysterious objects, two
types of weddings). Because sUBJ position is of great relevance in the grammatical codification of Italian constituent questions (cf. §§ 2-3), the presence of two
items encourages the use of explicit sUBJ rather than the more common null
sUBJ. In order to make sure that learners do use it, they are not allowed to collect all the information they require to play the game first on one item and then
on the other, but must ask questions about them alternatively, thus specifying
an item each time. In our study, five tasks were specifically targeted to elicit constituent questions, other tasks served as distractors.
Out of the whole corpus thus collected, we analyse here constituent questions encoded in full sentences with lexical Vs whose a-structure maps canonically onto f-structure, such as dare (give) and comprare (buy). In other words,
we do not consider passives or sentences with unaccusative and so-called exceptional Vs (Pinker 1984), as well as copular and presentative sentences (for these
‘nonverbal predicates’, cf. Kroeger 2005: ch. 10). Furthermore, in this study we
consider only sentences in which the focal element bears an argument GF or the
nonargument GF ADJ. This means that we leave for future analysis those with
perché (‘why’) and come (‘how’), which may involve subordination (regarding
why in english L2, cf. § 3.1, ch. 2, this volume). Finally, formulas such as quanto costa? (‘how much is it?’) have been discounted. Altogether, we analyse 372
constituent questions, corresponding to an average of circa 29 for each of the
12 learners, compared to 33 by Ga, the native speaker of Italian.
Camilla Bettoni and Giorgia Ginelli
5. The analysis
The empirical support for the interlanguage development of Italian constituent
questions hypothesised in (6) is presented cross-sectionally in (22).
(22) Cross-sectional study of the development of Italian constituent questions based on the
Prominence Hypothesis: interrogatives
As shown in (22), in-situ questions belonging to the canonical word order stage are
few and produced by only some of the learners, as in (23)-(24), as well as by Ga,
the native speaker.
(23) Ve
quale musica?
prefer-2.sG which music?
[which music do you prefer?]
(24) La
all’università a che ora?
stays-3.sG at university at what time?
[until what time does he stay at university?]
All our learners have also reached the XP canonical word order stage in so far as
they produce 8 or more structures where QPs are clause-initial and fronted, followed by canonical word order. There is also ample evidence that all learners can
use a wide range of QWs and QPs – even Pe, for example, as shown in (25). As a
matter of fact, in this regard, he displays a range of forms that is already similar to
Ga’s, the native speaker.
8. Acquiring constituent questions in Italian as a second language
(25) Pe
a. dove
questi fiori?
where you
these flowers?
[where do you find these flowers?]
b. che cosa fa
in casa
a sera?
do-3.sG at home in the evening?
[what does (he) do at home in the evening?]
c. chi
questa borsa?
who has-3.sG this bag?
[who has this bag?]
d. quanto tempo tu
questi amiche amici?
how long
know-2.sG these friends?
[how long have you known these friends?]
e. qualche [quale] artista preferisce?
which artist
[which artist does (he) prefer?]
There is, however, significant variation among the 12 learners in the distribution
of the different structures belonging to this XP canonical word order stage, concerning two factors: (i) the use of sUBJ vs null sUBJ; and (ii) the addition of a topicalised element. First, the less advanced learners – in particular Pe, Lu and Ve – use
null sUBJ quite sparingly, and overuse instead pronominal sUBJ in pragmatically
unmotivated contexts, as exemplified in (25)-(26): Ve 13 times out of 20, Pe and
Jh 6 times out of 13 and 15 respectively, Lu 4 out of 8. In (25b), Ve even uses both
a pronominal and a referential sUBJ; needless to say, in this latter case preverbal
sUBJ is ungrammatical.
(25) a. Ve
quanto tempo tu
con Marco?
how much time you spend-2.sG with Marco?
[how much time do you spend with Marco?]
quale ora
lei Daniela va
at what time she Daniela go-3.sG sleep?
[at what time does Daniela go to sleep?]
(26) Pe
quanto tempo fa loro
how long ago
[how long ago have they met?]
As learners progress, they display a more frequent use of null sUBJ. so, Cr and
La already show some progress in two ways: their figures for null sUBJ are high-
Camilla Bettoni and Giorgia Ginelli
er (they both use null sUBJ 4 times out of 16), and their use of sUBJ, although
still ungrammatically preverbal, is more pragmatically justified, and referential
rather than pronominal, as in (27). Finally, and most proficiently, ni, el, Ma, ev,
and Me use null sUBJ exclusively, like Ga does. This results in grammatical sentences, as in (28)
(27) Cr
quando Daniela deve
iniziare lavorare?
Daniela must-3.sG start
[when must Daniela start to work?]
(28) el
quante domande
how many questions must-1.sG make?
[how many questions must (I) make?]
still within this stage, the second significant variation among the learners concerns
the addition of a topicalised element. Once learners are able to front FOC, the
addition of a topicalised constituent in the prominent clause-initial position would
not seem very costly, provided TOP bears the ADJ function. Yet only 2 of our 12
learners topicalise ADJ, and do so once each, compared to Ga who does it 3 times.
The grammaticality (or ungrammaticality) of the full sentence with this addition
is independent of fronted ADJ, and depends exclusively on what follows it: in our
corpus the sentences are null sUBJ, as in (29)-(30), so the outcome is targetlike,
even for Ve, who is the weakest learner.
(29) Ve
dopo lavorare che cosa fe [fa]?
after work
[after work what does (she) do?]
(30) el
adesso quale
in mano?
which have-2.sG in hand?
[now which one do you have in the hand?]
Although the use of a topicalised ADJ is rare among the learners, it is not altogether rare for some of them to preface their sentences with an external element.
We do not consider this element a DF TOP, however, because it is independent
of the structure that follows. In fact it is then repeated within the sentence, where
it bears a variety of GFs, such as sUBJ in (31)-(34), OBJ in (32), and ADJ in (33).
notice that the following sentence in (31) repeats sUBJ, whereas in (34) null
sUBJ follows.
(31) Lu
roberto a che ora
in la sera?
roberto at what time roberto
eat-3.sG in the evening?
[roberto, at what time does roberto eat in the evening?]
8. Acquiring constituent questions in Italian as a second language
(32 ni
i fiori
dato i fiori
a te?
the flowers who have-3.sG given the flowers to you?
[the flowers, who has given the flowers to you?]
(33) Me
sul lago
cosa hanno
fatto sul lago?
on the lake what have-3.PL done on the lake?
[on the lake, what have they done on the lake]
(34) Ve
roberto cosa fe [fa]?
roberto what do-3.sG?
[roberto, what does he do?]
Further up the developmental hierarchy, at the noncanonical word order stage,
learners acquire functional assignment, and can thus free up canonical word order
and place sUBJ postverbally. As (22) shows, all our learners have also reached this
stage – even Ve and Pe, who still have problems with the lexicon and verbal inflection, as exemplified in (35), where the latter first uses qualche instead of the targeted interrogative adjective quale, and then variously attempts preferi, preferà and
preferiscia instead of the targeted third person singular preferisce (‘(he) prefers’) of
the V preferire.
(35) Pe
a. qualche cibo preferi preferà Marco?
what food
[which food does Marco prefer?]
b. qualche musica preferiscia Marco?
what music
prefer-3.sG Marco?
[which music does Marco prefer?]
However, although all learners can place sUBJ postverbally with what appears
to be nativelike online processing, if we look at the distribution of sV and Vs,
we can see that some learners are more advanced than others. That is, as we
move from the left towards Ga on the right in (22), we can see that more beginner learners alternate postverbal with preverbal sUBJ; more advanced learners
do it less, in favour of postverbal sUBJ, thus becoming more accurate; until
they no longer do so, with ni, el, Ma, ev, and Me producing only grammatical sentences.
still within the noncanonical word order stage, like at the previous stage,
some questions are preceded by topicalized constituents. These can bear the
ADJ function as in (36), and do not imply further linguistic constraints in the
grammatical codification of the sentence; the outcome is thus grammatical.
(36) ni
Camilla Bettoni and Giorgia Ginelli
per pranzo cosa ha
fatto Maurizio ieri?
for lunch what have-3.sG done Maurizio yesterday?
[for lunch what has Maurizio done yesterday?]
On the other hand, several constraints are operative when the topicalized constituent bears the OBJ function. As we have seen in §§ 2-3, in order to codify
the sentence correctly learners must be able, first, to identify the GFs of all nPs
and, secondly, to unify the number and gender features of TOP with the clitic,
and the past participle if V is in analytical form. Among our 12 learners, three
have reached this highest level within the last stage in the developmental hierarchy. Their figures are not robust, but then also Ga produces only one OBJ
topicalisaton among his 30 questions. Coupled with the fact that this is a crosssectional study with only 12 learners, this means that our data does not allow
us to have a clear developmental picture of the full range of variation within this
last stage. First, all 6 occurrences of OBJ topicalisation are null-sUBJ sentences.
This is quite appropriate, but it can tell us nothing about the placing of sUBJ.
All we can say in this regard is that, if we look at the figures of Vs and sV
sequences for these three learners in all their production, we notice that they
always place sUB accurately in postverbal position. secondly, all three learners
always mark the TOP function with formally accurate clitics, althought Ma
seems somewhat hesitant when she produces i and le before getting li right, as
in (37). This means that we do not have enough variation in the data to prove
that the default masculine singular lo clitic might initially be overused.
(37) Ma i fiori
chi i le ti
to you them-MAsC.PL have-3.sG given-MAsC.sG?
flowers-MAsC.PL who
[the flowers who has given them to you?]
Thirdly, Ma and Me produce only one OBJ topicalisation each, both of them
with V in analytical form. Having no synthetic form to compare them with,
in order to prove our hypothesis that the number and gender agreement will
be marked last on the past participle, it is irrelevant to know that Me marks
the past participle correctly in (38), and Ma does not, using dato instead of
dati in (37).
(38) Me elisa
dove l’
elisa-FeM.sG where her-FeM.sG have-2.sG
[elisa where have (you) met her?]
The only learner providing evidence in support of this hypothesis of a step within
a stage is ev, who produces four OBJ topicalisations: two with V in synthetic form
8. Acquiring constituent questions in Italian as a second language
which are fully accurate, as in (39a), and two with V in analytical form which are
accurate except for the missing agreement on the past participle, as in (39b).
(39) a.
ev roberta
da quanto tempo la
roberta-FeM.sG since how long
her-FeM.sG know-2.sG?
[roberta for how long have you known her?]
i fiori
dove li
nella casa?
flowers-MAsC.PL where them-MAsC.PL have-3.sG put-MAsC.sG in the house?
[the flowers where has (he) put them in the house?]
6. Conclusion
In this exploratory study of the development of constituent questions in the Italian
interlanguage of L2 learners, our empirical cross-sectional data fully supports the
hypothesis presented in § 3. Despite the great care taken in devising suitable elicitation tasks, numbers of occurrences for some of the structures, especially at the
highest stage, are low, but this is not surprising, given both their infrequent occurrences also in native speakers’ production, and their wide range. This range
involves, on the one hand, the presence or absence of sUBJ, and its position when
present; and on the other, several different GFs associated with focusing question
constituents and topicalisations.
To summarise our findings, we have seen that all our learners are lexically
ready to produce constituent questions in so far as even the weakest among them
produce as wide a range of QWs and QPs as the native speaker does. syntactically,
all our learners can focus the question constituent by fronting it. The prime reason
for this is that fronted FOC is the Italian way of marking interrogatives, and hence
the most pervasive QP position in the input. This also explains why only a few
learners in our data appear to traverse the canonical word order stage by producing
in-situ questions. In this regard, we draw the reader’s attention to a similarity with
German L2 in chapter 9 regarding the quick way in which learners appear to disentangle canonical word order thanks to the strength of the FOC position in constituent questions. This may very well show that canonical word order is not as relevant in questions as in declaratives. Or indeed it may even turn out that declaratives and constituent questions have different canonical word orders.
Besides placing QPs in initial position, all our learners can also place sUBJ
postverbally. This means that all of them have reached the noncanonical stage, and
are therefore – stagewise – quite advanced. However, if we look at accuracy beyond
emergence, we notice fair differences among the learners in terms of the proportion of structures which are pragmatically justified and syntactically accurate, and
Camilla Bettoni and Giorgia Ginelli
those which are not. As we move along the developmental continuum, on the one
hand, pragmatically, learners gradually curb their tendency to overuse redundant
pronominal sUBJ until their null-sUBJ structures become as frequent as those
produced by the native speaker. On the other, syntactically, when referential sUBJs
are motivated by discourse or pragmatic reasons, learners gradually reduce their
preverbal ungrammatical sUBJs until they produce only targetlike postverbal
grammatical ones.
The addition of a topical element in front of the focal QP is an altogether
rare occurrence in both the learners’ and the native speaker’s data. When learners
do use it, no progress is discernible among them as long as the added topical element is associated with the ADJ GF and requires no further formal constraints.
On the other hand, when the topic is associated with the OBJ function and costly constraints are required, we have a clear indication that, among our 12 learners, only three of them have moved ahead of the others and produce at least one
OBJ topicalisation, although with some variation in the accuracy of morphological agreements.
We can conclude that the hypotheses we put forward in (22) are fully supported by the empirical data of this small cross-sectional study. They would have been
falsified if they had contradicted the implicational relationship among the predicted stages. But predictions have turned out positively, thus strengthening the case
for PT’s universal hierarchy presented in (35), chapter 1, this volume. As well as
generally confirming the hierarchy for the development of constituent questions in
a nonconfigurational language like Italian, this study shows specifically that in such
a language it may be easier to move from one stage to the next in terms of emergence of syntactic structures, than to progress in terms of accuracy by either pragmatically discerning when to use alternatives (i.e., null sUBJ vs full sUBJ; fronted
vs in-situ QPs) or morphologically increasing the range of inflections within a stage
(i.e., default vs nondefault number and gender markers).
Acquiring V2 in declarative sentences and
constituent questions in German as a second
Louise Jansen* and Bruno Di Biase**
*Australian National University, **University of Western Sydney
1. Introduction
German is a well-studied language within the PT framework and its predecessor, the Multidimensional Model (cf. Meisel, Clahsen & Pienemann 1981;
Clahsen, Meisel & Pienemann 1983; Pienemann 1981, 1984, 1989, 1998;
Clahsen 1984; Ellis 1989; Boss 1996, 2004; Jansen 2008; Baten 2013).
However, no study has yet focused on constituent questions specifically for
German, nor tested the Topic Hypothesis (Pienemann, Di Biase & Kawaguchi
2005) for this language or indeed its updated version, the Prominence
Hypothesis, proposed in chapter 1, this volume. Furthermore, no study has yet
conceptually and empirically compared the development of interrogative and
declarative sentences, as proposed in this volume, in order to assess whether this
generates any significant theoretical consequences. This chapter aims at filling
these gaps on the basis of data collected for a study of the development of
German syntax by Jansen (2008). Since the 2008 corpus includes content questions, we have the opportunity to investigate their development and compare it
with that of declaratives in those learners who do produce both. In what follows, we will first present some relevant principles of German syntax, and
hypothesize the developmental hierarchies for declaratives and interrogatives
according to the Prominence Hypothesis (§ 2). There follows a brief consideration of the literature on the acquisition of German word order (§ 3) reporting,
in particular, on results of two historically significant studies conducted within
the PT traditional background (i.e., ZISA/Clahsen, Meisel & Pienemann 1983;
Pienemann 1989). Then, in (§§ 4-5), the methodology of this study and the
results of the current investigation on content questions will be presented, along
with those for declaratives in Jansen’s (2008) study, as guided by two main
research questions:
Grammatical development in second languages, 259-274
Louise Jansen and Bruno Di Biase
• Do learners of German L2 follow the developmental schedule according to the
Prominence Hypothesis?
• Does noncanonical word order develop in both constituent questions and declaratives at the same time and in the same way?
Finally, in the discussion (§ 6), we will summarise and interpret the results of all
three studies (i.e., ZISA, Pienemann, and our own), and conclude that learners of
German L2 (a) indeed follow the PT schedule, (b) clearly begin their production
of noncanonical word order in constituent questions much earlier than in declaratives, and (c) appear to produce questions with noncanonical word order categorically at their onset, whereas with declaratives the path from emergence to acquisition seems much more gradual.
2. German syntax and the Prominence Hypothesis
There are two main peculiarities in German word order: German is a so-called
‘verb-second’ (V2) language, and its canonical word order differs according to
whether the lexical V is inflected or not. Let us look at these in turn.
First, in the main clauses of a V2 language the inflected V obligatorily occupies the second position in c-structure. A corollary of V2 is that XP SVX is
ungrammatical in German, as in (1), because this would force the V into third
(1) *heute
Hence, if any constituent bearing a function other than SUBJ (whether ADJ or
argument function) occupies the sentence-initial position usually associated with
SUBJ/TOP, SUBJ must follow the finite V, as in (2a-b).
(2) a. heute
spiele ich
[today I play tennis]
b. Tennis
spiele ich
[tennis I play today]
9. Acquiring V2 in declarative sentences and constituent questions in German as a second language
Second, as pointed out by Zobl (1986), canonical word order in German is SVO
in main clauses with lexical Vs only, whereas the lexical V is clause-final in main
clauses with AUX-V or MOD-V, as shown in (3a-b).
(3) a. ich
spiele Tennis
[I play tennis]
b. ich
[I have played tennis]
Thus, learners with a V-last L1, such as Turkish may first hypothesize that the
V-last they hear in the input is canonical for German (Zobl 1986) as found, for
instance, by Haberzettl (2005). We will come back to this point in (§ 3) below.
With regard to constituent question formation, German is a fronting language,1 which means that FOC, the discourse function associated with question
words or phrases (QP) is expressed, syntactically, in sentence-initial position
(Mycock 2007). Parallel to the DF TOP (for a brief outline of these categories,
cf. § 2.2, ch. 1, this volume) in declarative sentences, the DF FOC expressed in
the QP can be linked to any GF, whether ADJ as in (4a), or argument as in (4b).
(4) a. wann
spielst du
[when do you play tennis?]
b. was
spielst du
[what do you play today?]
Thus the V2 rule applies to both topicalised declaratives and constituent questions.
That is, SUBJ is postverbal whenever TOP or FOC, bearing any other GF, is in
sentence-initial position.
Let us now hypothesise how learners acquire the German V2 rule in declaratives and constituent questions – or, in other words (i.e., those used in this volume), how they learn to move beyond the German SVO/SOV canonical word
1 No distinction is made here between single-fronting and multiple-fronting languages
(cf. Mycock 2007).
Louise Jansen and Bruno Di Biase
order. Based on the universal progress outlined with the Prominence Hypothesis in
chapter 1, this volume, we illustrate the development of German declaratives and
constituent questions in (5).
(5) Developmental stages hypothesised for German L2 syntax based on the Prominence
Hypothesis: declaratives and constituent questions
For declaratives, we hypothesise that the German canonical word order, SVO or
SOV, will emerge first. Topicalization followed by canonical order will come about
next, followed finally by noncanonical word order. For constituent questions, the
developmental hypothesis is parallel to that of declaratives. That is, the single question word stage is followed by a stage where an SVO or SOV order is used with insitu QP FOC. The following stage will see a QP FOC in first position, followed
by canonical word order. This is not standard German, but L2 learners are hypothesized to go through this stage regardless of whether their L1 exhibits V2 or not (cf.
Pienemann & Håkansson 1999; Pienemann, Di Biase, Kawaguchi & Håkansson
2005). Finally noncanonicity will emerge.
9. Acquiring V2 in declarative sentences and constituent questions in German as a second language
3. Literature review
Before dealing with the development of questions we should briefly refer to the
development of German canonical word order in L2 learners. As alluded to above,
in her longitudinal study of four L2 German child learners, two of whom with
Turkish L1, Haberzettl (2005: 159) finds that in their early data samples finite Vs
are predominantly in final position; from this she concludes that her results for the
two Turkish learners do not support PT. On the other hand, data of an adult
Turkish learner from an earlier longitudinal study by Klein & Carroll (1992: 162172), re-analysed by Schwarz & Sprouse (1994), shows that inflected V forms
appear systematically in second position, and uninflected forms in final position.
Furthermore, in their longitudinal analysis of an Italian learner of German L2,
Klein & Carroll (1992: 131) observe “some degree of uncertainty”, with both SVO
and SOV occurring in the same utterances. How do all these findings affect our
proposed schedule in (5)? Contrary to Haberzettl’s interpretation, her data does not
contradict PT because, unlike past interpretations of SVO as the universal canonical order, PT now hypothesises language-specific canonical orders (cf. § 4.2.1, ch.
1, this volume), and for German we hypothese both the SOV and SVO orders.
Neither, of course, do any of the Haberzettel’s, Klein & Carroll’s and Schwarz &
Sprouse’s findings contradict PT’s Developmentally Moderated Transfer
Hypothesis (Pienemann, Di Biase, Kawaguchi & Håkannson, 2005) because
German canonical order includes both the L1 Turkish SOV and the Italian SVO.
With regard to the development of questions, to our knowledge there are no
studies focusing specifically on their acquisition in L2 German, although two historically significant studies do include relevant data. These are (a) the ZISA crosssectional study on the naturalistic acquisition of German by 45 learners from
Romance language backgrounds (Meisel, Clahsen & Pienemann 1981; Clahsen,
Meisel & Pienemann 1983), and (b) a longitudinal study on the instructed acquisition of German by three university learners from an Australian English background by Pienemann (1987, 1989, 1998). The explanatory frameworks applied
to these two studies are different from the one proposed here. Whereas this volume
uses LFG, which is a nonderivational approach, studies (a) and (b) assume a grammatical framework that is derivational (transformational), and hence group what
for us is the noncanonical word order of declaratives and questions, emerging at
the upper stage of the developmental path, under the category of ‘inversion’. Next
we summarise the main results of these studies.
In (6) below we present cross-sectional data for the 25 learners, among the 45
investigated in the ZISA project, who produce constituent questions. Our table is
constructed on the basis of two ZISA tables in order to facilitate the comparison
between questions and declaratives, and further reorganised in order to make it
more reader friendly in the context of this volume. In this connection, a couple of
Louise Jansen and Bruno Di Biase
points need to be raised. First, unlike all studies in this volume and Jansen’s presented below, the ZISA analysis includes sentences with copula (cf. examples (2)
and (7) in Clahsen, Meisel & Pienemann 1983: 128-129). Second, the ZISA study
only includes the ‘inversion’ data, expressed as ratios of the application of the
SUBJ-V inversion rule. So, ‘1’ means that ‘inversion’ is categorically applied (100%
of the times), 0.25 (or other fractions) means that the rule is applied 25% (or n%)
of the times; and ‘0’ that it is never applied (0%); an empty cell means no context
for the rule application is produced. Because only obligatory contexts were included originally, we assume that nonapplication of ‘inversion’ (marked as 0 in the
table) represents the production of structures of our XP canonical word order stage.
Note also that the ZISA tables distinguish whether learners produce five or more
contexts for ‘inversion’, or less than five. In the latter case the ratio is represented
in brackets in (6) below as in the original ZISA tables. Finally, we maintain the
original division in three columns and the original labels, but we add a row with
(6) Application of the SUBJ-V inversion rule (i.e., production of structures at the noncanonical word order stage) in questions (first column) and declaratives (last three columns) –
Reconstructed from Clahsen, Meisel & Pienemann (1983: 130-134, 145)
9. Acquiring V2 in declarative sentences and constituent questions in German as a second language
the labels used in this volume to clarify what corresponds to what, although we will
not labour on the significance of the different functions of TOP.
The results shown in (6) support the Prominence Hypothesis because five
learners use XP canonical word order exclusively (Antonio, Pascua, Angelina,
Eliseo and Rosemarie), three learners (Janni, Agostino and Pinto) use noncanonical word order categorically in both declaratives and questions, and the remaining
17 learners use noncanonical word order at least sometimes. With respect to
whether or not noncanonicity develops at the same time and in the same way in
questions and declaratives, the table shows that XP canonical order is either copresent or entirely absent in declaratives and questions in all learners with the
exception of four (Toni, Leornardo, Carmela and Lolita), who produce noncanonical word order in declaratives but not in questions. So, at least for these latter four
learners, the table marks the emergence of noncanonicity for declaratives before
questions. We note however that, after emergence, many more learners (11 of
them) show categorical use of noncanonicity with questions, whereas only 4 (Janni,
Agostino, Pintos, Maria S.) show categorical use with declaratives. We will take up
this point in our discussion below.
In (7) we present the results of Pienemann’s (1989; also 1987 and 1998) oneyear longitudinal study of three ab-initio Australian university learners of German,
which he conducted in order to compare their development to that of the naturalistic learners investigated in the ZISA cross-sectional study. As the grey area in the
table shows, however, only one of his three learners was recorded for the whole year.
Pienemann’s data has been treated here in the same way as ZISA’s have in (6).
(7) Application of the SUBJ-V inversion rule (i.e., production of structures at the noncanonical
word order stage) in questions and declaratives – ratio figures from Pienemann (1989: 68-70)
Louise Jansen and Bruno Di Biase
Also the results from Pienemann’s learners support the Prominence Hypothesis in
so far as the learners reach the XP canonical word order stage (marked with a 0 ratio
in the table) before or at the same time as the noncanonical word order stage
(Vivien is an exception to this but, significantly, only in questions). Furthermore,
noncanonicity emerges in questions long before declaratives in all three instructed
learners, unlike in ZISA’s cross-sectional study of naturalistic learners. In fact, all
declaratives are produced without ‘inversion’ by the three learners, with the only
exception of Steven, who produces one or two declaratives with ‘inversion’ in week
11 and then another few in the very last week, marked respectively with the ratios
of .5 and 1.2 Remarkably, also, noncanonicity is virtually categorical in questions,
as the exceptions are rare and concern only the readings for weeks 3 and 11 for
Steven and week 9 for Vivien.
4. The study
Our study on content questions in German L2 comprises a subset of the 16 learners who produce them out of the original 21 in Jansen’s (2008) study, which dealt
with declaratives only. All 16 learners are adult native speakers of English, ranging
in age from 18 to 49 years; 8 are female and 8 male, and they are all students
enrolled in a German course at an Australian tertiary institution. In terms of length
of instruction, 10 learners (the first from left to right in (8) below) would be comparable to Pienemann’s (1989) learners at week 13 because they had one semester
of instruction. Only two of these learners have been to a German speaking country for one or two weeks as tourists. As for the remaining 6 learners, one of them
(CR) had studied German for 3 semesters, and the other 5 (rightmost in the table)
had studied the language for 5 semesters, including topical courses, such as literature and culture studies, delivered in German. Exceptionally, LD had completed
a Science German course, had a more varied instructional background in German,
including private tuition, and at the time the sample was taken was teaching a
beginner German class in school. None of the learners had significant exposure to
naturalistic spoken input, such as during conversation in a bilingual GermanAustralian home, or instruction in an immersion setting. However, many indicated that they had occasional exposure to naturalistic input outside of class, such as
watching German films, reading in German or conversing with other nonnative
speakers of German.
2 The general lack of raw numbers in the original studies (both ZISA’s and Pienemann’s)
is rather unfortunate. In these two cases, however, we can desume the exact numbers of
these occurrences from the original study.
9. Acquiring V2 in declarative sentences and constituent questions in German as a second language
To elicit the data, learners were asked to meet and have a conversation with
a native speaker of German, whom they had never met before. The topic of the
conversation was “getting to know one another”. The declared purpose was to
“provide a speech sample for a research project on second language acquisition”.
It was emphasized, to both the learners and the native speakers, that although
the researcher was interested in a sample of the learner’s speech this should not
encroach on spontaneity, the conversation should not be like an interview, and
learners too were expected to ask questions. The conversations lasted about 45
minutes. They were audio recorded and transcribed for analysis.
In line with current PT analyses, and in contrast with the two studies considered in our literature review above (i.e., Clahsen, Meisel & Pienemann 1983 and
Pienemann 1989), only sentences containing a lexical V are included in this study
to the exclusion of copular sentences (e.g., es ist gut [‘it’s good’]) and presentatives
(e.g., es gibt nichts [‘there is nothing’]). Excluded are also sentences with expletive
SUBJ (e.g., es regnet [‘it rains’]) and those which are structurally ambiguous or otherwise insufficiently audible to be analysed; verbatim repetitions from the interlocutor, and typical (semi)formulaic questions frequently used in early classroom
communication (e.g., wie sagt man x? [‘how does one say x?’]). Repeated identical
sentences (i.e., those with the same lexicon as well as the same structure) are counted once. Thus the total number of sentences analysed here for the 16 learners is
1372. Among them, 85 are questions, and 1267 are declaratives. Among the latter
1059 display canonical word order, and 208 have a nonSUBJ element as TOP in
first position. Among the questions, only 2 have in-situ FOC, and the rest display
fronted FOC.
5. Results
Results of the analysis are illustrated in (8). The learners are referred to by a code
and ordered according to the range of structures they actually produce. The leftmost column shows the structures with FOC representing QPs, which may be
arguments or ADJs; TOP represents fronted elements other than SUBJ in declaratives, which also may be arguments or ADJs. The numbers in the cells represent
the frequency of the structure in the data. For instance, reading the learner YJ’s
declarative results from bottom to top, we see that she produces 62 structures with
canonical order, 4 XPTOP with canonical order and 3 XPTOP with noncanonical
order. As for questions, she produces two (in-situ) canonical word order questions,
and three XPFOC with noncanonical order.
As (8) clearly shows, all learners produce canonical SVO/SOV sentences, and
do so in greater numbers than all the other types of structures. Results for declaratives show that when learners add a topicalised element to their sentences all of
Louise Jansen and Bruno Di Biase
(8) Cross-sectional study of the learners’ development of German: declaratives and constituent
questions (expanded from Jansen 2008)
them produce ungrammatical fronted XP with canonical word order. Canonical
order is always produced as SVO, as in (9a-b), except when SR produces also SOV
once, shown in (10). However, 9 of the 16 learners produce also targetlike V2
structures in declaratives, as in (11a-b), and have thus reached the noncanonical
word order stage.
(9) a. RA
b. CR
(10) SR
(11) a. ST
wenn wir sind in Deutschland wir
findet ein family
when we are in Germany
finds a family
[when we are in Germany we will find a family]
viele Freunden ich
habe in Adelaide
many friends
have in Adelaide
[I have many friends in Adelaide]
in Deutschland gehen
in Germany
[then I go to Germany]
in Europa lernen viele Leute viele Sprachen
in Europe learn
many people many languages
[In Europe many people learn many languages]
9. Acquiring V2 in declarative sentences and constituent questions in German as a second language
b. CR
nach England gegegen
to England
[in 1972 I went to England]
As can be appreciated from these examples, the range of structural exponents of
TOP is wide, including: subordinate clauses as in (9a), NPs as in (9b), adverbs as
in (10), and PPs as in (11a).
Results with questions differ significantly, and point to a strong role for FOC
in bringing about noncanonical word order. First, only one learner (YJ) uses in situ
FOC, as in (12), which belongs to the canonical word order stage. Interestingly,
this learner also produces a FOC preceded by ADJ, as shown in (14), the only
structure of this type in the whole data set.3
(12) YJ
arbeitest da
seit wieviele Jahren?
you work
there since how many years?
[you have been working there for how many years?]
Second, whereas all 16 learners produce ungrammatical declaratives with TOP
SVO structures, none of them produce questions with ungrammatical FOC SVO
structures. Thus not only have they all reached the higher stage, but their word
order is also applied categorically, including when they produce marginally acceptable structures, as MR in (13), as well as complex structures, as in (14), where FOC
is preceded by TOPADJ.
(13) MR wie lange hast Sie
gelibt ins Australisch
how long have you lived in Australian
[how long have you been living in Australia?]
(14) YJ
willst du
after that what
[after that what do you want to do?]
3 This complex structure, together with its accurate morphology, points to a more mature
use of in-situ FOC questions, which is not unknown in native speakers speech (cf.
Bettoni & Ginelli’s comments with regard to Italian L2 in-situ questions, ch. 8, this
book). Cf. also this same learner’s complex structure in (14).
Louise Jansen and Bruno Di Biase
As with declaratives, also with interrogatives the range of structural exponents of
the fronted element is wide: question phrases in (12)-(13), as well as question
words in (14), may function as ADJ or argument GFs.
Admittedly, our figures are not always robust, as some of the learners (e.g., CP
and RA) produce only one question with noncanonical word order. However,
questions such as those in (15a-b) appear to be constructed online (i.e., nonformulaically) because a subsequent check confirmed that their Vs (i.e., wohnen (‘live’)
and gehen (‘go’) respectively) are used also in other structural contexts.
(15) a. CP
b. RA
wohnst du ?
[where do you live?]
gehen wir ?
[where do we go? (he is asking himself aloud)]
6. Discussion
To recap, our cross-sectional study of 16 learners investigates the development
of content questions in German L2 and compares the results both internally,
with the development of declaratives, and externally, with relevant studies in
German L2 such as ZISA’s and Pienemann’s. These two studies are different
from ours not only because of their method of counting and presenting frequencies as ratios instead of raw figures, but also because they include copulas
and presentatives. Significantly, however, a methodological approach proposed
in this volume, but not yet applied empirically elsewhere (cf., however, the theoretical treatment of developmental syntax in English L2, § 3.1, ch. 2, this volume), is justified in the current study. We are referring to Bettoni & Di Biase’s
requirement for analytically sorting out questions from declaratives (cf. §§ 4.2.1
and 4.3, ch. 1, this volume) in order to clarify possible effects of different sentence types, such questions and declaratives, on timing of emergence and developmental patterns.
Our first research question asked whether or not the Prominence Hypothesis
is supported for German L2. Results in all three studies, despite their differences in
time, space and linguistic environment, are indeed compatible with the developmental schedule for German L2 based on the Prominence Hypothesis which we
formulated comprehensively for both declaratives and questions in (5) above. In all
9. Acquiring V2 in declarative sentences and constituent questions in German as a second language
cases where noncanonical word order emerges, it does so either at the same time,
or after canonical word order.
Our second research question asked whether noncanonicity develops at the
same time and in the same way in declaratives and in questions. The answer
appears to be negative for both variables. In terms of time of emergence, noncanonicity in questions emerges long before than in declaratives. In Pienemann’s
longitudinal study, in one of the learners (Steven), noncanonical word order
(‘inversion’, in his terminology) emerges in questions in week 3, and in declaratives
in week 11. In the other two learners, it emerges in questions in week 1 (Vivian)
and week 7 (Guy), and does not emerge at all in declaratives. This earlier emergence in questions is confirmed in the current cross-sectional study, where all 16
learners use noncanonicity in questions, but seven of them fail to do so in declaratives. In contrast, in four out of the 25 ZISA learners noncanonicity emerges in
declaratives before it does in questions, thus providing some counter-evidence to
the contrast between questions and declaratives in terms of timing of emergence.
In sum, noncanonicity in questions precedes noncanonicity in declaratives (with
the four ZISA exceptions).
In terms of developmental patterns, noncanonicity in questions appears to
be categorical, whereas in declaratives there is a persistent survival of XP with
canonical word order even in those learners who do produce noncanonical word
order comfortably. Thus, the difference between questions and declaratives is
actually dramatic. Not only do all the 16 learners in our study produce postverbal SUBJ with fronted FOC in questions, but they do so categorically. This
finds strong similarities in Pienemann’s study, where the number of contexts is
generally robust, particularly so in Guy’s figures. The number of contexts in the
ZISA study also provides some support for noncanonical word order in questions to emerge in a categorical fashion, certainly more readily in questions than
in declaratives. The laborious progress of ‘inversion’ in the presence of fronted
elements had also been noted by Clahsen (1984) in his longitudinal study of
three ZISA learners, who observed that it is a difficult rule to learn. What has
not been noticed so far, however, is that such slow and noncategorical progress
may be confined to declaratives.
What remains to be explained, then, is why categorical noncanonicity in
word order emerges earlier in questions, and lags so far behind in declaratives.
What is specific to questions that can account for this? A couple of reasons may
be put forward, at least for German L2. An often invoked possible reason is L1
transfer. The informants in the present study and in Pienemann’s are from an
English-speaking background, and English requires noncanonical word order in
questions but not in declaratives. However, English requires do-support for
questions unless the lexical V is already supported by an AUX or MOD V. In
Louise Jansen and Bruno Di Biase
German, on the other hand, such do-support is unknown and the lexical V is in
sentence final position in the presence of AUX or MOD V. So, in spite of some
similarities, the differences seem to be quite significant, and transfer, if any,
would require extra computations in working memory. Testing the transfer
hypothesis would in any case require triangulation with at least one other language with no postverbal SUBJ in constituent questions (cf. the DMLT
Hypothesis in Pienemann, Di Biase, Kawaguchi & Håkansson 2005). A second
reason may be that unlike TOP in declaratives, FOC in questions is obligatory
(Mycock 2007), and therefore reliable in the input and close to one-to-one
form-function mapping (cf. Andersen 1993; Ellis & Collins 2009, among others). Moreover, and most significantly, FOC in constituent questions is lexicalised transparently through a small, closed set of items, whose lexical specification includes focus, namely the wh- question words (Horvath 1986: 188). By
contrast, TOP can be lexicalised in a virtually infinite number of ways.
Therefore, the input for TOP is not only less frequent but also highly variable.
Note also that in the very frequent default declaratives TOP usually coincides
with SUBJ, which occupies the first slot in c-structure. This contrasts with the
obligatory postverbal position of SUBJ in constituent questions (the only exception being when SUBJ itself is questioned). Postverbal SUBJ may thus be computed as the default position for SUBJ in content questions.
7. Conclusion
The results presented in this chapter, including those of the historical studies,
support the Prominence Hypothesis. They document a robust continuity in
SLA studies, but they also show that further analytical work based on sound
theoretical principles can bring out hitherto unsuspected insights. Thus the
separation of declaratives from questions enables the identification of significant
differences in the acquisition patterns of noncanonical word orders. These differences are, on the balance of evidence, first that noncanonical orders emerge
in questions before they do in declaratives, as confirmed in Pienemann’s (1989)
longitudinal study, and second that noncanonical word order in questions in
German L2 is appropriated categorically by Jansen’s (2008) learners, nearly so
by Pienemann’s, and tendentially also by ZISA’s (1983). Hence, acquisition patterns are different: predominantly categorical in questions, and predominantly
gradual in declaratives, where the canonical word order of the lower stages persists alongside the target-like noncanonical one of the higher stage even in significantly advanced learners.
A limitation on this conclusion is that the data for categoriality in questions is
small in some learners given the naturalistic nature of the data collection method.
9. Acquiring V2 in declarative sentences and constituent questions in German as a second language
More robust data collected with focused tasks would better test these early results.
If this chapter offers some possible explanations for the differences in the acquisition of noncanonical word order in questions versus declaratives, further research
may reveal whether different GFs (argument or nonargument functions) and different elements (words, phrases or subordinate clauses) linked to FOC or TOP do
or do not play a role in such remarkable differences, and may help establish a more
unified and theoretically grounded explanation.
Thanks are due to Camilla Bettoni for her help in discussing and organising the issues
dealt with in this chapter. We also wish to thank Wayan Arka for his comments on an
earlier version of the chapter, as well as Gabriele Pallotti and two anonymous reviewers
for their insightful comments.
Exploring Processability Theory-based hypotheses
in the second language acquisition of a child
with autism spectrum disorder
Tonya G. Agostini* and Catherine T. Best**
*University of Western Sydney, **University of Western Sydney and
Haskins Laboratories
1. Introduction
Delays and deficits in structural language and social communication are common features of autism spectrum disorder (ASD), but the exact nature of these
problems is unclear. In particular, little is known about grammatical development in children with high functioning autism spectrum disorder (HFASD).
An understanding of how language develops in this population may provide
valuable insight into how underlying processing difficulties contribute to speech
delays. The aim of this study is to explore whether a child with HFASD is able
to acquire a second language. This study adopts Pienemann’s (1998) classic
Processability Theory (PT) approach in order to predict and measure how early
inflectional morphology develops in this child when acquiring Italian L2, compared with typically developing children.
1.1. High functioning autism spectrum disorder
Autism spectrum disorder (ASD) is a lifelong neurodevelopmental disorder first
described by Kanner (1943). The criteria for diagnosing ASD are based on impairments in two domains: (i) social communication and interactions, and (ii) restricted and repetitive behaviours and interests, often obsessive in nature (APA 2013).
Language outcomes vary greatly (Kim, Paul, Tager-Flusberg & Lord 2014), but
functional use of language prior to age 5-6 years is predictive of increased communication and independent living skills in adulthood (e.g., Lord & Venter 1992;
Howlin, Goode, Hutton & Rutter 2004). Some distinct language features in children with ASD who acquire spoken language are: a prolonged period of echolalia,
pronoun reversals, repetitive play with words, significantly better expressive language skills than receptive ones (which is the opposite of typically developing chil-
Grammatical development in second languages, 275-290
Tonya G. Agostini and Catherine T. Best
dren), difficulties in perceiving and attending to speech, and an unusual and idiosyncratic use of language (cf. Boucher 2012; Kim et al. 2014 for overviews).
Pragmatic language deficits are universal in ASD and lead to difficulties in maintaining conversations. Social interaction skills are thus impaired.
High Functioning Autism Spectrum Disorder1 (HFASD) is a label used by
researchers and clinicians to refer to children diagnosed with ASD but without
intellectual impairment, that is, with IQ above 70 (e.g., Mesibov, Shea & Adams
2001). Even among verbal children with HFASD there is considerable heterogeneity in language skills in the areas of semantics, morphology, syntax and phonology.
At least two language-based groups have been identified: those with normal, or
age-appropriate, linguistic skills (HFASD-N), as measured on standardised tests,
and those with language impairments (HFASD-I), some of which are similar to
those found in Specific Language Impairment (SLI: Kjelgaard & Tager-Flusberg
2001; Tager-Flusberg & Joseph 2003; Tek, Mesite, Fein & Naigles 2014). The participant in this study falls into the first HFASD-N group, with age-appropriate linguistic skills at the time of the study, despite experiencing language delays during
the preschool years.
1.2. Grammatical development in autism spectrum disorder
Longitudinal studies during early stages of L1 acquisition could provide valuable
insights into the underlying linguistic mechanisms that lead to delays and deficits
experienced in ASD (Tager-Flusberg 2004). However, relatively few ASD studies
have investigated grammatical development in either L1 (Boucher 2012; Kim et
al. 2014) or L2 acquisition (Ohashi et al. 2012). This is due partly to the fact that
a diagnosis occurs between 2 and 4 years of age (Coonrod & Stone 2005; Filipek
1 The term HFASD here replaces High Functioning Autism (HFA) to reflect the new
DSM-5 (APA 2013) diagnostic terminology. HFA previously referred to a child diagnosed with Autistic Disorder, who has no intellectual impairment, IQ>70, as categorised
under the criteria of the Diagnostic and Statistical Manual of Mental Disorders –
Fourth edition (DSM-IV) (APA 1994). Importantly, a diagnosis of autistic disorder
implied onset prior to 3 years of age with initial language delays. This was in contrast to
other high functioning individuals, for example, those with Asperger Syndrome which
impliesd an onset later than 3 years of age and no early language delays. However,
because the DSM-5 (APA 2013) has since replaced its five distinct pervasive developmental disorders (i.e., autistic disorder, Asperger Syndrome, Rett’s Disorder, child disintegrative disorder, pervasive developmental disorder not otherwise specified (PDDNOS)) with one overarching category of Autism Spectrum Disorder (Volkmar, Reichow,
Westphal & Mandell 2014) we will use HFASD, and then specify the participant’s language history and status at the time of the study.
10. Exploring PT-based hypotheses in the L2 acquisition of a child with autism spectrum disorder
et al. 1999; Tager-Flusberg 2005), and partly to the sparsity of longitudinal studies since the introduction of robust and well-accepted diagnostic criteria.2
A recent study reports that children with ASD produce syntactically less complex utterances than typically developing children matched on non-verbal IQ,
despite a similar vocabulary size (Eigsti, Bennetto & Dadlani 2007). With regard
to early morphological development, studies on English-speaking children with
ASD report conflicting results (Tager-Flusberg, Paul & Lord 2005). Some claim
morphological development is deviant (e.g., Bartolucci, Pierce & Streiner 1980;
Howlin 1984), others that it develops the same way as in typically developing children (e.g., Jarrold, Boucher & Russell 1997; Tager-Flusberg & Calkins 1990;
Waterhouse & Fein 1982; Tek, Mesite, Fein & Naigles 2014). This controversy has
partly been resolved with the identification of at least the two distinct groups
among verbal children with ASD mentioned above, that is, those with apparently
normal age-appropriate linguistic ability, and those with language impairments in
phonological processing and grammatical morphology, some of which are similar
to those found in children with SLI (Kjelgaard & Tager-Flusberg 2001; Roberts,
Rice & Tager-Flusberg 2004; Tager-Flusberg & Joseph 2003; Tek et al. 2014).
Most recently, Tek et al. (2014) investigated grammatical development longitudinally in eight participants with a diagnosis of ASD and later categorised as
highly verbal (ASD-HV) (age range = 2;2 to 3;1) in comparison to a group of 18
typically developing children (age range = 2;3 to 2;8), over a period of 12 months.
They tracked increases of productive use of several aspects of basic grammatical
abilities and found that the ASD-HV and TD groups both increased productive
use of most of the same grammatical structures and at a similar pace. However,
authors later noted that the ASD-HV group had been receiving an average of 14
hours per week of Applied Behaviour Analysis (ABA) therapy, which directly targets communication and speech production (Tek et al. 2014: 83). This greatly limits any conclusions that could be inferred from the results, and furthermore highlights the difficulties of obtaining true natural spontaneous L1 speech data for very
young children diagnosed with ASD.
Researchers have started to monitor speech development of potentially
high-risk (HR) infant siblings of older children with ASD. Because some of
them may eventually receive an ASD diagnosis, this would potentially allow retrospective reconstruction of their language development (Tager-Flusberg 2005;
Zwaigenbaum et al. 2005; Hudry et al. 2014). However, this approach is high-
2 That is, the diagnostic criteria provided in the DSM-IV (APA 1994) and ICD-10
(WHO 1993) and, more recently, the changes in diagnostic criteria and terminology
published in the DSM-5 (APA 2013). The latter are hoped to be more sensitive to ASD
behavioural symptoms in infants and toddlers under 3 years of age.
Tonya G. Agostini and Catherine T. Best
ly resource intensive and has a relatively low likelihood of payoff, as the recurrence of ASD within the same family is estimated at 18.7% (Ozonoff et al.
2011). In these studies, to date, researchers have only administered broad language measures at periodic time intervals of 6 months or greater (cf. Hudry et
al. 2014 for an overview). In one such longitudinal study, authors report that
even with a large number of HR infant siblings involved, they had modest sample sizes because of the necessary separation of the larger HR infant siblings
group into smaller diagnosis-outcome-based subgroups, that is, those later diagnosed with ASD, those with atypical development but not ASD, and those with
typical development (Hudry et al. 2014). In this study we take the novel alternative approach of examining the beginning acquisition of an L2 in a child with
Studies of L2 and bilingual language acquisition in ASD are very few. In the
absence of empirical studies, Jellinek, Toppelberg, Snow & Tager-Flusberg (1999)
suggest that children with ASD should avoid L2 learning and bilingual education,
reasoning that their pragmatic language deficits and difficulty in initiating and
maintaining conversation will make them poor L2 learners. Yet there is evidence
that at least some children with ASD can learn an L2 (e.g., Hambly & Fombonne
2012, 2014; Kanner 1971; Kay-Raining Bird, Lamond & Holden 2012; Ohashi
et al. 2012; Petersen, Marinova-Todd & Mirenda 2012; Seung, Siddiqi & Elder
2006; Valicenti-McDermott et al. 2012). However, these L2 studies in ASD have
been largely observational and/or exploratory and report a wide variety of outcomes (cf. Hambly & Fombonne 2014 for an overview). None of them report on
grammatical development. Most useful would be information about how children
manage the earliest building blocks of L2 acquisition. In this study we look at the
development of productive use of L2 inflectional morphology in NPs. Mastering
inflectional morphology requires the child to grasp that alternating a specific
phonological subunit changes the meaning of a word.
1.3. The targeted structures
Italian L2 was chosen for this study because of its morphological richness and morphophonological suffix alternation patterns at NP level affecting Ns and their
modifiers, such as adjectives (e.g., gatt-o ner-o, ‘black.M.SG CAT.M.SG’). As a stembased language, its nouns and modifiers take an obligatory suffix that indicates
number and gender through a matrix of final vowel alternations, as represented in
(1). For the analysis that will follow in § 3, it is worth noting that two of the three
main classes of nouns and adjectives have –i in their plural form. For this reason,
–i is considered the default plural ending, and is usually acquired before the more
marked plural –e ending (Di Biase 2002; cf. also ch. 3, § 3.1, this volume). Unlike
English, which adds the morpheme –s to noun in order to indicate plurality, Italian
10. Exploring PT-based hypotheses in the L2 acquisition of a child with autism spectrum disorder
number and gender features are fused into one vowel at the end of the noun and
adjective. Furthermore, within a NP, the same values of both these features of the
head N must be expressed on all nominal modifiers, such as determiners, demonstratives and adjectives. In other words, for our purposes here, learners must learn
to compute the agreement within a NP, and mark adjectives in attributive position
with the number and gender values of the head N. In this study we ignored agreement of the article, which according to Di Biase (2002) is not convincing evidence
of the activation of phrasal procedure, for two different reasons: on the one hand,
because of its frequency it is often formulaically learned with the N; on the other
hand, because of the complexity of its form-function mapping, its completely accurate use is usually achieved very late in both L1 and L2 acquisition (Caselli,
Leonard, Volterra & Compagnoli 1993).
(1) Ending alternation in the main classes of Italian nouns and adjectives (after Vincent
2011, examples added)
PT provides the theoretical framework (Pienemann 1998; ch. 1, this volume) for
our examination of the acquisition of Italian number/gender morphology by our
participant. This theory serves our purpose well for two main reasons. Firstly, it is
psychologically plausible and formally testable (cf. ch. 1, this volume). Secondly, it
has already been applied to Italian L2 (cf. ch. 3, § 3), and tested for both adults (Di
Biase & Kawaguchi 2002) and typically developing children (Di Biase 2002),
allowing a straightforward basis of comparison for our participant’s pattern of morphophonological acquisition. PT proposes a universal hierarchy of specific procedural skills which allow for the development of L2 morphology according to a predictable and implicational order (cf. ch. 1, § 4.2.1, this volume). In this study we
consider the first three procedures: the lemma access procedure, yielding single
words and formulas used in an unanalysed way; the category procedure, yielding
lexical form variation; and the phrasal procedure, computing agreement between
Tonya G. Agostini and Catherine T. Best
nouns and adjectives, as they apply to Italian L2 (cf. the developmental stages based
on these procedures illustrated for Italian in § 3.1, ch. 3, this volume, and partly
repeated here in (2)).
(2) First developmental stages hypothesised for Italian L2 inflectional morphology in noun
phrases (after Di Biase & Kawaguchi 2002)
1.4. Research questions
The broad objective of this study was to investigate whether a child with HFASDN is able to learn Italian L2 grammar beyond the production of single words and
formulaic expressions. Specifically, for moving from the lemma access stage to the
first grammatical stage of form variation, we address two interrelated aspects of
morphophonological alternation in the L2 that do not occur in the child’s L1
(English). Development in Italian L2 requires that an English-speaking child learn
two grammatical principles: the final vowel of an Italian N must be changed in
order to mark plural number, and the specific vowel alternation used depends on
the class of the N. This entails two novelties for an English-speaking child: first,
number is marked by changing a final vowel, rather than by adding a consonantal
suffix as in English; and secondly, Ns divide into classes that require different vowel
alternations, unlike English, where purely phonotactic principles guide pronunciation of the final /–s/ for number-marking. Then, in order to progress to the
phrasal procedure stage, gender and number values must be not only computed
separately on head Ns and adjectives, but also unified between them. Because of
the different N and adjective classes, our learner must be able to produce such NPs
as gatto nero / gatti neri (‘black cat / black cats’) and gallina nera / galline nere (‘black
chicken / black chickens’). In this study the stimuli avoided N and adjective agreements with a combination of different vowel suffixes, such as gatto verde (‘green
cat’), gallina verde (‘green chicken’) for singular contexts, and galline verdi (‘green
chickens’) for plural ones.
10. Exploring PT-based hypotheses in the L2 acquisition of a child with autism spectrum disorder
It is already known that early language development is delayed in the majority of children with ASD. Hence we address two possible sources of that delay:
first, the social and communication deficits associated with ASD may decrease
language input and interaction; secondly, language delays in ASD may be caused
by underlying language processing difficulties/deficits. In the former case, if the
primary cause of language delay is social, then a child with HFASD-N with no
apparent cognitive and linguistic deficits should be able to acquire the means to
mark the plural on Ns and compute the agreement within NP when taught
under instructional conditions suited to the child’s social and attentional limitations (cf. below). In the latter case, that is, if language delays are caused by language processing difficulties/deficits, we predict that, even under suitable
instructional conditions, a child with HFASD-N learning Italian L2 should still
demonstrate difficulties with acquiring the new morphophonological and grammatical principles of Italian. This would entail that our participant is slower than,
or different from the typical children studied by Di Biase (2002), and that, in
particular, the child with HFASD-N will first overextend the default singular
forms to plural contexts, and then use the more common default vowel –i for
marking plural Ns and adjectives for a longer period, and be delayed in learning
to apply the rarer non-default –e ending.
2. The Study
2.1. The participant
The child participant, Chris (a pseudonym), is a six-year old boy from an
Australian English L1 background. His parents became concerned about his language and social development prior to age 2;0 when he seemed to stop learning
new words, had difficulty with social interactions, and frequently exhibited high
levels of distress. Chris also failed to use gesture to communicate, displayed very little eye contact, and seemed to use language out of context, in an echolalic manner.
At times he appeared deaf, not responding to his mother’s voice nor turning his
head towards the speaker when his name was called.
Chris was first diagnosed with ASD at age 2;4 by a developmental paediatrician, in accordance with DSM-IV criteria set (APA 1994). He immediately
began weekly sessions of speech and language intervention. By age 5;1, his overall level of cognitive and language development and functioning was within the
normal range, according to results in the Brigance (1991) K-1 screen, which
assesses receptive language, expressive language and cognitive skills. As a result,
Chris’ diagnosis was reclassified to HFASD: his grammatical development was
appropriate to his age, but his conversational skills were pragmatically impaired.
Tonya G. Agostini and Catherine T. Best
At the time of the study, Chris was aged 6;5 and his CELF-IV (Semel, Wiig &
Secord 2003) results placed his speech production within normal range.
However, he still presented with some receptive language difficulties relating to
comprehension of pronouns.
Chris’ paternal grandfather is a native Italian speaker, and both parents
studied Italian L2 at university level. From age 3;0, the child had received some
limited exposure to Italian L2 by means of Italian stories and cartoons.
Moreover, he had attended an Italian story-time group for preschoolers. At the
start of the study, he was thus able to count to 20, and label many animals, food
items and colours in Italian, but he had acquired no grammar and was at the
lemma access stage.
2.2. The data
To investigate the participant’s grammatical development in Italian L2, a shortterm longitudinal study was carried out. Over a 12-week period, Chris received 9
lessons of developmentally planned and communicative style teaching of Italian
(cf. Di Biase 2002). Lessons lasted one hour each and were provided in a one-toone teaching environment by an Italian native speaker. Data was collected five
times, including a pre-test and post-test, then transcribed by the researcher, and
checked against the video tape by an Italian native speaker. Rather than analysing
Chris’ linguistic production during the lessons, the acquisition of the targeted
structures was monitored by means of specific tests. This is because, during the lessons, contexts contrasting the targeted number and gender forms for the same referent occurred rarely, making it difficult to determine whether emergence of these
structures had occurred. Furthermore, during instruction sessions, it is difficult to
determine how much the teacher’s corrective feedback affects the child’s performance. On the other hand, the tests were set at the very beginning of the lesson, and
the teacher was instructed to refrain from providing any corrective feedback during the testing tasks. The five tests make up the corpus for this study. In addition
to the pre-test in week 1 (t1) and delayed post-test in week 18 (t5), the other three
tests occurred in lesson 4 /week 4 (t2), lesson 7 /week 7 (t3), and lesson 9/week 12
(t4). Each test lasted 5 to10 minutes.
All five data collection sessions used picture-naming tasks. The stimuli were
wordless picture cards, designed to elicit contrasting singular and plural Ns and
adjectival NPs, and the child was asked to name the pictures in Italian, including
their colour. An example of the stimuli is provided in (3). Most vocabulary items
used for the tasks were familiar to the informant, but new items were also included to assess the child’s skills in generalising number and gender feature markers.
Other strategies designed to minimise formulaic responses (i.e., language ‘chunks’
or vocabulary items rote learnt in their plural or singular forms) included variation
10. Exploring PT-based hypotheses in the L2 acquisition of a child with autism spectrum disorder
of presentation (e.g., photographic stickers were used in one session, and handdrawn pictures in the next); shifts in language domains (e.g., food items in one session, and farm animals in the next); change in the quantities of repeated plural
probes (e.g., 47 penguins in one session, and 2 penguins in the next). The changes
in the stimuli for each testing session helped ensure the learner was not producing
unanalysed language chunks or routines he associated with specific stimuli, an
important control to assure linguistic rather than associative rote knowledge, especially when testing children with ASD.
(3) An example of the stimuli used for the picture-naming tasks
In order to provide an indication of the child’s progress over the period of the study,
in (4) we present the numbers of Italian word types produced by Chris for each
testing time, and the mean length of turns (MLT), that is, the average number of
words per utterance he produced. When calculating MLT, the number of Italian
tokens in a turn excludes: English or ambiguous words (e.g., okay or no, which can
be English or Italian), repetitions, incomplete-, unclear- or echoic-items, fillers and
hesitations. Also excluded are counting routines up to the final number followed
by a N (e.g., in un due tre giraffe, only tre giraffe is considered). The number of turns
refers to any turn in which Chris supplied Italian tokens. When more than one
probe response was produced in a turn, only the longest response was counted for
that turn. As the table shows, both the vocabulary size and MLT increased consistently from one session to the next. Cumulative vocabulary increased more than
sixfold (from 24 word types in t1 to 165 in t5), and also MLT scores show good
progress, almost doubling from 1.4 at t1 to 2.3 by t5.
(4) Italian language production during testing times
Tonya G. Agostini and Catherine T. Best
3. Results
In accordance with PT’s developmental and non-normative perspective, all occurrences of plural marking in the learner’s system are counted, irrespective of accuracy. So, for instance, if the learner supplies a default plural –i suffix to mark a plural context, this counts as plural marking even if the native Italian system would
use a different suffix for that context. Following Pienemann (1998: 132-140), the
distributional analysis presented in (5) places plural –i and –e markings on Ns in
obligatory contexts at the category procedure stage, and number and gender agreement between N and adjective(s) at the phrasal procedure stage.
(5) Chris’s progress from the pre-test to the (t1) delayed post-test (t5)
At t1, the pre-test provided the learner with the opportunity to produce 9 nominal forms in the plural, 6 of them with an –i suffix and 3 with an –e suffix. No contexts were provided for adjectival forms on their own or NPs requiring agreement.
In the –i context, Chris produced two target-like forms. These however cannot be
taken as sufficient evidence for productive use of number variation at the category
procedure stage, as specified in (45), § 5, chapter 1 of this volume. This is because,
although cagnolini (‘little dog’), shown in (6), is correct, pesci (‘fish’) is used also in
a singular context, so there is an instance of oversuppliance, as shown in (7).
(6) t1 Chris
uno due tre quattro cinque sei *cagnollino
[one two three four five six little dog-MASC.SG]
Teacher ah [filler]
cagnollini sei cagnollini
[little dog-MASC.PL six little dog-MASC.PL]
10. Exploring PT-based hypotheses in the L2 acquisition of a child with autism spectrum disorder
(7) t1 Chris
*pesci (in response to a picture of one fish)
Teacher guarda . quanti? uno due tre quattro . quattro?
[look . how many? one two three four . four?]
In the remaining four plural –i contexts, the default singular forms were used: three
times with an –o N (cinque *gattino, ‘five cat’; due *cavallo, ‘two horse’; due *elefanto, ‘two elephant’), and once with an –e N (*tigre, ‘tigre’). Likewise, in all three plural –e contexts, Chris invariantly produced the singular suffix –a in both plural and
singular contexts (e.g., un giraffa vs. tre *giraffa, ‘one giraffe’ vs. ‘three giraffe’). We
can conclude that at the beginning of the instructional period, Chris was still at the
lemma access stage.
At t2, after three weeks of Italian instruction, Chris showed remarkable
progress. The test provided more varied contexts for producing plurals: 14 singleN contexts, 12 of them for –i, and two for –e; one single-adjective context in –i;
and two phrasal contexts with both items ending in –i. All single Ns but one were
correctly inflected for plural, whether the singular vs. plural alternation is –o vs. –i
(un coniglio vs. due conigli, ‘one rabbit’ vs. ‘two rabbits’), –e vs. –i (un maiale vs. tre
maiali, ‘one pig’ vs. ‘three pigs’), or –a vs. –e (due mucche vs. una mucca, ‘two cows’
vs. ‘one cow’). Furthermore, there was evidence that this target-like production
extended to new lexical items. As shown in (8), when presented with the picture of
an unknown referent (a camel), Chris first produced un camel (‘a camel’) with
Italian phonology, which the teacher recasted with its proper singular –o suffix as
cammello; then, with no prompts, albeit after a self-repair, Chris correctly produced
otto cammelli (‘eight camels’).
(8) t2 C
ah well oh eh un camel (spoken with an Italian accent)
[one-MASC.SG camel]
cammello un cammello
[camel-MASC.SG one-MASC.SG camel-MASC.SG]
è un cammello e questi?
[it is one-MASC.SG camel-MASC.SG and this-MASC.PL?]
uno d oh I’m counting them uno due tre quattro cinque sei sette otto
[one two three four five six seven eight]
otto cammellia li ... cammelli
[eight ... camels-MASC.PL]
Tonya G. Agostini and Catherine T. Best
On the other hand, at t2 Chris missed a plural suffix once, when presented with a
picture of ‘many’ penguins. Despite their high number, he decided to count them
all in Italian, but needed scaffolding from the teacher beyond number 20. When
he finally counted the forty-seventh penguin independently, he produced quarantasette *panguino (‘forty-seven panguin’, with a mispronunciation of a vowel in the
stem and, more importantly, a singular ending). Furthermore, Chris also oversupplied plural –e on a N in a singular context (un l g la *galline vs. due la galline, ‘one
chicken’ vs ‘two chickens’), and left the unmarked default –e ending on the single
adjective, producing *marrone (‘brown’) instead of marroni with reference to sette
cammelli (‘seven camels’) said by the teacher.
We can conclude that after only three weekly lessons, Chris had clearly moved
one step up in the developmental path, from the lemma access stage to the category procedure stage. On the other hand, when provided with two contexts for a
structure requiring the activation of the phrasal procedure, Chris failed to mark the
agreement on the adjective, whether in the singular that ends in –o (elefanti *grigio,
‘grey elephants’) or in –e (*marrone asini, ‘brown donkeys’, with English word
order), albeit in both cases the N was correctly marked as plural. As a matter of fact,
it is worth reporting here that, before producing these two NPs, when first required
to name animals and their colours, Chris seemed agitated and began with false
starts and repetitions even if the adjective was rosa, which remains unchanged in all
contexts. We could infer that at this stage the very production of a two word NP
seemed to be a problem for Chris.
At t3, Chris continued to improve. He correctly produced four examples of
plural –i marking on single Ns, using singular –o only once (due *pinguino, ‘two
penguin’). Interestingly, this occurred when Chris was prompted to produce this N
with two adjectives (i.e., ‘two black and white penguins’), even though during the
same task he had just produced the correct plural form on its own. Now that he
could categorically mark Ns with confidence, Chris began to also mark adjectives
at the category procedure stage, and produced a correct plural –i marker on a single adjective in one context (bianchi ‘white,’ referring to ‘penguins’) out of three. In
the other two contexts the adjectives were used in their default singular form. A
clear step forward at t3, however, was shown by agreement within NP, when Chris
produced two target-like singular vs. plural agreement contrasts (l’elefanti grigi vs.
una l’elefante grigio, ‘the grey elephants’ vs. ‘one grey elephant’; un gatto nero vs. due
gatti neri, ‘one black cat’ vs. ‘two black cats’) out of a total of 5 contexts. In both of
these cases, the plural adjectives require an –i suffix like their head Ns. In the
remaining 3 contexts the adjectives bear default singular forms: pinguini *nero
(‘black.M.SG penguin.M.PL’); due gatti *arancione (‘two orange.M.SG cat.M.PL’); and
pulcini *giallo (‘yellow.M.SG chick.M.PL’).
At t4, there are 7 examples of plural –i marking on Ns, five of which contrast
with singular –o contexts, and two with singular –a contexts (e.g., una torta.F.SG vs.
10. Exploring PT-based hypotheses in the L2 acquisition of a child with autism spectrum disorder
due *torti.F.PL, ‘one cake’ vs. ‘two cakes’). These latter two cases are not target-like,
yet they were scored with a plus in (5) not only because Chris supplied a plural suffix but also because this overgeneralisation of –i provided further evidence that he
was applying it productively. For the first time now, however, Chris also used –e for
marking plural on Ns, and produced three examples of this suffix, albeit two of
them were not target-like insofar as they contrast with singular –o contexts (e.g., un
fungo vs. tre *funghe, ‘one mushroom’ vs. ‘three mushrooms’). On the other hand,
the third example contrasts with singular –a contexts accurately (mela vs. mele,
‘apple’ vs. ‘apples’). This step forward with the expansion of the –e production at
category level was further confirmed at phrasal level. Whereas at t3 there were only
two cases of –i + –i agreement, at this test session Chris produced three of –e + –e
agreements, one of which is shown in (9).
(9) t4 C
una mela
[an apple-FEM.SG]
bravo che colore?
[good what colour? ]
I don’t want . sso [(re)d]
una mela rossa
[a red-FEM.SG apple-FEM.SG]
una mela rossa
[a red-FEM.SG apple-FEM.SG]
e queste?
[and these?]
oh quattro mele rosse
[oh four red-FEM.PL apple-FEM.PL]
Notice, however, that here again Chris seemed to be reluctant to say longer phrases, and when pushed further his resistance became explicit: ‘I don’t want three
words’. Nevertheless, we can conclude that his progress at t4 consisted mainly of a
consolidation of the –i plural marker and an expansion of the –e marker, not only
at category level but also within the phrase.
At t5, 6 weeks after the final 9th lesson, Chris’ progress continued insofar as
he appeared to overcome his initial feeling of distress in producing longer phrases.
He ventured in this direction in two different ways. First, he began to respond to
task probes with three-word grammatical phrases consisting of a numeral, the head
N and an adjective: this happened altogether 31 times out of 40 at t5. Of these 40
three-word phrases, 16 were in a plural context, and 7 out of them had the –i
Tonya G. Agostini and Catherine T. Best
marker on both elements, which he over-generalised several times in place of –e, as
in quattro *banani *gialli (‘four yellow bananas’). Secondly, for the first time, Chris
tried to produce NPs with more than one adjective. Although he was unsuccessful
in marking the plural for the second adjective (e.g., pinguini neri *bianco, ‘white
and black penguins’), the attempt is evidence of further ease in handling longer
stretches of his L2 language.
In sum, our initial research question posed in § 1.4 can now be answered positively. First, we have shown convincingly that our participant with HFASD-N was
able to acquire the morphophonology of an L2. Furthermore, we have illustrated
that he acquired the plural marking on Ns and adjectives along the developmental
path predicted by PT generally, and tested specifically among children learning the
same Italian structures by Di Biase (2002). That is, Chris produced singular and
plural contrasts first on single Ns and adjectives at the category procedure stage,
and then also on Ns and adjectives at the phrasal procedure stage. Because no
phrasal structure was produced before he could activate the category procedure,
PT’s implicational hierarchy is respected. Finally, data shows that Chris gradually
learned to overcome his initial reluctance to produce longer strings in Italian, and
eventually produced NPs that were both three word long and grammatically target-like.
With regard to the pace of his progress, a comparison between Chris and the
typically developing children in Di Biase’s (2002) study is not straightforward even
though in both studies the instructional period targeted the same structures and
lasted 12 weeks. On the one hand, Chris was 6;5 years old, had received minimal
previous exposure to Italian, and was then taught for the study in a one-to-one situation. On the other hand, Di Biase’s children were two-to-three years older, had
already been learning Italian for three years (albeit without any progress beyond the
lemma access stage), and were taught Italian all together in a large class. In (21) we
show the delayed post-test results for Di Biase’s children in the control and experimental groups regarding both single plural forms and phrasal agreement. Notice
that in the Di Biase study both groups received focus-on-form type treatment similar to Chris’, with the only difference that the experimental group received corrective feedback only on targeted structures. Figures in the table indicate that, while
the overall improvement was remarkable, and particularly so in the experimental
group, three children in the control group had not progressed beyond the lemma
access stage, and that among those who had, the default –i marker was by far more
accurate than the –e marker. Finally, even in the experimental group, who received
the optimal treatment, most children (i.e., all but three: Chr, Kat and Lau) produce
target-like plural agreement in less than half the required contexts. It is thus reasonable to conclude that Chris’s progress was also well within the range of the average
typically developing children. Further, he seemed to develop each new stage at a
fairly rapid pace.
10. Exploring PT-based hypotheses in the L2 acquisition of a child with autism spectrum disorder
(21) Delayed post-test production of plural forms and phrasal plural agreement in Di Biase’s
(2002) study
4. Conclusion
This is a first study worldwide of grammatical development in L2 with a child with
HFASD-N. Our results show that a child with HFASD-N can learn an L2 beyond
the acquisition of unanalysed single words and formulas. In a 12 week period, as
his Italian L2 utterances become longer, Chris developed from producing mostly
invariant single words to being able first to mark number in Ns, and then to form
agreement between head Ns and their modifiers within NPs. This development
Tonya G. Agostini and Catherine T. Best
follows the same route as that of typically developing children reported in Di Biase
(2002), and at a comparable pace, indeed perhaps a faster one.
In § 1.4 we mentioned two possible sources of early language delay in ASD.
The first hypothesis claims that social and communication deficits can lead to a
lower amount of linguistic input, and hence a slower pace of acquisition; the second that an underlying language processing deficit/difficulty leads to difficulties in
acquiring L2 grammar. The evidence gathered in this study clearly indicates that
Chris does not display underlying linguistic deficits; rather, with instruction that is
structured appropriately to address his social/communication difficulties, he was
able to acquire Italian inflectional morphology in NPs by the normal route and at
a comparable (or faster) speed to normal children. He even demonstrates the ability to mark plural with both the default plural –i suffix and the more marked nondefault plural –e suffix. This discrimination was acquired earlier than expected and
may reflect the detail-focused processing style of children with ASD (Happé & Frith
2006) and/or ability to hyper-systemise (Baron-Cohen 2009) novel L2 grammar.
Recent studies have shown that bilingualism in typically developing children
improves cognitive functioning, particularly executive functions such as attention
shifting and inhibitory control (Bialystok & Craik 2010). Since children with ASD
are posited to have deficits in executive functions (Ozonoff, Pennington & Rogers
1991), SLA may prove to be of considerable cognitive benefit for these children,
rather than posing problems for them as argued by Jellinek, Toppelberg, Snow &
Tager-Flusberg (1999). Future research would need to involve large group studies
of children with HFASD and typically developing controls. It would also need to
enlarge its scope in many directions, including investigations not only of more
structures dealt with by PT, but also of earlier stages of morphophonological development, such as phonological constancy (Best et al. 2009), that is, the ability to
recognise spoken words even in the face of phonetic variability in the words’ pronunciation across a variety of speakers and regional accents. Also, a finer-grained
analysis of speech production at a phonemic level may bring to light important differences in the way children with HFASD perceive and produce novel speech and
articulatory gestures.
The authors wish to thank Bruno Di Biase for guidance throughout the project, and
Camilla Bettoni for help in organising the chapter; then, the college of Arts (University
of Western Sydney), and Co.As.It. (Sydney), for financial assistance and providing an
excellent native speaking teacher of Italian; and, finally, speech pathologist Debra-Ann
Tanne for monitoring the progress of the participant for the duration of the study.
Connecting CALL and second language
development: e-tandem learning of Japanese
Satomi Kawaguchi
University of Western Sydney
1. Introduction
More than 20 years have passed since Computer Assisted Language Learning
(CALL) was introduced into L2 classrooms. The advantages of CALL in SLA
are well documented (e.g., Thorne 2008). In particular, recent advancements in
multimedia technology allow for ever more authentic communication
exchanges in L2 interaction. Text message exchange known as chat offers great
opportunities for the language learner to interact with a native speaker of a language instantly regardless of their physical distance. Among CALL activities,
chat is one of the most researched areas (e.g., Donaldson & Kotter 1999;
Schwienhorst 2002; Toyoda & Harrison 2002; Iwasaki & Oliver 2003; Kotter
2003; O’Rourke 2005). Most studies in the area of chat are conducted within
an interactionist approach. The Interaction Hypothesis (Long 1996) claims that
interaction plays an important role in L2 acquisition, which is promoted
through negotiation of meaning such as clarification requests and confirmation
checks. In L2 learning, language input (Krashen 1985; Long 1996) and learner
output (Swain 1985, 1995) play crucial roles. CALL can enhance both. In particular, online chat creates authentic contexts for interaction where the language
learner has opportunities to receive meaningful input and output. Furthermore,
negotiation of meaning and feedback on his/her output from the native speaker may promote ‘noticing’ the gap (Schmidt & Frota 1986) between the current
state of the interlanguage and the target language.
The type of CALL activity chosen in this chapter is electronic tandem (etandem for short, Cziko 2004), an activity in which participants engage in telecollaboration (Ware & Cañado 2007) via text-based Synchronous Computer
Mediated Communication (SCMC). In L2 e-tandem learning, a group of L2
students engages in a learning interaction with another group of students who
are native speakers of that language. These, in turn, are also learners of a second
language, which is the native language of the first group. So, each group is, alter-
Grammatical development in second languages, 291-306
Satomi Kawaguchi
natively, learning from, or teaching the other group (cf. Lewis & Walker 2003
for detailed explanations of tandem language learning). In recent years, research
on e-tandem has grown in the field of SLA. Studies on e-tandem learning show
how learners can use negotiation opportunities qualitatively and quantitatively
when communication problems occur (Iwasaki & Oliver 2003; Sotillo 2005;
Lee 2006; Ware & O’Dowd 2008; Bower & Kawaguchi 2011). Other CALLrelated studies deal with technological design, students’ evaluation or perception and cultural issues. So far few studies on e-tandem have examined the effectiveness of CALL activities on L2 development (e.g., Bower & Kawaguchi
2011; Iwasaki & Oliver 2003). For learners, teachers and researchers it is important to be aware of language development when experimenting and practicing
CALL in order to make it more meaningful.
This chapter presents an early attempt to link CALL and PT. I will use as
an example e-tandem learning via chat between learners of Japanese L2 in
Australia and learners of English L2 in Japan, with the focus on Japanese L2. I
will analyse chat logs according to PT’s developmental stages, and measure the
learners’ L2 production as the text chatting activities progress over a two-month
period. Furthermore, I will attempt to link CALL to my main research line in
this field, which aims at extending the application of PT’s Steadiness Hypothesis
across learning modalities, and test whether the production of text chat language
conforms to the PT schedules formulated for orality. The Steadiness Hypothesis
was originally proposed by Pienemann (1998: 273-297) when he convincingly
showed that acquisitional sequences are not affected by different communicative
tasks. So far only Håkansson & Norrby (2007) and Rahkonen & Håkansson
(2008) have investigated steadiness in a modality other than speech, namely writing. Rahkonen & Håkansson’s (2008) cross-sectional study on L2 Swedish writing looks at formal and semiformal writing. Their results show that acquisition
patterns in the two types of writing present both similarities to, and differences
from PT sequence for speech. That is, using the old PT terminology of these two
researchers, the sequence in speech is ‘tense < VP agr < INV’; in the semiformal
writing corpus ‘tense = VP agr < INV’; and in the formal writing corpus ‘tense =
VP agr =INV’. Thus, acquisitional patterns may differ significantly according to
whether corpora include semiformal writing closer to speech or formal writing.
So far, no study has been undertaken on SCMC within the framework of PT. It
is interesting to examine whether the L2 text production via SCMC shows a
close relationship with PT stages in speech, since it occurs in real-time interaction and is thus closer to online speech production than either formal or semiformal writing.
In order to bridge the gap in research the following research questions are
addressed in this chapter:
11. Connecting CALL and second language development: e-tandem learning of Japanese
• Do tandem learning activities conducted through chat promote L2 production?
More specifically, does the number of word types produced by L2 speakers increase
during e-tandem learning activities?
• Does online production in text messaging (written production) show development
as measured by PT developmental stages?
• Is development across modalities steady in terms of the trajectory of morphological and syntactic structures defined by PT’s Steadiness Hypothesis for oral production?
2. Synchronous CMC and SLA research
Communication technology enables us to interact with others in various ways. For
example, e-mail exchange is an asynchronous way of Computer-Mediated
Communication (CMC), whereas text chat is synchronous, the “most interactive
end of the CMC spectrum” (Paramskas 1999: 17). As such, online text messaging
shares many characteristics with face-to-face conversation. Participants have far less
time to edit a message than in other types of CMC such as composition in blog or
e-mail (Levy & Stockwell 2006). Because of real time interaction, they can negotiate meaning in a way similar to face-to-face conversation (Blake 2000; Toyoda &
Harrison 2002; Iwasaki & Oliver 2003). Furthermore, the text-based medium can
lead learners to noticing more problematic L2 language than face-to-face communication (Lai & Zhao 2006), and may increase students’ attention to linguistic
form (Warschauer 1996).
Payne & Whitney (2002) claim that, in L2 learning, chat may achieve better
outcomes than face-to-face activities because the L2 learner benefits from slower
language processing while at the same time having to process utterances largely ‘on
the fly’, that is, with little or no advance planning, as is the case of face-to-face verbal communication. These authors demonstrate that a group of learners with
blended activities (both face-to-face and SCMC) improve their oral proficiency
better than an equivalent group with face-to-face learning only, when measured by
Oral Production Interview ratings on comprehensibility, fluency, vocabulary use,
grammar, and pronunciation.
Payne & Whitney (2002) also show that SCMC can develop the same cognitive mechanisms underlying spontaneous oral communication, and thus facilitate
the acquisition of L2 speaking skills. Part of their claim is based on Levelt’s (1989)
Speech Model (cf. § 2.1, ch. 1, this volume) in relation to the role of working
memory for language processing. Working memory has a limited capacity, and can
thus attend only to a limited amount of information immediately in real-time.
There is a trade-off between its information processing role and its storage role in
order to cope with the incremental nature of language production/comprehension
Satomi Kawaguchi
(Kahnemann 1973; Carpenter & Just 1989). This means that if a speaker processes information more efficiently, more working memory space becomes available to
devote to other components of language processing and storing. Beginning L2
learners need to place attentional resources on various steps of language processing.
For example, they need to search for appropriate words from their mental lexicon
and place them within an appropriate syntactic frame with appropriate functional
assignment. Furthermore, they need to determine the correct morphosyntactic
forms and select corresponding phonological units. If these processes are not automatic, a burden is placed on the phonological processing (Baddeley 2007). In L1,
lexical access and articulation are largely automatic, and the speaker needs to pay
attention only to conceptualization and careful delivery. On the other hand, at the
earlier stages of L2 learning, most processing components are controlled, resulting
in speech with longer and more frequent pauses, etc. (Poulisse 1999). PT also uses
Levelt’s (1989) Model and claims that L2 acquisition is constrained by the learner’s current state of procedural skills. As particular procedural skill components are
automatised, more attentional resources become available to the L2 speaker, who
can thus achieve further language processing, such as more fluent speech or/and
construction of higher stage-structures.
What may then be the advantages of using text chat communication in
SLA? First, the speed of information exchange is a little slower than in speaking
simply because one cannot write as fast as one speaks. This gives L2 learners
opportunities to process L2 messages at a slower pace while going through a
similar language processing as in face-to-face communication. Secondly, the
availability of previous messages (the visual co-text) can help learners to reduce
the amount of information to be stored in working memory. Therefore, they
can free up more attentional resources for the L2 lexicon and structures while
maintaining pace in interaction. Thus learners may gain greater benefit from
slower language processing while going through similar processing as may occur
in face-to-face communication involving conceptualization, formulation, articulation (typing) as well as comprehension of the interlocutor’s utterance.
Thirdly, Payne & Whitney (2002) believe that chat communication may provide opportunities for learners to experience a sort of “conversation simulator”
which may be effective especially for less confident, shyer or linguistically weaker students, who tend not to take full advantage of interaction with teacher or
fellow students in face-to-face settings. Furthermore, the text-based medium
may increase students’ attention to linguistic form (Warschauer 1996) due to
the slower information exchange.
Chat logs obtained from a tandem project can therefore offer an interesting
opportunity to investigate whether L2 development follows PT stages also in this
communication mode. In order to measure learners’ oral proficiency, Payne &
Whiteney (2002) use Oral Production Interview rating on comprehensibility, flu-
11. Connecting CALL and second language development: e-tandem learning of Japanese
ency, vocabulary usage, syntax, and pronunciation with two examiners rating on a
50 point scale. Although interrater reliability was high (0.86 for the pre-test and
0.94 for the post-test), it might be fruitful to assess L2 development using a more
objective measurement such as the PT stages.
3. The e-tandem project
This section describes a tandem learning project1 using chat via instant messaging
between language classes at the University of Western Sydney in Australia and at
the Kanda University of International Study in Japan. This project was organised
as an out-of-class L2 activity at each institution aiming at collaborative learning in
a more authentic context.
The participants are 21 second-year students of L2 Japanese in Australia, and
21 first-year students of L2 English in Japan.2 The two groups are compatible,
albeit not perfectly matched in several ways. First, the levels of L2 English in Japan
are generally higher than those of L2 Japanese in Australia, because Japanese students learn English for six years at secondary school as a compulsory subject before
starting university, whereas most Australian students start learning Japanese at university. Secondly, all students of English in Japan are native speakers of Japanese,
whereas the students of Japanese in Australia are a mixed group, reflecting the multicultural nature of Australian society. So, according to the answers to a questionnaire enquiring about their background, 10 students are foreigners, four identified
themselves as immigrants of ethnic background, and the rest as Australians. In
order to enhance compatibility, when the project started all learners wrote a short
self-introduction so that their teachers could match the tandem pairs according to
mutual interests.
The project includes three chat sessions, distributed over two months, each
lasting at least 30 minutes in each of the two languages. The first session was conducted during class time to ensure that everything worked; the second and third
sessions were then organized by tandem pairs autonomously.
Before each chat session, students were given a broad conversation topic. For
session 1 this was more oriented towards the here-and-now, and dedicated to self-
1 This study was supported by the University of Western Sydney, LTAP (Learning &
Teaching Action Plan).
2 Two second-year students at Kanda University of International Study and four thirdyear students at University of Western Sydney also volunteered on several occasions in
order to replace students who were unable to attend, and thus match the number of
Satomi Kawaguchi
introduction and one’s family, then it gradually moved on to more challenging
areas: session 2 was about university life, and session 3 about each other’s culture
and related controversial issues. Participants were required to do the following:
• organise the date and the time of the two subsequent sessions via e-mail, and agree
on the topics for conversation;
• check and study vocabulary which may be useful for conversing on the agreed
• log in and participate in the chat at the designated time, and converse on the chosen topics half the time in English and half in Japanese at each session;
• submit to their teachers their observations and chat logs after each session;
• study the tandem partners’ L2 production in the chat log, and send them some
corrections and suggestions via e-mail after each chat.
In § 4 below I will report on the analysis of L2 Japanese data only. Because not all
participants attended all chat sessions, some students had a three-way chat or different partners across sessions. I thus selected five learners of Japanese who participated in all three sessions with the same tandem partner.
As mentioned above, the e-tandem sessions were an out-of-class activity
organised over two months as part of the Japanese L2 course. In the face-to-face
classroom, new lexicon and the following grammatical structures were introduced
weekly during this period:
potential forms of the verb;
conditional clause;
various interrogative sentences (asking about subject/object/adjunct);
auxiliary verbs;
plain forms of adjective, copula and verb;
sentential modification for noun.
Also, learners might have learned new words and structures from their partners
during the sessions. However, learning a new structure as knowledge is different
from using it in time-constrained conditions. In order to produce the structure
online, processing efficiency needs to be attained. In my analysis here, although
development will be measured as the three sessions progress, the main focus is on
language use rather than acquisition. Indeed, the e-tandem project is not
designed to test the learner’s acquisition of particular morphological structure
(e.g., verbal inflections) and syntactic structure (e.g., passive voice) in order to
identify his/her PT stages. Instead, what is clearly seen from the chat log is each
L2 learner’s language use (such as selection of lexicon and syntactic structures) in
real time interaction.
11. Connecting CALL and second language development: e-tandem learning of Japanese
In order to exemplify some of the characteristics of text chat such as the pattern of turn taking, an excerpt from session 3 by Lee and his Japanese partner
Mayumi is reproduced in the Appendix. Text chat and face-to-face conversation
share many similarities including negotiation of meaning, such as the confirmation
checks in turn 2 and 3, and the clarification requests in turns 17-18. Also, since the
conversation is between a learner of Japanese and a native speaker, code-switching
can occur (e.g., turns 2 and 5). On the other hand, there are some differences
between this modality and face-to-face conversations. For example, the pattern of
turn taking is different. In chat, turns are counted whenever a typist posts the text
message by pressing the “enter” key. Thus, unlike in face-to-face conversation, participants may have different turn-taking opportunities, because a fast typist may
post another text message (e.g., turns 7, 9 and 16) before his/her partner responds
to the previous one, which may result in uneven distribution of turns across interactants. Sometimes a typist may accidentally post a turn before he/she finishes a
sentence and continues in the next turn (e.g., turns 25-26). Conversational
sequence can be interrupted by some unexpected turns (e.g., turn 20) due to the
delay/gap between the time of posting a message and the time of receiving the
response. Furthermore, sometimes a participant may move on to a new topic
before a negotiation is resolved (e.g., turn 20).
4. Language Development
4.1. Chat production
The numbers of turns produced in each chat session by our five learners are shown
in (1). In the table “total” indicates the turns produced by both interlocutors,
namely the native speakers of Japanese (NSs) and the L2 learner (NNSs); “NNS”
indicates the turns produced by the L2 learner only; and the numbers in brackets
indicate the turns produced in English.3 So, for example, Chaz’s session 2 comprises 109 turns, with Chaz himself posting 66 of them, three of which are written in
English. Looking at all five learners, we can see that the text messaging activity
seems successful, in so far as the number of turns sent in L2 Japanese increases for
all of them after the first break-in session, although more so in session 2 than in
session 3.
3 Codeswitches were encouraged by the topic, especially in session 3, when controversial
issues in each other’s culture were discussed. For example, two pairs talked about “whaling” in Japanese. One of these pairs sometimes used English words/sentences when
communication problems occurred.
Satomi Kawaguchi
(1) Number of turns by learner and session
Because a turn may consist of one word or a few sentences, a more precise measure of the amount of language used is the word count in terms of tokens, as
shown in the graph in (2). Here we can see that all five learners increase their
tokens from session 1 to session 2, but then four out of five decrease them in
session 3, as they do with their turns. Dani is the only learner who increases the
number of both turns and words in the third session. The reason for smaller
production in this session can be found in the topic of the conversation: whereas the topic for session 2 is “university life” and could be handled more descriptively, the topic for session 3 engages the learners on controversial issues relating to each other’s culture. Its argumentative nature may have slowed down the
conversation quantitatively as learners would have required more time for conceptualising the message. In the graph we also notice that, if the curve seems
similar across all learners except Dani, there is great variation among them in
the number of actual tokens produced. So, in session 2, for example, Chris posts
61 word tokens and Chaz 513.
(2) Number of words (tokens) by learner and session
11. Connecting CALL and second language development: e-tandem learning of Japanese
4.2. The lexicon
Using the Key Word In Cotext concordance programme (KWIC, created by
Nihon University and available online), I calculated the size of the lexicon used
during the chat sessions in terms of types (i.e., different words). This is illustrated
in the two graphs in (3)-(4). The graph in (3) shows number of types used by each
learner in each session. Considering the results of the token analysis in the graph
in (2), it is not surprising to see that as the tokens vary, so do the types; that is,
numbers for types increase for all students between session 1 and 2, but keep
increasing only for Dani between sessions 2 and 3.
(3) Number of words (types) by learner and session
Because the topic of conversation changed in each chat session, and learners were
required to study vocabulary relating to the agreed conversational topics prior to
the chat session, we expected learners to produce some different lexical items every
time. In order to measure this, the graph in (4) shows the cumulative number of
word types without including those already used in the previous session; that is, for
each learner, the types of session 1 are added to those of session 2, minus those
already used in session 1; and the types of sessions 1 and 2 are added up to those
(4) Cumulative number of word types by learner and session
Satomi Kawaguchi
of session 3, minus those already used in sessions 1 and 2. Thus the figures in this
graph indicate the actual size of the lexicon used over the three sessions, and show
that the number of types steadily increased at every session for all learners.
Comparing learners, however, we can see great variation among them, as we have
already noticed with regards to their word tokens. For example, the figure for Chris
in all three sessions is 83, for Dani 166, and for Chaz 354.
4.3. Morphology and syntax
The grammatical development of the five learners from session 1 to session 3 is
illustrated in (5). For the description of PT stages for Japanese L2 morphology and
syntax, I refer to chapter 4, §§ 2-3, this volume. However, as a matter of convenience, the first stage of lemma access is omitted here, because all learners already
safely reached it. In the analysis, as mentioned in chapter 1, § 5, this volume, lexical or formal variation is used as the acquisition criterion for morphology, whereas for syntax one occurrence of the relevant structure is considered sufficient.
(5) The learners’ progress over the three chat sessions
11. Connecting CALL and second language development: e-tandem learning of Japanese
As (5) shows, Chris’s numbers for both morphological and syntactic structures
are small, because most of his turns are single words or fragments of phrases.
However, he shows some progress during the three sessions. Morphologically,
he stays at the category procedure stage in all three sessions, but he produces a
variety of forms: polite (affirmative), negative and question forms in session 1,
then adding the plain form in sessions 2 and 3. Syntactically, he moves one
stage up with both word order and lexical mapping in session 2, then proceeds
no further.
Also Iwan produces few sentences in the three sessions as he often uses fixed
expressions. In terms of morphology, he produces only structures requiring category procedure (i.e., V inflection) in all three sessions. In terms of syntax involving word order, he is at the XP canonical word order stage at session 1 and does
not progress further by session 3. In terms of syntax involving mapping, he is at
the default mapping stage at session 1, and progresses to default mapping and
to the additional argument stage at session 2. We can conclude that his progress
is limited.
Dani’s progress, on the other hand, is remarkable in several ways.
Although, like Chris, he starts off in session 1 with only low numbers for the
earliest stage in both morphology and syntax, he then moves up two stages in
both the morphological and the lexical mapping schedules. Furthermore, morphologically both the numbers and range of structures increase substantially in
sessions 2 and 3. That is, he produces 2 V inflections (one negative and the
other desiderative) in session 1, which belong to the category procedure stage.
Then he adds more V inflections, including past tense and plain forms, as well
as V-te V structure expressing progressive aspect in session 2, which belongs to
the phrasal procedure stage. He also produces noncanonical case marking once
in sessions 2 and 3. Finally, although only one token of noncanonical case marking belonging to the S-procedure stage is produced, it can be assumed that he
has acquired it by session 3 because he has already produced this structure with
different lexical items in session 2. In terms of syntax, dramatic achievement is
observed. Despite producing only sentences with canonical word order and
default mapping in session 1, Dani constructs sentences with benefactive and
passive Vs in sessions 2 and 3.4 He also produces XP plus canonical order in session 3. In addition, in each of the last two sessions he actively uses further new
structures learned in class: conditional clause in session 2, and sentential noun
modification in session 3. Compared to Chris, Dani then progresses enormous4 Dani’s production of passive sentence is as follows: 日本で 何頭クジラが 毎年
殺されるか (Nihon-de nantoo kujira-ga maitoshi korosareru ka) “In Japan how many
whales are killed every year?”.
Satomi Kawaguchi
ly over the two months, shooting ahead by two stages both morphologically and
syntactically, and using twice the number of lexical types. He is also the only
learner who never codeswitches to English.
Lee is at the category procedure stage in morphology in session 1 as he produces V inflection 23 times. In this session, he also uses one V-te V structure, which
however is insufficient to place him at the phrasal procedure stage yet. In session 2,
remarkably, he produces the V-te V and other combinatorial structures 19 times,
and thus safely reaches the phrasal procedure stage. In session 3, he produces morphological structures requiring the category procedure 11 times and the phrasal
procedure 20 times, and attempts noncanonical case marking (i.e., S-procedure)
twice, but both these cases turn out to be wrong. Thus Lee’s morphological progression is from the category procedure stage in session 1 to the phrasal procedure
stage in sessions 2 and 3. As for syntax, in session 1 Lee is already at the noncanonical word order stage when, besides structures of the lower stages, he also produces
one TOPOBJ SV structure. In sessions 2 and 3 this rule is consolidated. In terms of
lexical mapping, he is at the intermediate stage of default mapping and additional
argument in session 1, and remains there in sessions 2 and 3.
Chaz also improves over the sessions. In session 1 he starts off with robust figures for both morphology at the phrasal procedure stage and syntax at the intermediate lexical mapping stage, and he is the learner who progresses furthest in syntax
in the subsequent two sessions. In morphology he then seems to reach the S-procedure stage with just one tentative production of noncanonical case marking in
session 2, shown in brackets in (5). In syntax, on the other hand, he comfortably
reaches the noncanonical word order stage by producing several OBJ topicalisations. Furthermore, he produces one benefactive structure in session 2.
In sum, the analysis of grammatical development over the three sessions indicates that the answers to my research questions formulated in § 1 are all positive.
The first question asks whether the cumulative number of words produced by L2
speakers increases during e-tandem. As the sessions progress, all learners’ use more
language, in terms of both turn numbers and word tokens. Also word types
increase dramatically, although with significant individual differences (almost
1:4.5). The same tendency is apparent in the development of morphology and syntax. Two learners (Dani and Lee) out of five progress to new morphological stages.
With regard to syntax, three learners progress on word order (Chris, Dani and
Chaz), and four extend their mapping (the exception is Lee).
The second question asks whether developmental sequences in text chat follow the trajectory defined by PT for oral production. They certainly do, as these
results demonstrate, with no learner skipping stages. As for the third question,
whether PT stages hold across modalities, the answer is positive and detailed results
also show that chat may produce results closer to speech than formal and informal
writing. Thus, e-tandem chat is shown to be an effective, guided L2 learning activ-
11. Connecting CALL and second language development: e-tandem learning of Japanese
ity which can increase L2 production. With regard to the large individual differences in students’ learning outcomes, at this stage we cannot say whether they are
due to a different response to this set of activities.
5. Conclusion
CALL is now a natural part of L2 learning (Chambers & Bax 2006). This requires
new roles for the language teacher, who must no longer just teach “knowledge” of
the L2 in the classroom, but also design an L2 learning environment which promotes and supports meaningful L2 activities (Thomson 2007). CALL can offer
these activities for out-of-class practice, thus offering new opportunities to achieve
better L2 outcomes. However, there may be a pitfall with CALL if SLA theories are
not taken into account. For example, CALL activities may seem to suite learners’
current needs and lifestyles, but they run the risk of being just an enjoyable experience and produce little L2 progress if teachers do not monitor the L2 learners
progress using a reliable measurement such as PT stages.
This chapter shows that PT has a positive potential to contribute to the field
and is capable to connect CALL and SLA. Further, the Steadiness Hypothesis is
confirmed not only across tasks but also across modalities. This suggests a great
potential for an online extended version of PT’s Rapid Profile (cf., e.g., Pienemann
& Mackey 1993; Keßler 2007; Pienemann & Keßler 2010) in monitoring L2
grammatical development with CALL (especially text messaging) by teachers or
learners themselves.
A chat log excerpt from session 3 with Lee (L, learner) and Mayumi (M, native speaker)
Satomi Kawaguchi
11. Connecting CALL and second language development: e-tandem learning of Japanese
Ågren, M. 2009. Morphological development in written L2 French: A Processability perspective. In Research in Second Language Acquisition: Empirical Evidence Across
Languages, D. Keatinge & J.-U. Keßler (eds), 121-151. Newcastle, UK, Cambridge
Scholars Publishing.
Aissen, J. 2003. Differential object marking: Iconicity vs. economy. Natural Language and
Linguistic Theory 213, 435-448. Glossologia 39 (2-3), 7-20.
Alsina, A. 1996. The Role of Argument Structure in Grammar. Stanford, CA, CSLI
Anđelković, D. 2000. Razumevanje imeničke morfologije srpskog jezika na predškolskom
uzrastu [The acquisition of inflectional case system in Serbian]. Psihologija 33 (1-2),
Andersen, R.W 1984. The one-to-one principle of interlanguage construction. Language
Learning 34, 77-95.
Andersen, R.W. 1993. Four operating principles and input distribution as explanations for
underdeveloped and mature morphological systems. In Progression and regression in
language, K. Hyltenstam & A. Viborg (eds), 309-339. New York, Cambridge
University Press.
Andorno, C., Bernini, G., Giacalone Ramat, A. & Valentini A. 2003. Sintassi della frase
semplive. In Verso l’Italiano. Percorsi e Strategie di Acquisizione, A. Giacalone Ramat
(ed.), 116-154. Roma, Carocci.
Andrews, A. & Manning, C. 1999. Complex Predicates and Information Spreading in LFG.
Stanford, CA, CSLI Publications.
APA. 1994. Diagnostic and Statistical Manual of Mental Disorders. Washington, DC,
American Psychiatric Association.
APA. 2013. Diagnostic and statistical manual of mental disorders, 5th edition, and clinical
utility. The Journal of nervous and mental disease 2019 (2), 727. doi:
Artoni, D. & Magnani, M. 2013. LFG contribution in second language acquisition research: The development of case in Russian L2. In Proceedings of the LFG13 Conference.
M. Butt & T.H. King (eds), 69-89. Stanford, CA, CSLI Publications.
Asudeh, A. 2004. Resumption as Resource Management. PhD thesis, Stanford University,
Asudeh, A. & Toivonen, I. 2010. Lexical-Functional Grammar. In The Oxford Handbook
of Linguistic Analysis, B. Heine & H. Narrog (eds), 425-458. Oxford & New York,
Oxford University Press.
Grammatical development in second languages, 307-332
Auer, P. (ed.) 1998. Code-Switching in Conversation: Language, Interaction and Identity.
London, Routledge.
Avrutin, S. & Rohrbacher, B. 1997. Null subjects in Russian inverted constructions.
Proceedings of the Fourth Annual Workshop on Formal Approaches to Slavic Linguistics.
Ann Arbor, MA, Slavic Publications.
Backhouse, A.e. 1993. The Japanese Language: An Introduction. Oxford, Oxford University
Baddeley, A. 2007. Working Memory, Thought, and Action. Oxford, Oxford University
Bailey, N., Madden, C., & Krashen, S. 1974. Is there a “natural sequence” in adult second
language learning? Language Learning 24 (2), 235-243.
Baron-Cohen, S. 2009. Autism: The empathizing-Systemizing (e-S) Theory. Annals of the
New York Academy of Sciences 1156, 68-80.
Bartolucci, G., Pierce, S. & Streiner, D. 1980. Cross-sectional studies of grammatical morphemes in autistic and mentally retarded children. Journal of Autism and
Developmental Disorders 10 (1), 39-50.
Baten, K. 2011. Processability Theory and German. Case acquisition. Language Learning
61 (2), 455-505.
Baten, K. 2013. The Acquisition of the German Case System by Foreign Language Learners.
Amsterdam, Benjamins.
Bates, e. 1976. Language and Context: The Acquisition of Pragmatics. New York, Academic
Bates, e. & MacWhinney, B. 1989. Functionalism and the Competition Model. In The
Cross-linguistic study of sentence processing, B. MacWhinney & e. Bates (eds), 3-73.
New York, Cambridge University Press.
Berman, R. & Slobin D.I. 1994. Relating Events in Narrative; Vol 1, A Crosslinguistic
Developmental Study. Hillsdale, NJ, erlbaum.
Best, C.T., Tyler, M.D., Gooding, T.N., Orlando, C.B. & Quann, C.A. 2009.
Development of phonological constancy: Toddlers’ perception of native- and
Jamaican-accented words. Psychological Science 20 (5), 539-542.
Bettoni, C. 1991. Language variety among Italians: Anglicisation, attrition, attitudes. In
Language in Australia, S. Romaine (ed.), 263-269. Cambridge, Cambridge University
Bettoni C. & Di Biase B. 2005. Sviluppo obbligatorio e progresso morfosintattico, un caso
di Processabilità in italiano L2. ITALS. Didattica e Linguistica dell’Italiano come Lingua
Straniera 2 (1), 27-48.
Bettoni, C. & Di Biase, B. 2011. Beyond canonical order. The acquisition of marked word
orders in Italian as a second language. In EUROSLA Yearbook 11, L. Roberts, G.
Pallotti & C. Bettoni (eds), 244-272. Amsterdam, Benjamins.
Bettoni, C., Di Biase, B. & Ferraris, S. 2008. Sviluppo sintattico e sviluppo morfologico:
ipotesi di corrispondenze nella Processabilità dell’italiano L2. In Competenze Lessicali
e Discorsive nell’Acquisizione di Lingue Seconde, G. Bernini, L. Spreafico & A. Valentini
(eds), 355-382. Perugia, Guerra.
Bettoni, C., Di Biase, B. & Nuzzo, e. 2009. Postverbal subject in Italian L2 – a
Processability Theory approach. In Research in Second Language Acquisition: Empirical
Evidence Across Languages, D. Keatinge & J.-U. Keßler (eds), 153-173. Newcastle,
UK, Cambridge Scholars Publishing.
Bettoni, C. & Fratter, I. 2013. Acquiring reflexive constructions in Italian as a second language – A Processability Theory account. Paper presented at the 23rd Annual
Conference, European Second Language Association (EUROSLA 23), Amsterdam, 2831 September 2013.
Bever, T.G. 1970. The cognitive basis for linguistic structures. In Cognition and the
Development of Language, J.R. Hayes (ed.), 279-362. New York: Wiley.
Bialystok, e. & Craik, F.I.M. 2010. Cognitive and linguistic processing in the bilingual
mind. Current Directions in Psychological Science 19 (1), 19-23.
Blake, B. J. 1994. Case. Cambridge, Cambridge University Press.
Blake, R. 2000. Computer mediated communication: A window on L2 Spanish interlanguage. Language Learning and Technology 4 (1), 120-136.
Bock, K. & Levelt, W. 1994. Language production. Grammatical encoding. In Handbook
of Psycholinguistics, M.A. Gernsbacher (ed.), 945-984. New York, Academic Press.
Bonilla, C.L. 2012. Testing Processability Theory in L2 Spanish: Can readiness or markedness predict development? Doctoral dissertation, University of Pittsburgh, PA.
Bonilla, C.L. 2014. From number agreement to the subjunctive: evidence for
Processability Theory in L2 Spanish. Second Language Research, 0267658314537291.
Bonilla, C.L. 2015. From number agreement to the subjunctive: evidence for
Processability Theory in L2 Spanish. Second Language Research 31 (1), 53-74.
Boss, B. 1996. German grammar for beginners: The Teachability Hypothesis and its relevance to the classroom. In Who’s Afraid of Teaching Grammar? (The University of
Queensland Papers in Language and Linguistics, vol. 1), C. Arbon´es-Sol`a, J. RolinIantziti & R. Sussex (eds), 93-100. Brisbane, University of Queensland Press.
Boss, B. 2004. Wann ich habe Zeit, ich koche gern: Zum erwerb der Inversion und
Nebensatzwortstellung durch australische Deutschstudierende. Deutsch als
Fremdsprache 41, 28-32.
Bossong, G. 1983-1984. Animacy and Markedness in Universal Grammar. Glossologia 23, 7-20.
Bossong, G. 1991. Differential object marking in Romance and beyond. In New Analyses
in Romance Linguistics. Selected Papers from the XVIII Linguistic Symposium on
Romance Languages, D. Wanner & D. Kibbee (eds), 143-170. Amsterdam,
Boucher, J. 2012. Research review: Structural language in autistic spectrum disorder –
characteristics and causes. Journal of Child Psychology and Psychiatry 53 (3), 219-233.
doi: 10.1111/j.1469-7610.2011.02508.x.
Bower, J. & Kawaguchi, S. 2011. Negotiation of meaning and corrective feedback in
Japanese/english eTandem article. Language Learning and Technology 15 (1), 41-71.
Bowles, M. & Montrul, S. 2009. Instructed L2 acquisition of Differential Object Marking
in Spanish. In Little Words: Their History, Phonology, Syntax, Semantics, Pragmatics, and
Acquisition, R.P. Leow, H. Campos & D. Lardiere (eds), 199-209. Washington,
Georgetown University Press.
Bresnan, J. 1998. Pidgin genesis in Optimality Theory. In Proceedings of the LFG98
Conference, Brisbane, M. Butt & T. Holloway King (eds), 1-18. Stanford, CA, CSLI
Publication, www.csli-publications.stanford.ed/LFG/3/lfg98.html.
Bresnan, J. 2000. Optimal Syntax. In Optimality Theory: Phonology, Syntax, and Acquisition,
J. Dekkers, F. van der Leew & J. van de Weijer (eds), 334-385. Oxford, Oxford
University Press.
Bresnan, J. 2001. Lexical-Functional Syntax. Oxford, Blackwell.
Bresnan, J. & Mchombo, S. 1987. Topic, pronoun and agreement in Chichewa. Language
63, 741-782.
Brigance, A.H. 1991. Brigance K & 1 Screen for Kindergarten and First Grade – Revised.
North Billerica, MA, Curriculum Associates.
Brown, R. 1973. A First Language. Harvard, MA, Harvard University Press.
Browne, W. 1993. Serbo-Croat. In The Slavonic Languages, B. Comrie & G.G. Corbett
(eds), 306-387. London, Routledge.
Carpenter, P.A. & Just, M.A. 1989. The role of working memory in language comprehension. In Complex Information Processing, D. Klahr & K. Ktovsky (eds), 31-68.
Hillsdale, NJ, erlbaum.
Caselli, C.M., Leonard, L.B., Volterra, V. & Campagnoli, M.G. 1993. Toward mastery of
Italian morphology: A cross-sectional study. Journal of Child Language 20, 377-393.
CeFR 2001. Modern Languages: Learning, Teaching, Assessment. A Common European
Framework of Reference. Cambridge, Cambridge University Press.
Chambers, A. & Bax, S. 2006. Making CALL work: Towards normalization. System 34,
Charters, H., Dao, L. & Jansen, L. 2011. Reassessing the applicability of Processability
Theory: The case of nominal plural. Second Language Research 27 (4), 509-533.
Cheng, Lai-Shen L. & Corver, N. (eds) 2006. Wh-Movement: Moving On. Cambridge,
MA, MIT Press.
Chini, M. & Ferraris, S. 2003. Morfologia del nome. In Verso l’Italiano. Percorsi e Strategie
di Acquisizione, A. Giacalone Ramat (ed.), 37-69. Roma, Carocci.
Choi, H.-W. 1999. Optimizing Structure in Context: Scrambling and Information structure.
Stanford, CA, CSLI Publications.
Choi, H.-W. 2001. Phrase structure, information structure, and resolution of mismatch. In
Formal and Empirical Issues in Optimality Theoretic Syntax. P. Sells (ed.), 17-62.
Stanford, CSLI Publications.
Chomsky, N. 1977. On Wh-movement. In Formal Syntax, P.W. Culicover, T. Wasow & A.
Akmajian (eds), 71-133. New York, Academic Press.
Clahsen, H. 1980. Psycholinguistic aspects of L2 acquisition. In Second Language
Development, S. Felix, (ed.), 57-79. Tübingen, Narr.
Clahsen, H. 1984. The acquisition of German word order: A test case for cognitive
approaches to L2 development. In Second Language, R.W. Andersen (ed.), 219-242.
Rowley, Newbury House.
Clahsen, H., Meisel, J. & Pienemann, M. 1983. Deutsch als Zweitsprache. Der Spracherwerb
ausländischer Arbeiter. Tübingen, Narr.
Clahsen, H. & Felser, C. 2006, Grammatical processing in languagelearners. Applied
Psycholinguistics 27, 3-42.
Clyne, M. 2003. Dynamics of Language Contact. Cambridge, Cambridge University Press.
Comrie, B. 1979. Definite and animate direct objects: A natural class. Linguisitica Silesiana
3, 13-21.
Comrie, B. 1989. Language Universals and Linguistic Typology. Oxford, Blackwell.
Comrie, B. 2011. Russian. In The Major Languages of the World, B. Comrie (ed.), 274-288.
London & New York, Routledge.
Conroy, L. 2007. extended Processability Theory and its application to L2-Swedish. In
Proceedings, In Between Wor(l)ds: Transformation and Translation, University of
Melbourne. http://repository.unimelb.edu.au/10187/2213
Cook, V. 1993. Linguistics and Second Language Acquisition. New York, St. Martin’s.
Coonrod, e.e. & Stone, W.L. 2005. Screening for autism in young children. In Handbook
of Autism and Pervasive Developmental Disorders, F.R. Volkmar, R. Paul, A. Klin & D.
Cohen (eds), 3rd ed., vol. 1, 707-729. Hoboken, NJ, Wiley & Sons.
Corbett, G. & Browne, W. (2011). Serbo-Croat: Bosnian, Croatian, Montenegrin,
Serbian. In The Major Languages of the World, B. Comrie (ed.), 330-346. London &
New York, Routledge.
Cziko, G. 2004. electronic tandem language learning (eTandem): A third approach to
second language learning for the 21st century. CALICO Journal 22 (1), 25-39.
Dalrymple, M. 2001. Syntax and Semantics: Lexical Functional Grammar. San Diego, CA,
Academic Press.
Dalrymple, M. & Nikolaeva, I. 2011. Objects and Information Structure. Cambridge,
Cambridge University Press.
Dardano, M. & Trifone, P. 1997. La Nuova Grammatica della Lingua Italiana. Bologna,
De Bot, K. 1992. A bilingual production model: Levelt’s ‘Speaking’ model adapted. Applied
Linguistics 13 (1), 1-24.
De Houwer, A. 1990. The Acquisition of Two Languages from Birth: A Case Study.
Cambridge, Cambridge University Press.
De Houwer, A. 2005. early bilingual acquisition: Focus on morphosyntax and the Separate
Development Hypothesis. In Handbook of Bilingualism: Psycholinguistic Approaches,
J.F. Kroll & A.M.B. De Groot (eds), 30-48. Oxford, Oxford University Press.
DeKeyser, R.M. 2005. What makes learning second-language grammar difficult? A review
of issues. Language Learning 55, 1-25.
Di Biase, B. 1998. Processability Theory and the acquisition of Italian L2. Paper delivered
at the European Science Foundation Conference: The Structure of Learner Language,
Acquafredda di Maratea, Italy, September 26-October 1, 1998.
Di Biase, B. 2000. Preface. In Stages of Development for English as a Second Language, M.
Johnston, v-vii. Sydney, LARC, University of Western Sydney.
Di Biase, B. 2002 Focusing strategies in second language development: A classroom-based
study of Italian L2 in primary school. In Developing a Second Language: Acquisition,
Processing and Pedagogy of Arabic, Chinese, English, Italian, Japanese, Swedish, B. Di
Biase (ed.), 95-120. Melbourne, Language Australia.
Di Biase, B. 2005. The Topic Hypothesis in Processability Theory. Paper presented at the
5th International Symposium on Processability, Second Language Acquisition and
Bilingualism. Deakin University, Melbourne, 26-28 September, 2005.
Di Biase, B. 2007. A Processability Approach to the Acquisition of Italian as a Second
Language: Theory and Applications. PhD thesis, Australian National University,
Canberra, http://hdl.handle.net/1885/6982.
Di Biase, B. 2008. Focus-on-form and development in L2 learning. In Processing Approaches
to Second Language Development and Second Language learning, J.-U. Keßler (ed.),
197-220. Newcastle, UK, Cambridge Scholars Publishing.
Di Biase, B. & Bettoni, C. 2007. Funzioni discorsive e Processabilità in italiano. In
Imparare una Lingua. Recenti Sviluppi Teorici e Proposte Applicative, M. Chini, A. De
Meo, P. Desideri & G. Pallotti (eds), 209-233. Perugia, Guerra.
Di Biase, B. & Bettoni, C. 2015. Subject realisation in Italian L2, a cross-sectional study of
production data. In Sintassi. Atti del 46° Congresso Internazionale di Studi della Società
di Linguistica Italiana, Siena, 27-29.09.2012, C. Bruno, S. Casini, F. Gallina & R.
Siebetcheu (eds), 589-604. Roma, Bulzoni.
Di Biase, B. & Hinger, B. 2011. Topic Hypothesis in Processability Theory: The case of
Spanish. Paper presented at The 11th Symposium on Processability Approaches to
Language Acquisition (PALA). Innsbruck. http://www.slashdocs.com/mwpqry/topichypothesis-in-processability-theory-di-biase-hinger-2011.html
Di Biase, B. & Kawaguchi, S. 2002. exploring the typological plausibility of Processabilty
Theory: Language development in Italian second language and Japanese second language. Second Language Research 18 (3), 272-300.
Di Biase, B. & Kawaguchi, S. 2012. Processability Theory. In Encyclopedia of Second
Language Acquisition, P. Robinson (ed.), 512-517. London, Routledge.
Diessel, H. 2004. The Acquisition of Complex Sentences. Cambridge: Cambridge University
Dik, S.C. 1997. The Theory of Functional Grammar. Berlin & New York, Mouton de
Dimitrijević, J. 2004a. Language shift and contact-induced change: Serbian in contact
with english in Australia. Melbourne Papers in Linguistics and Applied Linguistics 2, 119.
Dimitrijević, J. 2004b. Code-switching: Structure and meaning. Facta Universitatis 3 (1),
Dimitrijević-Savić, J. 2008. Convergence and attrition: Serbian in contact with english in
Australia. Journal of Slavic Linguistics 16 (1), 57-90.
Doi, T. & Yoshioka, K. 1987. Which grammatical strutures should be taught when?
Implications on the Pienemann-Johnston model for teaching Japanese as a foreign
language. Paper presented at the 9th Hawaii Association of Teachers of Japanese.
University of Hawaii at Manoa, Honolulu.
Doi, T. & Yoshioka, K. 1990. Jyoshi no shuutoku ni okeru gengo unyoojoo no seiyaku
–Pienemann-Johnston moderu no nihongo shuutoku kenkyuu e no ooyoo.
Proceedings of 1st Conference on SLA and Teaching 1, 23-33.
Donaldson, R.P. & Kotter, M. 1999. Language learning in cyberspace: Teleporting the
classroom into the target culture. CALICO Journal 16 (4), 531-557.
Doughty, C.J. 1991. Second language instruction does make a difference: evidence from
an empirical study of eSL relativization. Studies in Second Lnaguage Acquisition 13 (4),
Doughty, C.J. & Long, M.H. (eds) 2003. The Handbook of Second Language Acquisition.
Oxford, Blackwell.
Dryer, M.S. 1989. Large linguistic areas and language sampling. Studies in Language 13,
Dryer, M.S. 2013. Determining Dominant Word Order. In The World Atlas of Language
Structures Online, M.S. Dryer & M. Haspelmath (eds), ch. 26. Leipzig, Max Planck
Institute for evolutionary Anthropology. Available online at <http://wals.info/chapter/26>, accessed on 19-03-2015.
Dulay, H. & Burt, M. 1974. Natural sequences in child second language acquisition.
Language Learning 24 (1), 37-53.
Dyson, B. 2009. Processability Theory and the role of morphology in english as a second
language development: a longitudinal study. Second Language Research 25 (3), 355376.
eigsti, I.-M., Bennetto, L. & Dadlani, M. 2007. Beyond pragmatics: Morphosyntactic
development in autism. Journal of Autism and Developmental Disorders 37 (6), 10071023.
ellis, N. & Collins, L. 2009. Input and second language acquisition: The roles of frequency, form, and function. The Modern Language Journal 93, 329-335.
ellis, R. 1989. Are classroom and naturalistic acquisition the same? A study of the classroom acquisition of German word order rules. Studies in Second Language Acquisition
11, 305-328.
ellis, R. 1994. The Study of Second Language Acquisition. Oxford, Oxford University Press.
ellis, R. 2008. The Study of Second Language Acquisition. Second edition. Oxford, Oxford
University Press.
erteschik-Shir, N. 2007. Information Structure. The Syntax-Discourse Interface. New York,
Oxford University Press.
Falk, Y.N. 2001. Lexical-Functional Grammar: An Introduction to Parallel Constraint-based
Syntax. Stanford, CA, CSLI Publication.
Falk, Y.N. 2006. Subjects and Universal Grammar: An Explanatory Theory. Cambridge,
Cambridge University Press.
Farley, A.P. & McCollam, K. 2004. Learner readiness and L2 production in Spanish:
Processability Theory on trial. Estudios de Lingcüística Aplicada 40, 47-69.
Fava, e. 1995. Tipi di frasi principali. Il tipo interrogativo. In Grande Grammatica Italiana
di Consultazione, L. Renzi, G. Salvi & A. Cardinaletti (eds), vol. 3, 70-127. Bologna,
Il Mulino.
Filipek, P.A., Accardo, P.J., Baranek, G.T., Cook, e.H., Dawson, G. & Gordon, B. 1999.
The screening and diagnosis of autistic spectrum disorders. Journal of Autism and
Developmental Disorders 29 (6), 439-484.
Finegan, e. 2011. english. In The Major Languages of the World, B. Comrie (ed.), 59-85.
London & New York, Routledge.
Franck, J., Vigliocco, G. & Nicol, J. 2002. Subject-verb agreement errors in French and
english: The role of syntactic hierarchy. Language and Cognitive Processes 17, 371-404.
Franks, S. 1994. Parametric properties of numeral phrases in Slavic. Natural Language and
Linguistic Theory 12, 597-674.
Frey, W. 2004. A medial topic position for German. Linguistische Berichte 198, 153-190.
Gass, S.M. 1994. The reliability of second-language grammaticality judgments. In Research
Methodology in Second Language Acquisition, e.e. Tarone, S.M. Gass & A.D. Cohen
(eds), 302-322. Hillsdale, NJ, erlbaum.
Gass, S.M & Mackey, A. 2012. The Routledge Handbook of Second Language Acquisition.
New York, Routledge.
Gass, S.M. & Selinker, L. 2001. Second Language Acquisition. An Introductory Course. 2nd
edition. Hillsdale, NJ, erlbaum.
Givón, T. 1983. Topic continuity and word-order pragmatics in Ute. In Topic Continuity
in Discourse, T. Givón (ed.), 145-195. Amsterdam, Benjamins.
Givón, T. 2001. Syntax. Vols 1-2. Amsterdam, Benjamins.
Glahn, e., Håkansson, G., Hammarberg, B., Holmen, A., Hvenekilde, A. & Lund, K.
2001. Processability in Scandinavian second language acquisition. Studies in Second
Language Acquisition 25, 389-416.
Green, J.N. 2011. Spanish. In The World’s Major Languages, B. Comrie (ed.), 2nd ed., 197216. London & New York, Routledge.
Greenberg, J.H. 1963. Some universals of grammar with particular reference to the order
of meaningful elements. In Universals of Language, J.H. Greenberg (ed.), 58-90.
Cambridge, MA, MIT Press.
Grosjean, F. & Li P. 2013. The Psycholinguistics of Bilingualism. Malden, MA, WileyBlackwell.
Guijarro-Fuentes, P. 2011. Feature composition in Differential Object Marking.
EUROSLA Yearbook 11, L. Roberts, G. Pallotti & C. Bettoni (eds), 138-164.
Amsterdam, Benjamins.
Guijarro-Fuentes, P. 2012. The acquisition of interpretable features in L2 Spanish: Personal
a. Bilingualism: Language and Cognition 15, 701-720.
Haberzettl, S. 2005. Der Erwerb der Verbstellungsregeln in der Zweitsprache Deutsch durch
Kinder mit Russischer und Türkischer Muttersprache. Tübingen, Max Niemeyer.
Håkansson, G. 2001. Tense morphology and Verb-second in Swedish L1 children, L2 children and children with SLI. Bilingualism: Language and Cognition 4, 85-99.
Håkansson, G. 2005. Similarities and differences in L1 and L2 development: Opening up
the perspective: Including SLI. In Cross-linguistic Aspects of Processability Theory, M.
Pienemann (ed.), 179-198. Amsterdam, Benjamins.
Håkansson, G. & Norrby, C. 2007. Processability Theory applied to written and oral
Swedish. In Second Language Acquisition Research: Theory-construction and Testing, F.
Mansouri (ed.), 81-94. Newcastle, UK, Cambridge Scholar Publishing.
Håkansson, G., Pienemann, M. & Sayehli, S. 2002. Transfer and typological proximity in
the context of second language processing. Second Language Acquisition Research 18
(3), 250-273.
Håkansson, G., Salameh, e. & Nettelbladt, U. 2003. Measuring language development in
bilingual children: Swedish-Arabic children with and without language impairment.
Linguistics 41, 255-288.
Hambly, C., & Fombonne, e. 2012. The impact of bilingual environments on language
development in children with Autism Spectrum Disorders. Journal of Autism and
Developmental Disorders 42 (7), 1342-1352. doi: 10.1007/s10803-011-1365-z.
Hambly, C., & Fombonne, e. 2014. Factors influencing bilingual expressive vocabulary
size in children with autism spectrum disorders. Research in Autism Spectrum Disorders
8(9), 1079-1089. doi: http://dx.doi.org/10.1016/j.rasd.2014.05.013.
Hammond, L. 2005. Serbian: An Essential Grammar. London, Routledge.
Happé, F. & Frith, U. 2006. The weak coherence account, Detail-focused cognitive style
in autism spectrum disorders. Journal of Autism and Developmental Disorders 36 (1),
Harada, S. 1971. Ga-no conversion and idiolectal variations in Japanese. Gengo Kenkyuu
60, 25-38.
Hatch, e. & Lazaraton, A. 1991. The research manual: Design and Statistics for Applied
Linguistics. Boston, MA, Heinle & Heinle.
Hayashi, A., Tomlin, R.S. & Yokota, T. 2002. Attention detection and Japanese.
Unpublished manuscript.
Hinger, B. 2011. Sprache lehren – Sprache überprüen – Sprache erwerben: empirische
einsichten in den schulischen Spanischunterricht – eine Fallstudie. Habilitation thesis, University of Innsbruck, Innsbruck.
Hopp, H. 2007. Cross-linguistic differences at the syntax-discourse interface in off- and online L2 performance. In Proceedings of the 2nd Conference on Generative Approaches to
Language Acquisition North America (GALANA), A. Belikova & al. (eds), 147-158.
Somerville, MA, Cascadilla Proceedings Project.
Hopper, J.P. & Thompson, S.A. 1980. Transitivity in grammar and discourse. Language 56,
Horvath, J. 1986. FOCUS in the Theory of Grammar and Syntax in Hungarian. Dordrecht,
Howlin, P. 1984. The acquisition of grammatical morphemes in autistic children: A critique and replication of the findings of Bartolucci, Pierce, and Streiner, 1980. Journal
of Autism and Developmental Disorders 14 (2), 127-136.
Howlin, P., Goode, S., Hutton, J. & Rutter, M. 2004. Adult outcome for children with
autism. Journal of Child Psychology and Psychiatry 45 (2), 212-229.
Hualde, J.I., Olarrea, A. & escobar, A.M. 2001. Introducción a la Lingüística Hispánica.
Cambridge, Cambridge University Press.
Hudry, K., Chandler, S., Bedford, R., Pasco, G., Gliga, T., elsabbagh, M., … Charman, T.
2014. early language profiles in infants at high-risk for Autism Spectrum Disorders.
Journal of Autism and Developmental Disorders 44(1), 154-167. doi: 10.1007/s10803013-1861-4.
Huter, K. 1996. Atarashii no Kuruma and other old friends – the acquisition of Japanese
syntax. Australian Review of Applied Linguistics 19, 39-60.
Huter, K. 1997. “Onnanohito wa duressu desu. That’s why the lady is a dress”, Developmental
stages in Japanese second language acquisition. Australian Studies in Language
Acquisition 6, 1-39.
Ishikawa, A. 1985. Complex Predicates and Lexical Operations in Japanese. Doctoral dissertation, Stanford University, CA.
Itani-Adams, Y. 2007. Lexical and grammatical development in Japanese-english bilingual
first language acquisition. In Second Language Acquisition Research: Theory
Construction and Testing, Mansouri, F. (ed.), 173-198. Newcastle, UK., Cambridge
Itani-Adams, Y. 2008. One Child, Two Languages: Bilingual First Language Acquisition in
Japanese and English. PhD thesis, University of Western Sydney, Sydney.
Itani-Adams, Y. 2009. Development of discourse functions in Japanese and english bilingual first language acquisition. In Research in Second Language Acquisition: Empirical
Evidence Across Languages, J.-U. Keßler & D. Keatinge (eds), 41-66. Newcastle, UK,
Cambridge Scholars Publishing.
Itani-Adams, Y. 2011. Bilingual first language acquisition. In Studying Processability Theory,
M. Pienemann & J.U. Ke ler (eds), 121-130. Amsterdam, Benjamins.
Itani-Adams, Y. 2013. One Child and Two Languages: Acquisition of Japanese and English as
Bilingual First Languages. München, Lincom.
Iwasaki, J. 2008. Acquiring Japanese as a second language (JSL) in a naturalistic context: A
longitudinal study of a young child from a Processability Theory (PT) perspective. In
Second Language Acquisition and the Younger Learner: Child’s Play? J. Philp, R. Oliver
& A. Mackey (eds), 231-253. Amsterdam, Benjamins.
Iwasaki, J. 2013. A Processability Theory (PT)-based analysis of the acquisition of Japanese
morphology and syntax: A case of an intensive adult learner. Second Language 12, 2142.
Iwasaki, J. & Oliver, R. 2003. Chat-line interaction and negative feedback. Australian
Review of Applied Linguistics 17, 60-73.
Jackendoff, R.S. 1972. Semantic Interpretation in Generative Grammar. Cambridge, MA,
MIT Press.
Jakobson, R. 1959. Boas’ view of grammatical meaning. American Anthropologist 61, 139145.
Jakobson, R. 1971. Beitrag zur allgemeinen Kasuslehre. In Selected Writings II, R. Jakobson
(ed.), 23-71. The Hague, Mouton.
Jansen, L. 2008. Acquisition of German word order in tutored learners: A cross-sectional
study in a wider theoretical context. Language Learning 58, 185-231.
Jarrold, C., Boucher, J. & Russell, J. 1997. Language profiles in children with autism:
Theoretical and methodological implications. Autism 1 (1), 57-76.
Jellinek, M.S., Toppelberg, C.O., Snow, C.e., & Teger-Flusberg, H. 1999. Severe
developmental disorders and bilingualism. Journal of the American Academy of Child
& Adolescent Psychiatry 38(9), 1197-1199. doi: http://dx.doi.org/10.1097/00004583199909000-00027
Jia, G. & Fuse, A. 2007. Acquisition of english grammatical morphology by native
Mandarin-speaking children and adolescents: Age-related differences. Journal of
Speech, Language and Hearing Research 50, 1280-1299.
Johnston, M. 1985a. Syntactic and morphological progressions in learner English. Canberra,
Department of Immigration and ethnic Affairs.
Johnston, M. 1985b. SLA Research in the Adult Migrant education Program. Prospect: A
Journal of Adult Migrant Education 1 (1), 19-46.
Johnston, M. 1987. Understanding learner language. In Applying Second Language
Acquisition Research, D. Nunan (ed.), 2-44. Adelaide, National Curriculum Research
Centre, Adult Migrant education Program.
Johnston, M. 1994. Second Language Acquisition: A Classroom Perspective. Sydney, NLLIA
Language Acquisition Research Centre, University of Western Sydney.
Johnston, M. 1995. Stages of Acquisition of Spanish as a Second Language. Australian
Studies in Language Acquisition 4, 1-28.
Johnston, M. 1997. Development and Variation in Learner Language. PhD thesis,
Australian National University, Canberra.
Johnston, M. 2000. Stages of development for English as a Second Language. Melbourne,
Language Australia.
Jordan, G. 2004. Theory Construction in Second Language Acquisition. Amsterdam,
Kageyama, T. 1999. Word formation. In The Handbook of Japanese Linguistics, N.
Tsujimura (ed.), 297-325. Malden, MA, Blackwell.
Kahnemann, D. 1973. Attention and Effort. englewood Cliffs, NJ, Prentice-Hall.
Kallestinova, e.D. 2007. Aspects of word order in Russian. PhD dissertation, University of
Iowa, IA.
Kanner, L. 1943. Autistic disturbances of affective contact. Nervous Child 2 (2), 217-250.
Kanner, L. 1971. Follow-up study of eleven autistic children originally reported in 1943.
Journal of Autism and Developmental Disorders 1 (2), 119-145.
Kaplan, R.M. 1995. The formal architecture of lexical-functional grammar. In Formal Issues
in Lexical-Functional Grammar, M. Dalrymple, R.M. Kaplan, J.T. Maxwell III & A.
Zaenen (eds), 7-27. Stanford University, CSLI.
Kaplan, R.M. & Bresnan, J. 1982. Lexical-Functional Grammar: A formal system for
grammatical representation. In The Mental Representation of Grammatical Relations, J.
Bresnan (ed.), 173-281. Cambridge, MA, MIT Press.
Kay-Raining Bird, e., Lamond, e., & Holden, J. 2012. Survey of bilingualism in autism
spectrum disorders. International Journal of Language & Communication Disorders
47(1), 52-64. doi: 10.1111/j.1460-6984.2011.00071.x.
Kawaguchi, S. 2000. Acquisition of Japanese verbal morphology: Applying Processability
Theory to Japanese. Studia linguistica 54 (2), 238-248.
Kawaguchi, S. 2002. Grammatical development in learners of Japanese as a second language. In Developing a second language: Acquisition, processing and pedagogy of Arabic,
Chinese, English, Italian, Japanese, Swedish, B. Di Biase (ed.), 17-28. Melbourne,
Language Australia.
Kawaguchi, S. 2005. Argument structure and syntactic development in Japanese as a second language. In Cross-Linguistic Aspects of Processability Theory, M. Pienemann (ed.),
253-299. Amsterdam, Benjamins.
Kawaguchi, S. 2007. Lexical Mapping Theory and Processability Theory: A case study in
Japanese. In Second Language Acquisition Research: Theory-Construction and Testing, F.
Mansouri (ed.), 39-90. Newcastle, UK, Cambridge Scholars Publishing.
Kawaguchi, S. 2008. Language typology and Processability Theory. In Processing Approaches
to Second Language Development and Second Language learning, J.-U. Keßler (ed.), 89112. Newcastle, UK, Cambridge Scholars Publishing.
Kawaguchi, S. 2009a. Acquiring causative constructions in Japanese as a second language,
The Journal of Japanese Studies 29 (2), 273-291.
Kawaguchi, S. 2009b. Acquisition of non-canonical order in Japanese as a second language:
The case of causative structure. In Research in Second Language Acquisition: Empirical
Evidence Across Languages, D. Keatinge & J.-U. Keßler (eds), 213-239. Newcastle,
UK, Cambridge Scholars Publishing.
Kawaguchi, S. 2010. Learning Japanese as a Second Language. A Processability Perspective.
Amherst, NY, Cambria Press.
Kawaguchi, S. 2011. Japanese as a second language: A test case for the typological plausibility of PT. In Studying Processability Theory. An Introductory Textbook, M. Pienemann
& J.-U. Keßler (eds), 99-105. Amsterdam, Benjamins.
Kawaguchi, S. 2013. The relationship between lexical and syntactic development in
english as a second language. In Language Acquisition and Use in Multilingual
Contexts: Theory and Practice, A.F. Mattisson & C. Norrby (eds), 92-106. Lund,
Sweden, University of Lund, Traveaux de L’Institut de Linguistique De Lund.
Kawaguchi, S. (in press). Question constructions, argument mapping, and vocabulary
development in english L2 by Japanese speakers: A cross-sectional study. In
Developing and Assessing Second Language Grammars across Languages. J.-U. Keßler, A.
Lenzig & M. Liebner (eds). Amsterdam, Benjamins.
Kawaguchi, S. & Di Biase, B. 2005. Second language development at the syntax-pragmatics interface. Paper presented at the Seventh Annual International Conference of the
Japanese Society of Language Sciences, JSLS. Sophia University, Tokyo, June 25-26, 2005.
Kawaguchi, S., & Di Biase, B. 2012. Acquiring procedural skills in L2: Processability
Theory and skill acquisition. Studies in Language Sciences 11, 70-99.
Keatinge, D. 2008. Key issues in designing tasks for data elicitation purposes. Paper presented at The 8th Symposium on Processability Approaches to Language Acquisition
(PALA). Verona, 15-16 September 2008.
Keatinge, D. & Keßler J.-U. 2009. The acquisition of the passive voice in L2 english:
Perception and production. In Research in Second Language Acquisition: Empirical
Evidence Across Languages, D. Keatinge & J.-U. Keßler (eds), 67-92. Newcastle, UK,
Cambridge Scholars Publishing.
Keenan, e.L. 1979. On surface form and logical form. Studies in Linguistic Sciences 8,
Keenan, e.O. & Comrie, B. 1977. Noun phrase accessibility and Universal Grammar.
Linguistic Inquiry 8, 63-99.
Kempe, V. & MacWhinney, B. 1998. The acquisition of case-marking by adult learners of
Russian and German. Paper 218, Department of Psychology. Pittsburgh, PA,
Carnegie Mellon University.
Kempen, G. 1998. Comparing and explaining the trajectories of first and second language
acquisition: In search of the right mix of psychological and linguistic factors.
Bilingualism: Language and Cognition 1, 29-30.
Kempen, G. & Hoenkamp, e. 1987. An incremental procedural grammar for sentence formulation. Cognitive Science 1 (1), 201-258.
Keßler, J.-U. & Pienemann, M. 2011. Research methodology. In Studying Processability
Theory. An Introductory Textbook. M. Pienemann & J.-U. Keßler (eds), 84-96.
Amsterdam, Benjamins.
Kim, J.-B. & Sells, P. 2008. English Syntax: An Introduction. Stanford, CA, CSLI
Kim, S.H., Paul, R., Tager-Flusberg, H., & Lord, C. 2014. Language and Communication
in Autism. In Handbook of Autism and Pervasive Developmental Disorders, F.R.
Volkmar, S.J. Rogers, P. Rhea & K.A. Pelphrey (eds), 4th ed., vol. 1, 230-262.
Hoboken, NJ, Wiley & Sons.
King, T.H. 1995. Configuring Topic and Focus in Russian. Stanford, CA, CSLI Publications.
Kjelgaard, M.M. & Tager-Flusberg, H. 2001. An investigation of language impairment in
autism: Implications for genetic subgroups. Language and Cognitive Processes 16(2),
Klein, W. & Carroll, M. 1992. The acquisition of German. In Utterance Structure.
Developing Grammars Again, W. Klein & C. Perdue (eds), 123-188. Amsterdam,
Kormos, J. 2006. Speech Production and Second Language Acquisition. Mahwah, NJ,
Kotter, M. 2003. Negotiation of meaning and codeswitching in online tandems. Language
Learning and Technology 7 (2), 145-172.
Krashen, S. 1976. Formal and informal linguistic environments in language learning and
language acquisition. TESOL Quarterly 10 (2), 157-168.
Krashen, S. 1977. Some issues relating to the monitor model. In Teaching and Learning
English as a Second Language: Trends in Research and Practice: Selected Papers from the
Eleventh Annual Convention of Teachers of English to Speakers of Other Languages,
Miami, Florida, April 26 - May 1, 1977, 144-158. Washington, DC, Teachers of
english to Speakers of Other Languages.
Krashen, S. 1979. A Response to McLaughlin, “The Monitor Model: Some
Methodological Considerations”. Language Learning 29, 151-67.
Krashen, S. 1982. Principles and Practice in Second Language Acquisition. Oxford, Oxford
University Press.
Krashen, S. 1985. The Input Hypothesis: Issues and Implications. New York, Longman.
Kroeger, P.R. 1993. Phrase Structure and Grammatical Relations in Tagalog. Stanford, CA,
CSLI Publication.
Kroeger, P.R. 2004. Analysing Syntax: A Lexical-functional Approach. Cambridge,
Cambridge University Press.
Kroeger, P.R. 2005. Analysing Grammar: An Introduction. Cambridge, Cambridge
University Press.
Kroll, J.F. & de Groot, A.M.B. (eds) 2005. Handbook of Bilingualism: Psychological
Approaches. Oxford and New York, Oxford University Press.
Kuno, S. 1973. The Structure of the Japanese Language. Cambridge, MA, MIT Press.
Lai, C. & Zhao, Y. 2006. Noticing and text-based chat. Language Learning and Technology
10 (3), 102-120.
Laca, B. 2006. el objeto directo. La marcación preposicional. In Sintaxis historica de la lengua espanola, C. Company Company (ed.), vol. 1, 421-75. Distrito Federal,
Universidad Nacional Autónoma de México.
Lambrecht, K. 1988. There was a farmer had a dog: Syntactic amalgams revisited.
Proceedings of the 14th Annual Meeting of the Berkeley Linguistics Society 14, 319-339.
Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus and Mental
Representation of Discourse Referents. Cambridge, Cambridge University Press.
Lardiere, D. 1998. Case and tense in the “fossilized” steady state. Second Language Research
14, 1-26.
Lardiere, D. 2008. Feature-assembly in second language acquisition. In The Role of Formal
Features in Second Language Acquisition, J. Liceras, H. Zobl & H. Goodluck (eds),
106-140. Mahwah, NJ, erlbaum.
Lardiere, D. 2009. Some thoughts on the contrastive analysis of features in second language acquisition. Second Language Research 25, 173-227.
Larsen Freeman, D. & Long, M.H. 1991. An Introduction to Second Language Acquisition
Research. London, Longman.
Lee, H. 2001. Markedness and word order freezing. In Formal and Empirical Issues in
Optimality Theoretic Syntax, P. Sells (ed.), 63-127. Stanford, CA, CSLI Publications.
Lee, L. 2006. A study of native and nonnative speakers’ feedback and responses in SpanishAmerican networked collaborative interaction. In Internet-mediated Intercultural
Foreign Language Education, J. Belz & S. Thorne (eds), 147-176. Boston, Thomson
Leonetti, M. 2004. Specificity and object marking: The case of Spanish a. Catalan Journal
of Linguistics 3, 75-114.
Lepschy, A.L. & Lepschy, G. 1981. La Lingua Italiana: Storia, Varietà dell’Uso, Grammatica.
Milano, Bompiani.
Levelt, W.J.M. 1989. Speaking: From Intention to Articulation. Cambridge, MA, MIT Press.
Levelt, W.J.M., Roelofs, A. & Meyer, A.S. 1999. A theory of lexical access in speech production. Behavioral and Brain Sciences 22, 1-75.
Levin, B. & Rappaport Hovav, M. 2005. Argument Realisation. Cambridge, Cambridge
University Press.
Levy, M. & Stockwell, G. 2006. CALL Dimensions: Options and Issues in Computer-assisted
Language Learning. Mahwah, NJ, erlbaum.
Lewis,T. & Walker, L. (eds) 2003. Autonomous Language Learning in Tandem. Sheffield,
Academy electronic Publications.
Li, C.N. & Thompson, S.A. 1981. Mandarin Chinese: A Functional Reference Grammar.
Berkeley, CA, University of California.
Li Wei (ed.) 2006. The Bilingualism Reader. London, Routledge.
Loewen, S. 2009. Grammaticality judgment tests and the measure of implicit and explicit
L2 knowledge. In Implicit and Explicit Knowledge in Second language Learning and
Teaching. R. ellis, S. Loewen, R. erlam, J. Phelp, C. elder & H. Reinders (eds), 6593. Clevedon, Multilingual Matters.
Long, M.H. 1996. The role of linguistic environment in second language acquisition. In
Handbook of language acquisition; Vol. 2, Second Language Acquisition, W.C. Ritchie &
T.K. Bhatia (eds), 413-468. New York, Academic Press.
Long, M.H. 2015. Second Language Acquisition and Task-based Language Teaching.
Chichester, West Sussex, Wiley-Blackwell.
Lord, C. & Venter, A. 1992. Outcome and follow-up studies of High-Functioning Autistic
individuals. In High-functioning Individuals with Autism, e. Schopler & G.B. Mesibov
(eds), 187-200. New York, Plenum Press.
Lujan, M. 1996. Clitic-doubling and the acquisition of agreement in Spanish. In
Perspectives on Spanish Linguistics, J. Gutiérrez-Rexach & L. Silva Villar (eds), 119138. Los Angeles, UCLA, Department of Linguistics.
Macaro, e. (ed.) 2013. The Bloomsbury Companion to Second Language Acquisition. London
and New York, Bloomsbury.
Mackey, A. 1999. Input, interaction, and second language development: An empirical
study of question formation in eSL. Studies in Second Language Acquisition 21, 557587.
Mackey, A., Pienemann, M. & Thornton, I. 1991. Rapid Profile: A second language
screening procedure. Language and Language Education 1 (1), 61-82.
MacWhinney, B. 1997. Second language acquisition and the Competition Model. In
Tutorials in Bilingualism: Psycholinguistic Perspectives, A.M.B. de Groot & J. Kroll
(eds), 113-142. Mahwah, NJ, erlbaum.
Maiden, M. 1995. A Linguistic History of Italian. London, Longman.
Mansouri, F. 1995. The acquisition of Subject-Verb agreement in Arabic as a second language. Australian Review of Applied Linguistics 18, 65-85.
Mansouri, F. 1997. From emergence to acquisition: Development issues in agreement
marking among Australian learners of Arabic. The Australian Review of Applied
Linguistics 20, 83-104.
Mansouri, F. 1999. The Acquistion of Arabic as a Second Languagage: From Theory to Practice.
ASLA 7. Sydney, Language Acquistion Research Centre, Unversity of Western
Mansouri, F. 2002. exploring the interface between syntax and morphology in second language development. In Developing a Second Language: Acquisition, Processing and
Pedagogy of Arabic, Chinese, English, Italian, Japanese, Swedish, B. Di Biase (ed.), 5972. Melbourne, Language Australia.
Mansouri, F. 2005. Agreement morphology in Arabic as a second language: Typological
features and their processing implications. In Cross-Linguistic Aspects of Processability
Theory, M. Pienemann (ed.), 117-154. Amsterdam, Benjamins.
Mansouri, F. & Håkansson, G. 2007. Conceptualizing intra-stage sequencing in the learner language. In Second Language Acquisition Research: Theory Construction and Testing,
F. Mansouri (ed.), 95-117. Newcastle, UK, Cambridge Scholars.
Matsumoto, Y. 1996. Complex Predicates in Japanese. A Syntactic and Semantic Study of the
Notion of ‘Word’. Stanford, CA, CSLI Publications, and Tokyo, Kurosio Publishers.
Mayer, M. 1969. Frog, Where are You? New York, Dial Books for Young Readers.
McLaughlin, B. 1978. The monitor model: Some methodological considerations.
Language Learning 28, 309-332.
McLaughlin, B. 1980. Theory and research in second _language learning: An emerging
paradigm. Language Learning 30, 331-350.
Medojević, L. 2014. The Effect of First Year of Schooling on Bilingual Language Acquisition:
A study of Second and Third Generation Serbian-Australian 5-year-old Bilingual
Children from Processability Perceptive. PhD thesis, University of Western Sydney,
Meisel, J. 1989. early differentiation of languages. In Bilingualism Across Lifespan: Aspect of
Acquisition, Maturity and Loss, K. Hyltenstam & L. Odler (eds), 13-40. Cambridge,
Cambridge University Press.
Meisel, J., Clahsen, H. & Pienemann, M. 1981. On determining developmental sequences
in natural second language acquisition. Studies in Second Language Acquisition 3 (2),
Mesibov, G.B., Shea, V. & Adams, L.W. 2001. Understanding Asperger Syndrome and High
Functioning Autism. New York, Kluwer Academic/Plenum Publishers.
Montrul, S. 2004. The Acquisition of Spanish. Morphosyntactic Development in Monolingual and Bilingual L1 Acquisition and Adult L2 Acquisition. Amsterdam,
Montrul, S. 2008. Incomplete Acquisition in Bilingualism: Re-examining the Age Factor.
Amsterdam, Benjamins.
Montrul, S. & Bowles, M. 2008. Negative evidence in instructed Heritage Language acquisition: A preliminary study of Differential Object Marking. In Selected Proceedings of
the 2007 Second Language Research Forum, M. Bowles & al. (eds), 252-262.
Somerville, MA, Cascadilla Proceeding Project.
Montrul, S. & Bowles, M. 2009. Back to basics: Incomplete knowledge of Differential
Object Marking in Spanish heritage speakers. Bilingualism: Language and Cognition
12 (3), 363-383.
Montrul, S. & Sánchez-Walker, N. 2013. Differential Object Marking in child and adult
Spanish Heritage Speakers. Language Acquisition 20 (2), 109-132.
Müller, G. 1988. Zur Analyse subjektloser Konstruktionen in der Rektions-BindungsTheorie. That-trace-effekte und pro-Lizensierung im Russischen. Seminararbeit,
Frankfurt am Main, Goethe Universität.
Mycock, L. 2007. Constituent question formation and focus: A new typological perspective. Transactions of the Philological Society 105 (2), 192-251.
Næss, Å. 2007. Prototypical Transitivity. Amsterdam, Benjamins.
Nespor, M. & Vogel, I. 2007. Prosodic Phonology. The Hague, Mouton de Gruyter.
Nichols, J. 1986. Head-marking and dependent-marking grammar. Language 66, 56-119.
Nicoladis, e., Song, J. & Marentette, P. 2012. Do young bilinguals acquire past tense morphology like monolinguals, only later? evidence from French-english and Chineseenglish bilinguals. Applied Psycholinguistics 33, 457-479.
Noda, H., Sakota, K., Shibuya, K. & Kobayashi, N. 2001. Nihongo Gakushūsha no Bunpō
Shūtoku (Second Language Acquisition of Japanese Grammar). Tokyo, Taishukan.
Nordlinger, R. 1998. A Grammar of Wambaya. Canberra, Pacific Linguistics.
Nuzzo, e. 2012. Imparare la costruzione passiva in italiano L2. Studi Italiani di Linguistica
Teorica e Applicata 41 (1), 67-84.
Nuzzo, e. & Bettoni, C. 2011. Insegnamento mirato in varietà avanzate d’italiano L2: una
sperimentazione didattica nel quadro della Teoria della Processabilità. Rivista di
Psicolinguistica Applicata 11 (1-2), 33-47.
Ohashi, J.K., Mirenda, P., Marinova-Todd, S., Hambly, C., Fombonne, e., Szatmari, P., .
. . Thompson, A. 2012. Comparing early language development in monolingual- and
bilingual- exposed young children with autism spectrum disorders. Research in Autism
O’Rourke, B. 2005. Form-focused Interaction in online tandem learning. CALICO Journal
22 (3), 433-466.
Ortega, L. 2009. Understanding Second Language Acquisition. London, Arnold-Hodder
Ozonoff, S., Pennington, B.F. & Rogers, S.J. 1991. executive function deficits in highfunctioning autistic individuals: Relationship to theory of mind. Journal of Child
Psychology and Psychiatry 32 (7), 1081-1105.
Ozonoff, S., Young, G.S., Carter, A., Messinger, D., Yirmiya, N., Zwaigenbaum, L., …
Stone, W.L. 2011. Recurrence risk for Autism Spectrum Disorders: A baby siblings
research consortium study. Pediatrics 128 (3), e488-e495. doi: 10.1542/peds.20102825.
Pallotti, G. 2007. An operational definition of the emergence criterion. Applied Linguistics
28, 361-382.
Paramskas, P.C. 1999. The shape of computer-mediated communication. In Computer
Assisted Language Learning (CALL): Media, Design and Applications, K. Cameron
(ed.), 13-34. Lisse, The Netherlands, Swets & Zeitlinger.
Payne, S. & Whitney, P. 2002. Developing L2 oral proficiency through synchronous CMC:
Output, working memory, and interlanguage development. CALICO Journal 20 (1),
Perlmutter, D. & Moore, J. 2002. Language-internal explanation: The distribution of
Russian impersonals. Language 78, 619-650.
Petersen, J., Marinova-Todd, S., & Mirenda, P. 2012. Brief report, An exploratory study of
lexical skills in bilingual children with Autism Spectrum Disorder. Journal of Autism
and Developmental Disorders 42 (7), 1499-1503. doi: 10.1007/s10803-011-1366-y
Pickering, M.J., Branigan, H.P. & McLean, J.F. 2002. Constituent structure is formulated
in one stage. Journal of Memory and Language 46, 586-605.
Pienemann, M. 1981. Der Zweitspracherwerb ausländischer Arbeiterkinder. Bonn, Bouvier.
Pienemann, M. 1984. Psychological constraints on the teachability of languages. Studies in
Second Language Acquisition 6, 186-214.
Pienemann, M. 1985. Learnability and syllabus construction. In Modelling and Assessing
Second Language Acquisition, K. Hyltenstam & M. Pienemann (eds), 23-76.
Clevedon, Multilingual Matters.
Pienemann, M. 1987. Determining the effect of instruction on L2 speech processing.
Australian Review of Applied Linguistics 10, 83-113.
Pienemann, M. 1989. Is language teachable? Psycholinguistic experiments and hypotheses.
Applied Linguistics 10, 52-79.
Pienemann, M. 1998. Language Processing and Second Language Development: Processability
Theory. Amsterdam, Benjamins.
Pienemann, M. 2002. The procedural skill hypothesis for SLA. In An Integrated View of
Language Development, P. Burmeister, T. Piske & A. Rohde (eds), 43-56. Trier,
Wissenshftlicher Verlag.
Pienemann, M. 2003. Language processing capacity. In The Handbook of Second Language
Acquisition, C.J. Doughty & M.H. Long (eds), 679-714. Oxford, Blackwell.
Pienemann, M. 2005, An introduction to Processability Theory. In Cross-linguistic Aspects
of Processability Theory, M. Pieneamnn (ed.), 1-60. Amsterdam, Benjamins.
Pienemann, M. 2007. Processability Theory. In Theories in Second Language Acquisition, B.
VanPatten & J. Williams (eds), 137-154. Mahwah, NJ, erlbaum.
Pienemann, M. 2011a. Develpmental schedules. In Studying Processability Theory. An
Introductory Textbook. M. Pienemann & J.-U. Keßler (eds), 3-11. Amsterdam,
Pienemann, M. 2011b. explaining develpmental schedules. In Studying Processability
Theory. An Introductory Textbook. M. Pienemann & J.-U. Keßler (eds), 50-63.
Amsterdam, Benjamins.
Pienemann, M. 2011c. The psycholinguistic basis of PT. In Studying Processability Theory.
An Introductory Textbook. M. Pienemann & J.-U. Keßler (eds), 27-49. Amsterdam,
Pienemann, M., Di Biase, B. & Kawaguchi, S. 2005. extending Processability Theory, In
Cross-linguistic Aspects of Processability Theory, M. Pienemann (ed.), 199-251.
Amsterdam, Benjamins.
Pienemann, M., Di Biase, B., Kawaguchi, S. & Håkansson, G. 2005. Processing constraints on L1 transfer. In Handbook of Bilingualism: Psychological Approaches, J.F. Kroll
& A.M.B. de Groot (eds), 128-153. Oxford & New York, Oxford University Press.
Pienemann, M. & Håkanssson, G. 1999. A unified approach towards the development of
Swedish as L2: A processability account. Studies in Second Language Acquisition 21,
Pienemann, M. & Håkansson, G. 2007. Full transfer vs. developmentally moderated transfer. Second Language Research 23 (4), 485-494.
Pienemann, M. & Johnston, M. 1986. An acquisition-based procedure for second language assessment. Australian Review of Applied Linguistics 9, 92-122.
Pienemann, M. & Johnston, M. 1987. Factor influencing the development of language proficiency. In Applying second language acquisition research, D. Nunan (ed.), 45141. Adelaide, National Curriculum Research Centre, Adult Migrant education
Pienemann, M., Johnston, M. & Brindley, G. 1988. Constructing an acquisition-based
procedure for second language assessment. Studies in Second Language Acquisition 10,
Pienemann, M. & Keßler, J.-U. 2007. Measuring bilingualism. In Handbook of Applied
Linguistics; Vol. 5, Multilingualism, P. Auer & Li Wei (eds), 247-274. Berlin/New
York, Mouton/de Gruyter.
Pienemann, M. & Keßler, J.-U. 2010. Rapid Profile: Linguistic Profiling. http://groups.unipaderborn.de/rapidprofile/ (accessed 10.01.2010).
Pienemann, M. & Keßler, J.-U. (eds) 2011. Studying Processability Theory. An Introductory
Textbook. Amsterdam, Benjamins.
Pienemman, M. & Kessler, J-U. 2012. Processability Theory. In The Routledge Handbook
of Second Language Acquisition, S.M. Gass & A. Mackey (eds), 228-246. New York,
Pienemann, M., Keßler, J.-U. & Lenzing, A. 2013. Developmentally Moderated
Transfer and the role of L2 in L3 acquisition. In Language Acquisition and Use in
Multilingual Contexts: Theory and Practice, A.F. Mattisson & C. Norrby (eds), 142159. Lund, Sweden, University of Lund, Traveaux de L’Institut de Linguistique De
Pienemann, M. & Mackey, A. 1993. An empirical study of children’s eSL development
and rapid profile. Studies in Second Language Acquisition 15, 495-503.
Pienemann, M., Mackey, A. & Thornton, I. 1991. Rapid Profile: A Second Language
Screening Procedure. Canberra, National Languages Institute of Australia.
Pinker, S. 1984. Language Learnability and Language Development. Cambridge, MA,
Harvard University Press.
Pinker, S. 2004. Clarifying the logical problem of language acquisition. Journal of Child
Language 31, 949-953.
Pinker, S. & Prince, A. 1988. On language and connectionism: Analysis of a parallel distributed processing model of language acquisition. Cognition 28, 73-193.
Pickering, M.J., Branigan, H.P., & McLean, J.F. 2002. Constituent structure is formulated
in one stage. Journal of Memory and Language 46, 586-605.
Plag, I. 2008a. Creoles as interlanguage. Inflectional morphology. Journal of Pidgin and
Creole Languages 23, 114-135.
Plag, I. 2008b. Creoles as interlanguage. Syntactic structures. Journal of Pidgin and Creole
Languages 23, 307-328.
Plag, I. 2011. Pidgins and creoles. In Studying Processability Theory. An Introductory
Textbook, M. Pienemann & J.-U. Keßler (eds), 106120. Amsterdam, Benjamins
Polinsky, M. 1995. Double objects in causatives: Towards a study of coding conflict. Studies
in Language 19, 129-221.
Poulisse, N. 1999. Slips of the Tongue: Speech Errors in First and Second Language Production.
Amsterdam, Benjamins.
Qi, R. 2011. The Bilingual Acquisition of English and Mandarin: Chinese Children in
Australia. New York, Cambria Press.
Qi, R. & Di Biase, B. 2005. L1 and L2 patterns in the bilingual language development of
a Mandarin-english child. Paper presented at the 5th International Symposium on
Bilingualism. Universitat Autonoma de Barcelona, March 20-23, 2005.
Qi, R., Di Biase, B. & Campbell, S. 2006. The transition from nominal to pronominal person reference in the early language of a Mandarin-english bilingual child.
International Journal of Bilingualism 10, 301-329.
Rahkonen, M. & Håkansson, G. 2008. Production of written L2-Swedish – processability or input frequencies? In Processability Approaches to Second Language Development
and Second Language Learning, J.-U. Keßler (ed.), 135-164. Newcastle, UK,
Cambridge Scholar Publishing.
Radford, A. 2004. English Syntax: An Introduction. Cambridge, UK, Cambridge University
Real Academia española 2010. Nueva Gramática de la Lengua Española. Manual. Madrid,
Renzi, L., Salvi, G.P. & Cardinaletti, A. 2001. Grande Grammatica Italiana di
Consultazione. Bologna, Il Mulino.
Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar: A
Handbook of Generative Grammar, L. Haegemann (ed.), 281-337. Dordrecht, Kluwer.
Roberts, J.A., Rice, M.L. & Tager-Flusberg, H. 2004. Tense marking in children with
autism. Applied Psycholinguistics 25 (3), 429-448.
Robinson, P. (ed.) 2012. Routledge Encyclopedia of Second Language Acquisition. New York,
Rodriguez-Mondoñedo, M. 2008. The Acquisition of Differential Object Marking in
Spanish. Probus 20 (1), 111-145.
Rothman, J. & VanPatten, B. 2013. On multiplicity and mutual exclusivity: The case for
different SLA theories. In Contemporary Approaches to Second Language Acquistion,
M.P. García Mayo, M.J. Gutierrez Mangado & M. Martínez Adrián (eds), 243-256.
Amsterdam, Benjamins.
Sakai, H. 2008. An analysis of Japanese university students’ oral performance in english
using processability theory. System 36, 534-549.
Salvi, G. & Vanelli, L. 2004. Nuova Grammatica Italiana. Bologna, Il Mulino.
Sano, K. 1977. An experimental study on the acquisition of Japanese simple sentences
and cleft sentences. Descriptive and Applied Linguistics 10, 213-233.
Sasaki, Y. 1998. Processing and learning of Japanese double-object active and causative
sentences: An error-feedback paradigm. Journal of Psycholinguistic Research 27 (4),
Savić, M. & Anđelković, D. 2007. The role of input frequency in early language production: Children’s usage of Serbian prepositions. In Frequency Effects in Language
Acquisition: Defining the Limits of Frequency as an Explanatory Concept, I. Gülzow &
N. Gagarina (eds), 145-180. New York, Mouton de Gruyter.
Schachter, J. 1974. An error in error analysis. Language Learning 27, 205-214.
Schmidt, R.W. & Frota, S.N. 1986. Developing basic conversational ability in a second
language: A case study of an adult learner of Portuguese. In Talking to Learn:
Conversation in Second Language Acquisition, R. Day (ed.), 237-326. Rowley, MA,
Newbury House.
Schwarz, B.D. & Sprouse, R.A. 1994. Word order and nominative case in non-native language acquisition. In Language Acquisition Studies in Generative Grammar, T.
Hoekstra & B.D. Schwartz (eds), 317-68. Amsterdam, Benjamins.
Schwarze, C. 2002. Representation and variation: On the development of Romance auxiliary syntax. In Time over Matter: Diachronic perspectives in morphosyntax, M. Butt &
T. Holloway King (eds), 143-172. Stanford, CA, CSLI.
Schwarze, Ch. 2009. Grammatica della Lingua Italiana. Roma, Carocci.
Schwienhorst, K. 2002. evaluating tandem language learning in the MOO: Discourse
repair strategies in a bilingual internet project, Computer Assisted Language Learning
15 (2), 135-145.
Segalowitz, N. 2003. Automaticity and second languages. In The Handbook of Second
Language Acquisition, C.J. Doughty & M.H. Long (eds), 382-408. Oxford,
Selkirk, A.G. 1984. Phonology and Syntax. The Relation between Sound and Structure.
Cambridge, MA, MIT Press.
Sells, P. 1995. Korean and Japanese morphology from lexical perspective. Linguistic Inquiry
26, 277-325.
Sells, P. 1999. Lexicalist syntactic theories and language education. Paper presented at AILA
symposium, Waseda University, Tokyo. 5 August, 1999.
Sells, P. 2001. Introduction. In Formal and Empirical Issues in Optimality Theoretical Syntax,
P. Sells (ed.), 1-16. Stanford, CA, CSLI Publications.
Semel, e., Wiig, e.H. & Secord, W. 2003. Clinical Evaluation of Language Fundamentals.
4th ed. San Antonio, TX, The Psychological Corporation.
Serianni, L. 1996. Grammatica Italiana: Italiano Comune e Lingua Letteraria. Milano,
Seung, H., Siddiqi, S. & elder, J.H. 2006. Intervention outcomes of a bilingual child with
autism. Journal of Medical Speech – Language Pathology 14 (1), 53-63.
Shibatani, M. 1987. Japanese. In The Major Languages of East and South-East Asia, B.
Comrie (ed.), 127-170. London, Routledge.
Shibatani, M. 1990. The Languages of Japan. Cambridge, Cambridge University Press.
Shibatani, M. 1994. An integrational approach to possessor raising, ethical datives, and
adversative passives. In Berkeley Linguistic Society 20, 461-486.
Shibatani, M. 2011. Japanese. In The Major Languages of the World, Comrie, B. (ed.),
741-764. London & New York, Routledge.
Silva-Corvalan, C. 1994. Language Contact and Change. Oxford, Oxford University Press.
Skehan, P. 1998. A Cognitive Approach to Language Learning. Oxford, Oxford University
Slobin, D.I. 1997. The origins of grammaticisable notions: Beyond the individual mind.
In The Crosslinguistic Study of Language Acquisition; Vol. 5, Expanding the contexts, D.I.
Slobin (ed.), 265-324. Mahwah, NJ, earlbaum.
Slobin, D.I. & Bever, T. 1982. Children use canonical sentence schemas: A crosslinguistic
study of word order and inflections. Cognition 12, 229-65.
Song, J.J. 2012. Word Order. Cambridge, Cambridge University Press.
Sorace, A. 2003. Near-nativeness. In The Handbook of Second Language Acquisition, C.J.
Doughty & M.H. Long (eds), 130-151. Oxford,Blackwell.
Sorace, A. & Filiaci, F. 2006. Anaphora resolution in near-native speakers of Italian’, Second
Language Research 22 (3), 339-368.
Sotillo, S. 2005. Corrective feedback via Instant Messenger Learning activities in NS-NNS
and NNS-NNS dyads. Calico Journal 22 (3), 467-496.
Swain, M. 1985. Communicative competence: Some roles of comprehensible input and
comprehensible output in its development. In Input in Second Language
Acquisition, S.M. Gass & C.G. Madden (eds), 235-253. Rowley, MA, Newbury
Swain, M. 1995. Three functions of output in second language learning. In Principle and
Practice in Applied Linguistics: Studies in Honor of H.G. Widdowson, G. Cook & B.
Seidlhofer (eds), 125-144. Oxford, Oxford University Press.
Tager-Flusberg, H. 2004. Strategies for conducting research on language in autism. Journal
of Autism and Developmental Disorders 34 (1), 75-80.
Tager-Flusberg, H. 2005. Designing studies to investigate the relationships between genes,
environments, and developmental language disorders. Applied Psycholinguistics 26 (1),
Tager-Flusberg, H. & Calkins, S. 1990. Does imitation facilitate the acquisition of grammar? evidence from a study of autistic, Down’s syndrome and normal children.
Journal of Child Language 17 (3), 591-606.
Tager-Flusberg, H. & Joseph, R.M. 2003. Identifying neurocognitive phenotypes in
autism. Philosophical Transactions: Biological Sciences 358 (1430), 303-314.
Tager-Flusberg, H., Paul, R. & Lord, C. 2005. Language and communication in autism.
In Handbook of Autism and Developmental Disorders, F.R. Volkmar, R. Paul, A. Klin
& D. Cohen (eds), 3rd ed., vol. 1, 335-364. Hoboken, NJ, Wiley & Sons.
Tanaka, M. 2001. The Acquisition of Point of View and Voice in Japanese as a Foreign/Second
Language: The Case of Adult Learners of English, Korean, Chinese and Indonesian/Malay.
PhD thesis, International Christian University, Tokyo.
Tankó, G. 2005. Into Europe. The Writing Handbook. Budapest, Teleki László
Taube-Shiffnorman, M. & Segalowitz, N. 2005. Within-language attention control in second language processing. Bilingualism: Language and Cognition 8, 195-206.
Tek, S., Mesite, L., Fein, D., & Naigles, L. 2014. longitudinal analyses of expressive
language development reveal two distinct language profiles among young children
with Autism Spectrum Disorders. Journal of Autism and Developmental Disorders 44
(1), 75-89. doi: 10.1007/s10803-013-1853-4.
Thomson, C.K. 2007. Gakushuu kankyoo o dezain suru: Gakushuusha komyunithii to
shite no nihongo kyooshi yoosei koosu (Designing a learning environment: Teacher
training course as a learner community). Japanese Language Education Around the
Globe 17, 169-186.
Thorne, S. 2008. Computer-mediated communication. In Encyclopedia of Language and
Education, 2nd Edition; Vol. 4, Second and Foreign language education, N. Van DeusenScholl & N.H. Hornberger (eds), 325-336. New York, Springer Science & Business
Media, LLC.
Timberlake, A. 2004. A Reference Grammar of Russian. Cambridge, Cambridge University
Tippets, I. 2011. Differential Object Marking: Quantitative evidence for underlying
hierarchical constraints across Spanish dialects. In Selected Proceedings of the 13th
Hispanic Linguistics Symposium, L.A. Ortiz-López (ed.), 107-117. Somerville, MA,
Cascadilla Proceedings Project.
Tomlin, R.S. 1986. Basic Word Order: Functional Principles. London, Croom Helm.
Torrego, e. 1998. The Dependencies of Objects. Cambridge, MA, MIT Press.
Torrego, e. 1999. el complemento directo preposicional. In Gramática Descriptiva de la
Lengua Española; Vol. 2, Las Construcciones Sintácticas Fundamentales. Relaciones
Temporales, Aspectuales y Modales, I. Bosque & V. Demonte (eds), 1779-1805.
Madrid, espasa Calpe.
Torregrossa, J. & Bettoni, C. 2013. The development of english yes/no questions at the
syntax/phonology interface. Paper presented at the 23rd Annual Conference,
European Second Language Association (EUROSLA 23), Amsterdam, 28-31.2013.
Toyoda, e. & Harrison, R. 2002. Categorization of text chat communication between
learners and native speakers of Japanese. Language Learning and Technology 6 (1), 8299.
Tsujimura, N. 1996. An Introduction to Japanese Linguistics. Cambridge, MA, Blackwell.
Valicenti-McDermott, M., Tarshis, N., Schouls, M., Galdston, M., Hottinger, K., Seijo,
R., … Shinnar, S. 2013. Language differences between monolingual english and
bilingual english-Spanish young children with Autism Spectrum Disorders.
Journal of Child Neurology 28(7), 945-948. doi: 10.1177/0883073812453204
Vallduvi, e. & engdahl, e. 1996. The linguistic realisation of information packiging.
Linguistics 34, 459-519.
Van Valin, R.D. 2001. Introduction to Syntax. Cambridge, Cambridge University Press.
Van Valin, R.D. 2005. Exploring the Syntax-Semantics Interface. Cambridge, Cambridge
University Press.
VanPatten, B. 1996. Input Processing and Grammar Instruction: Theory and Research.
Norwood, NJ, Ablex.
VanPatten, B. 2007, Input processing in adult second language acquisition. In Theories in
Second Language Acquisition, B. VanPatten & J. Williams (eds), 115-135. Mahwah,
NJ, erlbaum.
VanPatten, B. & Cadierno, T. 1993. explicit instruction and input processing. Studies in
Second Language Acquisition 15, 225-241.
VanPatten, B. & Williams, J. (eds) 2007. Theories of Second Language Acquisition. Mahwah,
NJ, erlbaum.
Vermeulen, R. 2009. On the syntactic typology of topic marking: A comparative study of
Japanese and Korean. UCL Working Papers in Linguistics 21, 335-36.
Vigliocco, G., Butterworth, B. & Garrett, M.F. 1996. Subject-verb agreement in
Spanish and english: Differences in the role of conceptual constraints. Cognition
61, 261-298.
Vigliocco, G., Butterworth, B. & Semenza, C. 1995. Constructing subject-verb agreement
in speech: The role of semantic and morphological factors. Journal of Memory and
Language 34, 186-215.
Vigliocco, G. & Franck, J. 1999. When sex and syntax go hand in hand: Gender agreement in language production. Journal of Memory and Language 40, 455-478.
Vigliocco, G. & Franck, J. 2001. When sex hits syntax: Gender agreement in sentence production. Journal of Memory and Language 45, 368-390.
Vigliocco, G., Hartsuiker, R.J., Jarema, G. & Kolk, H.H.J. 1996. One or more labels on
the bottles? Notional concord in Dutch and French. Language and Cognitive Processes
11 (4), 407-442.
Vihman, M.M., & McLaughlin, B. 1982. Bilingualism and second language acquisition
in preschool children. In Verbal Processes in Children: Progress in Cognitive
Development Research, C.J. Brainerd & M. Pressley (eds), 35-58. New York,
Vincent, N. 2011. Italian. In The Major Languages of Eastern Europe, B. Comrie (ed.), 233252. London, Routledge.
Volkmar, F.R., Reichow, B., Westphal, A., & Mandall, D.S. 2014. Autism and the Autism
Spectrum: Diagnostic concepts. In Handbook of Autism and Pervasive Developmental
Disorders, F.R. Volkmar, S.J. Rogers, P. Rhea & K.A. Pelphrey (eds), 4th ed., vol. 1, 327. Hoboken, NJ, Wiley & Sons.
Wang, K. 2006. The Acquisition of the English Passive Construction by Chinese Learners of
ESL. Unpublished BA (Honours) thesis, University of Western Sydney, Sydney.
Wang, K. 2009. Acquiring the passive voice: Online production of the english Passive construction by Mandarin speakers. In Research in Second Language Acquisition: Empirical
Evidence Across Languages, D. Keatinge & J.-U. Keßler (eds), 93-117. Newcastle, UK,
Cambridge Scholars Publishing.
Wang, K. 2010. The Acquisition of English Passive Constructions by Mandarin Speakers: A
Developmental Perspective. Unpublished PhD thesis, University of Western Sydney,
Wang, K. 2011. ‘The Fish was eaten’: The Acquisition of the English Passive Construction by
Chinese Learners of ESL. Koln, LAP Lambert Academic Publishing.
Ware, P. & Cañado, M. 2007. Grammar and feedback: Turning to language form in telecollaboration. In Online Intercultural Exchange: An Introduction for Foreign Language
teachers, R. O’Dowd (ed.), 107-126. Clevedon, UK, Multilingual Matters.
Ware, P. & O’Dowd, R. 2008. Peer feedback on language form in telecollaboration.
Language Learning & Technology 12 (1), 43-63.
Warschauer, M. 1996. Comparing face-to-face and electronic discussion in the second lan-
guage classroom. CALICO Journal 13 (2), 7-26.
Waterhouse, L. & Fein, D. 1982. Language skills in developmentally disabled children.
Brain and Language 15 (2), 307-333.
Weyerts, H., Penke, M., Münte, T.F., Heinze, H.-J. & Clahsen, H. 2002. Word order in
sentence processing: An experimental study of verb placement in German. Journal of
Psycholinguistic Research 31, 211-268.
WHO 1993. The ICD-10 Classification of Mental and Behavioural Disorders: Diagnostic
Criteria for Research. Geneva, World Health Organization.
White, L. 1989. Universal Grammar and Second Language Acquisition. Amsterdam,
Wilson, F. 2009. Processing at the syntax-discourse interface in second language acquisition. PhD Thesis. University of edinburgh.
Wirbatz, K. 2014. Investigating the Fundamental Difference between L1 and L2 Acquisition
Based on Syntax: A Fresh Look at the Development of German L1 and L2. PhD thesis,
University of Western Sydney, Sydney.
Yamaguchi, Y. 2008. The early syntactic development in child L2 acquisition: What happens after “canonical order”? In Processability Approaches to Second Language
Development and Second Language Learnings, J.-U. Keßler (ed.), 245-266. Newcastle,
UK, Cambridge Scholars Publishing.
Yamaguchi, Y. 2009. The development of plural marking and plural agreement in child
english L2 acquisition. In Research in Second Language Acquisition: Empirical Evidence
Across Languages, D. Keatinge & J.-U. Keßler (eds), 9-39. Newcastle, UK, Cambridge
Scholars Publishing.
Yamaguchi, Y. 2010. The Acquisition of English as a Second Language by a Japanese Primary
School Child: A Longitudinal Study from a Processability Viewpoint. PhD thesis,
University of Western Sydney, Sydney.
Yamaguchi, Y. 2013. Child Second Language Learning: A Study of English as a Second
Language. Saarbrücken, LAP Lambert Academic Publishing.
Yamaguchi, Y. & Kawaguchi, S. 2014. Acquisition of english morphology by a Japanese
school-aged child: A longitudinal study. Asian EFL Journal 16 (1), 89-119.
Yip, V. & Matthews, S. 2007. The Bilingual Child: Early Development and Language
Contact. New York, Cambridge University Press.
Zhang, Y. 2002. A processability approach to L2 acquisition of Chinese grammatical morphemes. In Developing a Second Language: Acquisition, Processing and Pedagogy of
Arabic, Chinese, English, Italian, Japanese, Swedish, B. Di Biase (ed.), 29-44.
Melbourne, Language Australia.
Zhang, Y. 2004. Processing constraints, categorical analysis, and second language acquisition of the Chinese adjective suffix de (ADJ). Language Learning 54, 437-468.
Zhang, Y. 2005. Processing and formal instruction in the L2 acquisition of five Chinese
grammatical morphemes. In Cross-Linguistic Aspects of Processability Theory, M.
Pienemann (ed.), 155-177. Amsterdam, Benjamins.
Zhang, Y. 2007. Testing the Topic Hypothesis: The L2 acquisition of Chinese syntax. In
Second Language Acquisition Research: Theory Construction and Testing, F. Mansouri
(ed.), 145-172. Newcastle, UK, Cambridge Scholars.
Zhang, Y. & Widyastuti, I. 2010. Acquisition of L2 english morphology: A family case
study. Australian Review of Applied Linguistics 33 (3), 29.1–29.17. doi:
Zobl, H. 1986. Word order typology, lexical government and the prediction of multiple,
graded effects in L2 word order. Language Learning 36, 159-183.
Zwaigenbaum, L., Bryson, S., Rogers, T., Roberts, W., Brian, J. & Szatmari, P. 2005.
Behavioral manifestations of autism in the first year of life. International Journal of
Developmental Neuroscience 23 (2-3), 143-152.
About the authors
Tonya G. Agostini is a PhD candidate at the MARCS Institute at the University
of Western Sydney, Australia. Her current research programme investigates early
acquisition of Italian L2 in children with high functioning autism spectrum disorder. Her research interests include early word learning and morphophonological
development in L1 and L2 acquisition in typical and atypical populations, such as
those with autism spectrum disorder, specific language impairment and dyslexia.
Daniele Artoni is a PhD student at the University of Verona, where he also tutors
in Russian literature. His research interests are second language acquisition, with a
focus on grammatical development in Russian. His wider interests include literature and music.
Catherine T. Best is Professor and Chair of Psycholinguistic Research at the
University of Western Sydney, Australia, and has a long-standing affiliation as a
senior scientist at Haskins Laboratories, USA. She received her PhD in developmental psychology and human neuropsychology, and gained subsequent training
in phonetics and phonology. She held prior academic appointments at Columbia
University and Wesleyan University (USA). Her research focuses on how language
experience shapes cross-language speech perception in infants and adults, and
influences adults’ and young children’s recognition of spoken words across regional accents of their native language.
Camilla Bettoni is Professor at the University of Verona, Italy. A graduate from the
Universities of Padua and Edinburgh, she held teaching appointments in Scotland,
Australia, and Italy, and visiting positions in Germany (Konstanz), Sweden
(Stockholm), and Austria (Klagenfurt). Her research interests include second language grammar development; first language loss; cross-cultural pragmatics; and the
(socio)linguistics of migration, with particular attention to Italian, as acquired by
immigrants in Italy and lost by emigrants abroad in contact with English.
Bruno Di Biase, is Associate Professor at the School of Humanities and Communication Studies and at MARCS Institute, University of Western Sydney,
Australia. A close collaborator with Manfred Pienemann on many projects since
the establishment of the Language Acquisition Research Centre (LARC) at the
University of Sydney in the early 1990s, he received his PhD in linguistics at the
Australian National University. His research focuses on second language development and the applications of SLA research to language teaching and learning
in formal and natural language environments, including situations of migration
and language contact.
Grammatical development in second languages, 333-334
About the authors
Giorgia Ginelli (BA, Verona 2003; MA, Padova 2005; PhD, Verona 2010) wrote
her doctoral thesis on the acquisition of constituent questions in Italian L2. A teacher and teacher trainer, she is chairperson of Italiano con noi, a school of Italian
language and culture in Verona. Her interests include language teaching methodology, especially the deployment of drama and theater thecniques in the language
Barbara Hinger is Professor of language teaching methodology (Fachdidaktik) at
the University of Innsbruck, Austria. Her research areas are second language acquisition and classroom-based language assessment, with a focus on Spanish L2. She
is chairwoman of ÖGSD (Österreichischen Gesellschaft für Sprachendidaktik, the
Austrian Association of Language Pedagogy), and a member of the board of VERBAL (Verband fur Angewandte Linguistik, the Austrian affiliate of the International
Association of Applied Linguistics).
Louise Jansen is a Lecturer in Applied Linguistics and German Studies at the
Australian National University. Her primary research interest is second language
development, with a focus on Processability Theory. Her recent data-based work
includes the acquisition of L2 German syntax in instructed settings (Jansen 2008)
and nominal plural (Charters, Dao & Jansen 2011).
Satomi Kawaguchi is a Senior Lecturer in Japanese, second language acquisition,
and TESOL research methods at the University of Western Sydney. Her research
areas include Japanese L2 and English L2. She is interested in theory-construction
in SLA and research-driven-teaching, as well as in the use of new communication
technology and student-focused innovative pedagogies. She was awarded the
Australian Teaching & Learning Council Citation for outstanding contribution to
student learning in 2010.
Marco Magnani is a PhD student at the University of Verona, where he also tutors
undergraduate students in Russian language and literature. His research interests
include second language acquisition, with a focus on grammatical development in
Russian and in Italian.
Lucija Medojević is an applied linguist at the School of Humanities and
Communication Arts, University of Western Sydney, Australia. She holds a BA
Hons in Languages & Linguistics (University Medal) and a PhD from the
University of Western Sydney. Her research interests include bilingual/first/second
language acquisition, the development of narrative skills in bilingual children, and
the connection of biculturalism to creative writing.
Yumiko Yamaguchi, PhD, is an associate professor at the Foreign Language
Center, Tokai University, Japan. She obtained master’s degree in Education from
Boston University, USA, and a PhD in Applied Linguistics from the University of
Western Sydney, Australia. Her current research focuses on ESL/EFL acquisition
and language teaching methodology.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Related manuals

Download PDF