ABSTRACTS - MTA Nyelvtudományi Intézet

ABSTRACTS - MTA Nyelvtudományi Intézet
16th International Morphology Meeting
(IMM 16)
Budapest, May 29 – June 1, 2014,
Research Institute for Linguistics (Hungarian Academy of Sciences)
Main session
Poster session
Workshop 1: Compounds
Wokshop 2: Info-theoretic
Workshop 3: Inflection
May 29 – June 1
May 31, 13:30-15:00
May 30, 14:00-15:30,
June 1, 9:00-16:30
May 30, 16:00-18:00,
May 31, 10:10-12:30, 15:00-17:00
May 31, 10:10-12:30, 15:00-17:00,
June 1, 10:10-12:30
16th International Morphology Meeting
(IMM 16)
Budapest, May 29 – June 1, 2014
Organizing committee:
Ferenc KIEFER (Hungarian Academy of Sciences)
Mária LADÁNYI (Eötvös Loránd University)
Huba BARTOS (Research Institute for Linguistics, H.A.S.)
Péter REBRUS (Research Institute for Linguistics, H.A.S.)
Program Committe:
Jóhanna BARĐDAL (Ghent University)
Martin MAIDEN (University of Oxford)
Ingo PLAG (Heinrich Heine University, Düsseldorf)
Gregory STUMP (University of Kentucky)
Ferenc KIEFER (Hungarian Academy of Sciences)
Wolfgang U. DRESSLER (Austrian Academy of Sciences)
Mária LADÁNYI (Eötvös Loránd University)
Franz RAINER (Austrian Academy of Sciences)
Hans Christian LUSCHÜTZKY (University of Vienna)
Huba BARTOS (Research Institute for Linguistics, H.A.S.)
Péter REBRUS (Research Institute for Linguistics, H.A.S.)
Assistant organizers:
Veronika HEGEDŰS (Research Institute for Linguistics, H.A.S.)
Kinga GÁRDAI (Research Institute for Linguistics, H.A.S.)
Ildikó Emese SZABÓ (Eötvös Loránd University)
When is a Nominative an Object?
Jóhanna Barðdal (Ghent University)*
In traditional grammar the nominative has been equated with subject, irrespective of argument
structure and position in the clause. This includes nominatives of “inverse” predicates such as ‘like’ in
languages with non-canonically case-marked argument structures. However, beginning in the 1960s
with the general theorizing of grammatical structure, behavioral properties of subjects were identified.
This led to the recognition that behavioral subjects could be non-canonically case marked, for instance
in the accusative, dative and the genitive, and as a corollary that behavioral objects could be in the
nominative case.
A long-standing division between Icelandic and German in the literature takes for granted that
there are non-nominative subjects and nominative objects in Icelandic, like with the Dat-Nom case
frame. At the same time, apparent corresponding structures in German have been analyzed as
diametrical opposites, i.e. as exhibiting a Nom-Dat case frame (cf. Zaenen, Maling & Thráinsson
1985, Sigurðsson 1989). This alleged contrariety between Icelandic and German is in particular based
on two factors:
a) different behavior of the dative with regard to a subset of subject tests across the two
b) the apparent subject behavior of the nominative argument in German
This presentation focuses on the second difference between Icelandic and German, i.e. the apparent
subject behavior of the nominative argument in German Dat-Nom constructions and the contradiction
in its behavior, alternating between behaving as a subject and an object. This involves an analysis of
additional data relevant to grammatical relations and case marking, data that have not figured in the
earlier literature and are vital for a deeper understanding of the overarching problem. These data
involve alternating predicates, which behave in such a way that either argument, the dative or the
nominative, may take on subject properties while the other argument of the argument structure
behaves syntactically as an object, irrespective of its case marking. The verbal predicate, however,
remains constant.
A comparison between Icelandic and German shows that Icelandic Dat-Nom predicates are of
two types, a non-alternating and an alternating type, whereas German seems to exhibit only the
alternating type. On this assumption, the apparent subject behavior of the nominative in German is
easily explained, as such occurrences in fact involve the Nom-Dat construction and not the Dat-Nom
construction. Hence, the subject behavior of the nominative does not invalidate a subject analysis of
the dative in Dat-Nom constructions in German, only in Nom-Dat constructions. This, in turn, begs the
question of which factors decide on which of the two constructions, Dat-Nom or Nom-Dat, are used
with a given verb in a given context.
On a more general note, the concept of alternating predicates, i.e. predicates which
systematically alternate between two diametrically-opposed argument structure constructions, in this
case Dat-Nom and Nom-Dat, is new to both linguistic theorizing and empirical method. It will be
shown how predicates like these can be accounted for in Sign-Based Construction Grammar (Sag
2012), including their case marking, the alternating argument structure constructions, and the object
behavior of the nominative.
Sag, Ivan. 2012. Sign-Based Construction Grammar: An Informal Synopsis. In Hans C. Boas & Ivan
A. Sag (eds.), Sign-Based Construction Grammar, 69–202. Stanford: CLSI Publications.
Sigurðsson, Halldór Ármann. 1989. Verbal Syntax and Case in Icelandic. Doctoral dissertation. Lund:
Department of Scandinavian Languages.
Zaenen, Annie, Joan Maling & Höskuldur Thráinsson 1985, Case and Grammatical Functions: The
Icelandic Passive. Natural Language and Linguistic Theory 3: 441–483.
*Based on joint work with Thórhallur Eythórsson and Tonya Kim Dewey.
Lexical Synonymy and the Diachrony of Paradigm Structure
Research Centre for Romance Linguistics, University of Oxford
[email protected]
From detailed observation of patterns of morphological change, mainly but not
exclusively in Romance languages, I argue that such superficially disparate phenomena as
analogical levelling-out of stem allomorphy, the diachronic persistence of morphomic
alternation patterns and the genesis of ‘incursive’ suppletion invite unified explanation if we
recognize the crucial role played in morphological change by lexical synonymy in interaction
with the long-acknowledged tendency (e.g., Girard 1718; Clark 1993) for synonymy to be
Lexical synonymy obviously exists across the word-forms of inflexional paradigms:
indeed it serves as a classic (if obviously inadequate; cf. Spencer 2013) criterion for
distinguishing ‘inflexional’ and ‘derivational’ morphology. Analogical levelling-out of rootallomorphy in inflexional paradigms can plausibly be assigned to synonymy-avoidance, such
that multiplicity of forms sharing identical lexical semantic content is subject to synonymyavoidance strategies, rooted in principles of iconicity, and favouring the survival of just one
form (‘one meaning - one form’). One of the striking diachronic characteristics of morphomic
alternation patterns is that (if they are not themselves levelled out) they overwhelmingly
display diachronic coherence: any morphological innovation affecting any one of the
paradigm-cells over which a morphome is defined always, and identically, affects all the
others. I argue that coherence reflects a strategy rendering maximally predictable, and thereby
minimizing, the mismatch between multiplicity of form and unitary lexical meaning (i.e.,
synonymy). Crucial supporting evidence comes from the observation that otherwise wholly
robust morphomic structures disintegrate where even slight lexical semantic differences
emerge, i.e., where synonymy disappears.
Despite occasional denials that truly synonymous lexemes exist (e.g., Bloomfield
1933:145), if we admit Cruse’s view of synonyms (2011:142) simply as ‘words with
construals whose semantic similarities are more salient than their differences’ then there are
various circumstances under which speakers may find themselves unable to distinguish the
meaning of different lexemes. As argued in Maiden (2004) and elsewhere, the emergence of
synonymy between historically distinct lexemes (a scenario particularly common under
conditions of language contact — cf. Carstairs-McCarthy 2010:ch.4 — but also with a variety
of other causes) is a common source of suppletion, the synonymy being resolved (‘avoided’)
by merging or ‘accommodating’ the lexemes, effectively by assigning them (often,
morphomic) complementary domains.
However, my view that lexical suppletion is determined by synonymy-avoidance (and
is therefore of a kind with levelling and morphomic coherence) has been interestingly but, I
submit, incorrectly, challenged by Börjars and Vincent (2011), who adduce Scandinavian
adjectives meaning ‘small’ as an example of a suppletion (for number) which, allegedly, does
not involve synonymy. This type is actually found beyond Scandinavian (occurring, for
example, in Middle Breton and Middle Cornish), having a particularly interesting
manifestation in Megleno-Romanian (where ‘big’ also shows suppletion). Through analysis
of the Megleno-Romanian data I show that such behaviour in basic ‘size’ adjectives is in fact
strong evidence for the role of avoidance of lexical synonymy in determining suppletion, even
though that synonymy turns out to be ‘local’, in that it occurs only in plural cells. We shall
see that this is not the only case where ‘local’ synonymy between lexemes is sufficient to
result in suppletion.
In conclusion, I argue that avoidance of lexical synonymy drives and unites apparently
disparate forms of morphological change. My conclusions thus very closely support the
arguments about the role of synonymy-avoidance in the evolution of morphology developed
by Carstairs-McCarthy (e.g., 2010). I shall particularly underscore, however, the role of
distributional predictability in morphological change, suggesting — perhaps surprisingly —
that appeal to the ‘autonomy’ of morphology in such cases may be open to question.
BLOOMFIELD, L. 1933: Language. New York: Holt • BÖRJARS, K. and VINCENT, N. 2011. ‘The pre-conditions
for suppletion’, in Galani, A., Hicks, G. and Tsoulas, G. (eds) Morphology and its Interfaces. Amsterdam:
Benjamins, pp. 239-65 • CARSTAIRS-MCCARTHY, A. 2010 The Evolution of Morphology. Oxford: OUP •
CLARK, E. 1993: The Lexicon in Acquisition. Cambridge: CUP • CRUSE, A. 2011. Meaning in Language. An
Introduction to Semantics and Pragmatics. Oxford: OUP • GIRARD, G. 1718. La Justesse de la langue françoise,
http://visualiseur.bnf.fr/Visualiseur?Destination=Gallica&O=NUMM-50626 • MAIDEN, M. 2004. ‘When
lexemes become allomorphs: on the genesis of suppletion’, Folia Linguistica 38:227-56; • SPENCER, A. 2013.
Lexical Relatedness. Oxford: OUP
Homophony in morphology:
The acoustic properties of morphemic and non-morphemic word-final s in English
Heinrich-Heine-Universität Düsseldorf
[email protected]
Recent research on lexemes has shown that homophonous lexemes show striking phonetic
differences (e.g. Gahl 2008, Drager 2011), with important consequences for models of speech
production such as Levelt et al. (1999). Such models would need to be adjusted to
accommodate frequency effects at the lemma level and/or information of fine phonetic detail
into lexical representations (e.g. Johnson 1997).
These findings also pose the question of whether similar problems hold for allegedly
homophonous affixes (instead of free lexemes), or with affixes and phonologically identical
non-morphemic segments. If the same kind of effects were to be found also with affixes it
would seriously challenge traditional analyses. A good test case for this is English, which, in
the traditional view, has a number of homophonous bound {s} morphemes. Thus, plural,
genitive, genitive plural, the auxiliary forms of has and is, and the pronoun us (as in let’s) can
all be realized by /s/ (after voiceless segments) or /z/ after voiced segments (cf. plural boys,
genitive singular boy's, genitive plural boys', or cliticized boy's, see e.g. Bauer, Lieber & Plag
2013). The underlying assumption for such analyses is that the form of morphemes is
lexically specified only phonologically, i.e. without phonetic specification.
The present investigation challenges this assumption. First, we test Walsh & Parker’s
(1983) hypothesis that plural /s/ and /z/ differ acoustically from non-morphemic final /s/ and
/z/. Our data show a significant contrast between the two groups, with non-morphemic /s/ and
/z/ being longer than the plural allomorphs. We then present an investigation of almost 600
words from the Buckeye corpus (Pitt et al. 2007) that additionally includes words with the
other {s} morphemes mentioned above. We demonstrate that there are significant acoustic
differences between at least some of these morphemes (especially in duration). Mixed effects
regression models with a number of pertinent covariates (e.g. speech rate, position in the
phonological phrase, frequency etc.) show that, for example, plural {s} is significantly longer
than all other {s} morphemes.
At the theoretical level these findings challenge standard assumptions in
morphological theory, Lexical Phonology and models of speech production.
BAUER, LAURIE, ROCHELLE LIEBER & INGO PLAG. 2013. The Oxford reference guide to English morphology.
Oxford: Oxford University Press.
DRAGER, KATIE. 2011. Sociophonetic variation and the lemma. Journal of Phonetics 39(4). 694-707.
GAHL, SUSANNE. 2008. Time and thyme are not homophones: The effect of frequency on word durations in
spontaneous speech. Language 84(3). 474-496.
JOHNSON, KEITH. 1997. Speech perception without speaker normalization: An exemplar model. In Keith Johnson
(ed.), Talker variability in speech processing, 145-165. San Diego, CA: Academic Press.
LEVELT, WILLEM J. M., ARDI ROELOFS & ANTJE S. MEYER 1999. A theory of lexical access in speech
production. Behavioral and Brain Sciences 22(1). 1-75.
ERIC FOSLER-LUSSIER 2007. Buckeye corpus of conversational speech, 2nd release. Columbus, OH:
Department of Psychology, Ohio State University.
WALSH, THOMAS & FRANK PARKER. 1983. The duration of morphemic and non-morphemic /s/ in English.
Journal of Phonetics 11. 201-206.
Polyfunctionality, positional exponence and affix conflation
University of Kentucky
[email protected]
The only grammatically significant relation that exists between inflectional markings and
morphosyntactic properties is that of exponence (Stump 2001:11); but variable affix ordering
and morphotactic conditioning reveal that affixes participate in three sorts of exponence
relations: inherent, positional and conflated. The latter two relations contribute to the
polyfunctionality of a language’s inflectional morphology, as evidence from Swahili reveals.
In Swahili, the noun-class concord vi- is an inherent exponent of the property set {σ} =
{{GEND:ki-vi, NUM:plural}}. But when it appears in affix position –3 (see (1) below), it is
also a positional exponent of subject agreement{{ARG:sbj}}, as in (2a); and when it appears
in affix position –1, it is a positional exponent of object agreement {{ARG:obj}}, as in (2b).
Most of the Swahili noun-class concords exhibit this sort of positional exponence, one of the
sources of polyfunctionality in Swahili verb inflection.
A related source is the phenomenon of affix conflation, in which two affixes combine to
form a complex affix whose content combines that of its component affixes and whose
distribution may override the default morphotactics of one or both of these affixes. This
phenomenon is exemplified by Swahili relative verb forms, which inflect for the gender and
number of a relativized argument. A relative verb form’s relative affix is (in most cases) a
conflated affix consisting of two parts: (i) the noun-class concord appropriate for the gender
and number of the relativized argument; and (ii) the affix o, an inherent exponent of relative
agreement {{ARG:rel}}. Conflated relative affixes combine the content of these two parts;
for example, the noun-class concord vi- (an inherent exponent of {σ} = {{GEND:ki-vi,
NUM:plural}}) conflates with o to produce the relative affix vyo, a conflated exponent of
{{ARG:rel} ∪ σ}. In the absence of overt marking for tense or negation, a relative affix is by
default suffixal, appearing in affix position ±2 (cf. (3a)). But relative affixes themselves
conflate with TN2 affixes (affixes of tense or negation associated with position ±2); these
conflated affixes appear prefixally in position ±2 (cf. (3b)), overriding the relative affixes’
default suffixal morphotactics. Thus, the Swahili noun-class concords are polyfunctional
both because they serve as positional exponents of subject and object agreement and because
they conflate with o to form relative affixes (which may in turn conflate with TN2 affixes).
How are inherent, positional and conflated exponence to be modeled theoretically? A
reasonable approach is to distinguish declarations of affixal exponence from rules of affix
sequencing (Stump 1993: 174f), the former being the domain of inherent exponence relations,
and the latter, the domain of positional exponence relations; together, an exponence
declaration and a sequencing rule have the effect of a rule of exponence of the sort familiar in
inferential-realizational theories of morphology (Zwicky 1986, Stump 2001). Instances of
affix conflation can, correspondingly, be seen as involving the conflation of distinct
exponence declarations. Under such an approach, an exponence declaration is the pairing
x, σ of an affix x with a morphosyntactic property set σ; thus, the Swahili affix vi- has the
exponence declaration in (4a), which specifies its inherent exponence. This affix is subject to
the sequencing rules in (4b), which specify its positional exponence; these rules are
interpreted realizationally, as in (4c, d).
The exponence declaration in (4a) is also subject to conflation, where the conflation of
two exponence declarations x, σ and y, τ is the exponence declaration xy, s(σ, τ) ,
whose operator s is as defined in (5a, b). Thus, according to the conflation rule in (6a), the
conflation of vi, {{GEND:ki-vi, NUM:plural}} and o, {{ARG:rel}} (= vyo, {{ARG:rel,
GEND:ki-vi, NUM:plural}} ) is a relative affix, whose sequencing is determined by rule (6b);
and according to rule (7a), the conflation of the TN2 affix li, TNS:pst and the relative
affix vyo, {{ARG:rel, GEND:ki-vi, NUM:plural}} is a TN2 affix livyo, {TNS:pst, {ARG:rel,
GEND:ki-vi, NUM:plural}} , whose sequencing is determined by rule 7b .
This approach to modeling Swahili verb inflection correctly entails that by default, a
subject-agreement affix has the same form as the corresponding object-agreement affix.
Moreover, it correctly entails five facts about the relative affixes: (a) the relative prefixes
have the same form as the relative suffixes; (b) the relative suffixes are mutually exclusive
with the TN2 prefixes; (c) the relative suffixes are mutually exclusive with the relative
prefixes; (d) the relative prefixes require a TN2 prefix; and (e) the relative prefixes are
adjacent to the TN2 prefixes. This analysis makes it possible to say that each Swahili nounclass concord has a constant bit of content (of which it is an inherent exponent) but may
express additional content according to its position or to its conflation with other affixes.
(1) Affix positions in
tense or
negative subject
final relative
Swahili verb
prefix ha- agreement
vowel suffix
prefix siroot
(2) a. Vi-tabu vi-me-anguka.
σ-book SBJ:σ-COMPL-fall.down
‘The books have fallen down.’
(3) a. vi-tabu a-vi-soma-vyo
σ-book 3SG.SBJ-OBJ:σ-read-REL:σ
‘the books which he reads’
‘Have you seen the books?’
‘the books which he read’
(4) a. vi, {{GEND:ki-vi, NUM:plural}} is a noun-class affix. (read: the noun-class affix viis an inherent exponent of the property set {{GEND:ki-vi, NUM: plural}}.)
b. For any noun-class affix x, {σ} ,
Pref( x, {{ARG:sbj} ∪ σ} ) is a sequencing rule for Slot –3;
Pref( x, {{ARG:obj} ∪ σ} ) is a sequencing rule for Slot –1.
c. Interpretation: Pref( z, τ ) applies to the pairing of a stem Y with a property set ρ only
if ρ is an extension of τ; the result of applying Pref( z, τ ) to 〈Y, ρ〉 is 〈zY, ρ〉.
(Similarly, the application of Suff( z, τ ) to 〈Y, ρ〉 is 〈Yz, ρ〉.)
d. Given two property sets σ, τ: σ is an extension of τ iff for each x ∈ τ, either x ∈ σ or x
is a set such that for some y ∈ σ, y is an extension of x.
(5) a. Where σ and τ are morphosyntactic property sets, s(σ, τ) is smallest well-formed
extension of both σ and τ (and is undefined if no such property set exists).
b. Given two property sets σ, τ, τ is smaller than σ iff either (i) |σ| > |τ| or (ii) |σ| = |τ| and
σ is an extension of τ but τ is not an extension of σ.
(6) a. The conflation of a noun-class affix x, σ and the relative affix o, {{ARG:rel}} is a
relative affix.
b. For every relative affix x, σ , Suff( x, σ ) is a sequencing rule for Slot ±2.
(7) a. The conflation of a TN2 affix x, σ and a relative affix y, τ is a TN2 affix.
b. For every TN2 affix x, σ , Pref( x, σ ) is a sequencing rule for Slot ±2.
ASHTON, E. O. 1944. Swahili grammar. Essex: Longman. ● LIPPS, JONATHAN. 2011. A Lexical-Functional analysis
of Swahili relative clauses. MPhil thesis, Oxford University. ● STUMP, GREGORY T. 1993. Position classes and
morphological theory. In G. Booij & J. van Marle (eds.), Yearbook of Morphology 1992, 129-180. Dordrecht:
Kluwer. ● STUMP, GREGORY T. 2001. Inflectional morphology. Cambridge: Cambridge University Press. ●
ZWICKY, ARNOLD M. 1985. How to describe inflection. BLS 11, 372-386. Berkeley Linguistics Society.
Hyponymy in simplex and complex formations and the distinction
between skeleton and body
University of Patras
[email protected]
Hyponymy is a relation of inclusion and, in particular, it is the KIND/TYPE/SORT OF semantic
relation among words (e.g. whale is a hyponym of mammal). Contrary to semantic relations
such as synonymy, hyponymy is asymmetrical and hierarchical. That is, the relation which
holds between whale and mammal (whale<mammal) is not the same as the relation between
mammal and whale (mammal>whale).
The study of hyponymy with respect to the Morphology-Lexical Semantics interface is of
paramount importance since the hyponymy test is used by scholars for the identification of the
semantic head of the word. In this respect, door knob is headed by knob since the whole
compound is a ‘kind of’ knob (knob>door knob). The way hyponymy works in morphology,
however, (with the exception of Bauer, 1990) has not been studied in any detail. This is mainly
due to the fact that a framework which would allow one to conduct such a research has been
In this paper, I enquire into the use of hyponymy with respect to simplex and complex
configurations. In particular, I adopt the framework of Lieber (2004) and raise two questions.
The first question is whether hyponymy relates the complete lexical-semantic representations
of two items or only parts of them. On the latter assumption, hyponymy may hold between the
grammatical skeletons in which categorial information is encoded or between the pragmatic
bodies of two lexical items which consist of encyclopedic knowledge. The second question is
whether hyponymy should be considered as a relation between words or as a relation among
the meanings of words (for a discussion see Murphy, 2003). The former renders hyponymy a
paradigmatic lexical relation and, as a result, it should be considered important for lexical
organization. On the contrary, the latter shows that hyponymy is a relation among the things
lexical items describe and it should, therefore, not be represented in the lexicon. That is,
relations of hyponymy and hyperonymy are available as part of our knowledge of the world
and should not be part of the organization of the lexicon.
In order to answer these questions, I first examine the way hyponymy relates simplex lexical
items since the relations which hold between simplex items are also evident in the relations
among complex items (Lieber, 2004) and show that categorial information is not relevant to
hyponymy. Consider the lexical semantic representations of flower and rose in (1):
(1a) flower
[+material ([ ])]
{a bloom on a plant}
(1b) rose
[+material ([ ])]
{a fragrant flower}
Given that hyponymy is an asymmetrical relation, it is not accurate to claim that it relates the
skeletal parts of rose and flower, since the relation (if any) which holds between the skeletons
of rose and flower is the same as the relation between the skeletons of flower and rose, i.e. both
are [+material]. Rather, it seems that the relation of hyponymy holds between the bodies of the
two items. That is, based on the information {a fragrant flower}, rose is considered a hyponym
of flower. This is in accordance with the hierarchical and asymmetrical nature of hyponymy
since the relation which holds between the bodies of flower and rose (flower>rose) is not the
same as the relation between the bodies of rose and flower (rose<flower). That is, flower is a
hyperonym of rose and rose is a hyponym of flower.
Hyponymy can also hold between members of distinct lexical categories (Lyons, 1977). By
way of example, adjectives such as sweet/bitter and happy/sad can serve as hyponyms of nouns,
namely taste and emotion respectively. In a similar vein, a number of countable nouns may be
in a relation of hyponymy with other uncountable nouns. The noun chair, for example, is a
hyponym of furniture (also knife<cutlery, shirt<clothing). Of importance is that the relation of
hyponymy holds between the pragmatic bodies of two lexical items and not between the
skeletons in which categorial information is encoded. That is, no hierarchical and asymmetrical
relation can be established between sweet and bitter if we take into consideration that sweet is
an adjective and taste a noun. In a similar vein, no hyponymic relation can be established
between chair and furniture if we consider that chair is a mass noun and furniture a singular
count noun.
Another argument in favor of the proposal that hyponymy relates the bodies of two items
comes from the distinction between taxonomic and functional hyponymy (Miller, 1998). A dog,
for example, can be a taxonomic hyponym of animal and a functional hyponym of pet. Of
relevance to our discussion is that the relations of taxonomic and functional hyponymy are
available to us as part of the pragmatic bodies of lexical items.
The proposal that hyponymy and hyperonymy are only established between the bodies of
two items has implications for the way we classify complex configurations into headed and
headless. In particular, Bauer (2010:167) argues that headless compounds can fail the
hyponymy test in three ways: (a) they can fail to show a head element, (b) they may belong to
a lexical category which is not the category of their head element, and (c) their head may be of
the correct category, but with the wrong meaning. On the assumption that the lexical category
of an element (skeleton) is not relevant to hyponymy, these claims should be reassessed.
In more detail, that the hyponymy test applies to the bodies of two items can account for the
fact that compounds which exhibit a figurative meaning (e.g. metaphorical and metonymical
compounds) fail the hyponymy test. A compound such as hard hat (taken from Lieber, 2009:69)
can be both headed and headless depending on whether it denotes a ‘kind of’ hat (helmet) or a
person (worker). Notice that whether hard hat is headed or headless depends on whether a
relation of hyponymy can be established between the bodies of hat and hard hat; skeletal
information is not relevant since there is no change in the word-class of the compound in any
Configurations in which an anti-intersection adjective, that is, an adjective which requires
negation of the noun with which it combines, modifies a noun (e.g. Greek psevd-o-profitis
‘false-LE-prophet’) can also greatly inform our discussion. A similar problem arises in the case
of denominal derivatives with the prefix anti-. Formations such as anti-fascist and anti-hero
denote the ‘opposite of X’ and fail the hyponymy test (e.g. an anti-fascist is not a fascist). Of
importance is that these formations fail the hyponymy test as a result of their non-heads, psevdo‘pseudo, false, phony’ in psevd-o-profitis and anti- in anti-fascist.
Given that hyponymy only relates the pragmatic bodies and not the grammatical skeletons
of two items, the proposed analysis argues in favor of the idea that hyponymy should not be
represented in the lexicon. Relations of hyponymy and hyperonymy are available as part of our
knowledge of the world and should not be part of the organization of the lexicon.
BAUER, L. 1990. Be-heading the word. Journal of Linguistics 26, 1–31. • BAUER, L. 2010. The typology of
exocentric compounding. In S. Scalise and I. Vogel (Eds.), Cross-disciplinary issues in compounding, pp. 167–
176. Amsterdam: John Benjamins. • LIEBER, R. 2004. Morphology and lexical semantics. Cambridge: Cambridge
University Press. • LIEBER, R. 2009. A lexical semantic approach to compounding. In R. Lieber and P. Štekauer
(Eds.), The Oxford handbook of compounding, pp. 78–104. Oxford University Press. • LYONS, J. 1977. Semantics.
Cambridge: Cambridge University Press. • MILLER, G. A. 1998. Nouns in WordNet. In C. Fellbaum (Ed.),
WordNet: an electronic lexical database, pp. 23–46. Cambridge, MA: MIT Press. • MURPHY, M. L. 2003.
Semantic relations and the lexicon. Cambridge: Cambridge University Press.
Contrastive co-reference cross-linguistically: the issue of polyfunctionality in wordformation
University of Turin
[email protected]
In the languages of Europe we find complex formations – compounds or prefixed words,
such as Italian autodistruzione or German Selbstzerstörung ‘self-destruction’ –, in which the
semantic contribution of the first element (e.g. self-, auto-, selbst-) allows to manifest a
relation of co-reference between arguments or non-argumental participants – in the proposed
example: between the destroyer and the destroyed referent. Besides co-reference, elements
like auto- suggest a contrast between the specific configuration of action and a more usual
configuration in which, for example, the destroyer and what is destroyed are different entities.
Because of these two main contributions (co-reference and contrast) I define the function
expressed by these marks as contrastive co-reference (Angster 2012).
Apart from sparse works on single marks in specific languages (e.g. Mutz 2004 on Italian and
French), little attention has been until now devoted to this kind of formations in a crosslinguistic perspective. König (2011), a pilot study based on a cross-linguistic sample,
proposes two sets of so-called “reflexive compounds” comparing the functions of intensifiers
with functions performed by the elements found as first constituent in such compounds. The
first set corresponds to “adnominal reflexive compounds” in which co-reference is connected
to the remarkability of a patient (Selbstironie ‘self-irony’ = usually irony is directed toward
others); the second set is that of “adverbial reflexive compounds”, in which the agent is
remarkable (Selbsthilfe ‘self-help’ = usually the help is given by others).
First of all, data issued from 30 European languages (mainly Indo-European, but also from
Finno-ugric, Uralo-altaic and Afro-asiatic languages and a language isolate, namely Basque)
will support König’s conclusion that the main source for contrastive co-reference marks are
elements which have in syntax at least one of the functions typical of the formally
heterogeneous class of intensifiers (König/Gast 2006).
Secondly I will show that, despite the existence of different strategies connected with the
expression of contrastive co-reference, in no language it is possible to find a formal
differentiation connected to the functions proposed in König’s work.
Thirdly, besides the fact that formal differentiations in this realm are far from being clear-cut,
I will claim that, if a formal differentiation arises between two productive word-formation
strategies, this differentiation is connected to the semantic class which the second element of
a complex formation belongs to. For example in German two competing marks exist: selbstand eigen-. Hand in eigenhändig is an object in Croft’s (2001) sense and a formation like
selbsthändig does not exist and would not be well-formed. This conclusion is supported by
other languages in which a competition between different marks exists. For example, beside
German and several other Germanic languages, also most Slavic languages, Finnish and
Estonian and Modern Greek show a similar contrast between two marks. However most part
of the languages in the sample show only a mark related to the realm of contrastive coreference. In these languages either the CC-marks are not compatible with base-words
belonging to the class of objects (for example auto- in Romance languages and self- in
English), or the same mark used with deverbal base-words of the class of actions is applied
(see Lithuanian savi-, Hungarian ön-). An unavoidable difficulty in the study of the contrast
between CC-marks and marks of what we could define “contrastive possession” is that the
second function is more rare, the relevant processes are less productive and therefore the final
lexemes often have a rather idiomatic meaning.
Fourthly, functions that are evident in syntax may be less evident in the realm of word-
formation and what is sufficient to lead to a formal differentiation in syntax may be just a
different interpretation of a single formal strategy in word-formation. Defining a set of
possible functions confronting related syntactic and morphological strategies might give a
road-map for defining functions of affixes and word-formation processes. However, when one
works on word-formation on a cross-linguistic basis, only an evident formal differentiation
with stable form-function pairings in specific languages gives us a cross-linguistically reliable
generalization about a function. In absence thereof, we should just speak of different
interpretations of a wider function, even if these interpretations have a formal differentiation
in the relevant paraphrases. In this context, contrastive co-reference appears to lack reliable
formal differentiations in word-formation connected to different functions detectable instead
through a syntactic paraphrasis.
Selected references
ANGSTER, M. (2012) Marche di coreferenza contrastiva nella formazione delle parole. Una panoramica delle
lingue d’Europa. (unpublished PhD thesis)
CROFT, W. (2001). Radical Construction Grammar. Oxford : Oxford University Press.
GAST, V., HOLE, D., KÖNIG, E., SIEMUND, P., & TÖPPER, S. (2007). Typological Database of Intensifiers and
Reflexives. Version 2.0. http://www.tdir.org.
KÖNIG, E. (2011) “Reflexive nominal compounds” Studies in Language 35:1. 112-127.
KÖNIG, E. & GAST, V. (2006). “Focused assertions of identity: A tipology of intensifiers”. Linguistic Typology,
10, 223–276.
MUTZ, K. (2004). “Zur Argumentstruktur der deverbalen Ableitungen von auto-“. In: M. HUMMEL and R.
KAILUWEIT (eds), Semantische Rollen in der Romania. Tübinger Beiträge zur Linguistik 472 , 355–374.
Tübingen: Narr.
Holistic properties of morphological constructions: Evidence from Akan
Department of Linguistics, University of Ghana
[email protected]
In accounting for the properties of morphological constructions, one may adopt a sourceoriented view where every property of the whole is expected to be accounted for in the parts.
Alternatively, one may adopt a product-oriented view where the whole may have properties
that do not necessarily emanate from the properties of the constituents (cf. Haspelmath, 1989;
Zager, 1981).
Studies on Akan complex nouns have been largely source-oriented. Scholars have
assumed that properties of the whole either emanate directly from the constituents, e.g., the
syntactic category of complex nouns(Anderson, 2013; Anyidoho, 1990; Christaller, 1875;
Obeng, 2009), or arrived at through processes that modify corresponding properties in the
constituentse.g., tonal melody of complex nominals(Abakah, 2004, 2006; Anderson, 2013;
Anyidoho, 1990; Obeng, 2009).
Constructionist approaches to morphology have stressed the fact that morphological
constructions often have properties that do not emanate from those of their constituents. Such
properties are said to be holistic or output properties of the constructions themselves. In this
presentation, I will provide evidence of such holistic properties of complex nouns in Akan
(Kwa, Niger-Congo) in the form of the syntactic category and tonal melody of Akan
compounds as well as the semantics of a particular construction types which Appah (2013)
calls personal attribute nominal constructions
First, I will show that the syntactic category of Akan compounds is a constructional
property not depending on the syntactic category of the constituents. Thus, notwithstanding
the syntactic category of the constituents, the Akan compound is invariably nominal. For
example, all the compounds in (1) are nouns although there are two nouns (1a), a noun and an
adjective (1b), a verb and a noun (1c), noun and a verb (1d) and two verbs (1e).
(1) a.
‘to take’
‘to ask’
‘bad luck’
Secondly, I show that the tonal melody of Akan compounds may not necessary be accounted
for by tweaking those of their constituents. Dolphyne (1988) identified two types of
compounds based on their surface tonal melodies. In the first, all the tone bearing units
preceding the last constituent of the compound bear low tone, as in (2). In the second, the
constituents of the compound retain their underlying tonal melodies.
(2) a. àbùsùà-bɔ́
b. bàkà-nú
‘being a family member’
‘fishing in a lagoon’
‘education, learning’
The first tonal melody is regarded as a defining feature of Akan compounds and is normally
said to result from the spread of low from the left edge of the word. I argue that this tonal
melody does not define compoundhood in Akan because this tonal melody characterises other
non-compound lexical items like the personal attribute nominal constructions. As such, the
tonal melody should be regarded as a lexical tonal melody and a holistic property of
lexical(ized) items because it seems to be independent of the tonal melodies of the
Finally, I show that a large part of the semantics of personal attribute nominal
constructions is a holistic property attributable to the constructional schema and not to the
individual instantiating nominals. This is because, the constructions as a whole refers to a
property predicated of the possessor of the body-part occurring as the first constituent in the
construction. Thus, the construction as a whole only bears a non-direct semantic link to one of
the constituents.
I will present a construction morphology modelling of the lexical tonal melody and the
semantics of the personal attribute nominal constructions.
ABAKAH, E. N. (2004). The segmental and tone melodies of Akan. Ph.D. Dissertation, Norwegian University of
Science and Technology, Trondheim. ABAKAH, E. N. (2006). The Tonology of Compounds in Akan.
Languages and Linguistics, 17, 1-33. ANDERSON, J. C. (2013). Verb-internal compound formation in Akan.
Journal of West African languages, XL(1), 89-104. ANYIDOHO, A. (1990). On Tone in Akan Compound
Nouns. Paper presented at the the 19th West African Languages Congress, University of Ghana, Legon.
APPAH, C. K. I. (2013). Construction Morphology: Issues in Akan Complex Nominal Morphology. PhD
dissertation, Lancaster University, Lancaster. CHRISTALLER, J. G. (1875). A Grammar of the Asante and
Fante Language called Tshi (Chwee, Twi) based on the Akuapem dialect with reference to the other (Akan and
Fante) dialects. Basel: Basel Evangelical Missionary Society. DOLPHYNE, F. A. (1988). The Akan (TwiFante) language: Its sound systems and tonal structure. Accra: Ghana Universities Press. HASPELMATH, M.
(1989). Schemas in Hausa plural formation: product-orientation and motivation vs. source-orientation and
generation. Buffalo Working Papers in Linguistics, 89, 32-74. OBENG, S. G. (2009). Akan deverbal nouns. In
S. G. Obeng (Ed.), Topics in Descriptive and African Linguistics: Essays in Honor of Distinguished Professor
Paul Newman (pp. 96-110). Muenchen: LINCOM Europa. ZAGER, D. (1981). On orientation or, constraining
false analogy. Linguistics, 19(11-12), 1107-1131.
Interplay of agglutination, cumulation and overabundance:
non-canonical case-number paradigm in Adyghe
Institute of Slavic Studies of the Russian Academy of Sciences / Russian State University for
the Humanities / Sholokhov Moscow State University for the Humanities, Moscow
[email protected]
Circassian languages (Adyghe and Kabardian) belong to the North-West Caucasian
family well-known for the complexity of its polysynthetic verbal morphology. By contrast,
nominal morphology in Circassian languages is relatively simple. Putting aside possessive
prefixation, adverbial and coordinative suffixes, the Circassian nominal paradigm is constituted by two numbers (singular and plural) and there cases (absolutive, oblique and instrumental). Cf. the paradigm from Kabardian (e.g. Colarusso 1992: 51–52; Kumakhov & Vamling 2009: 22–25) in table 1, showing that case suffixes are uniform across numbers and that
number exponence is uniform across cases, implying a fully transparent agglutinative organization of inflection; the only complication arises with the instrumental, which is formally
based on the oblique.
Table 1. The Kabardian nominal paradigm
Instrumental -m-č ̣ʼe
-xe-m-č ̣ʼe
However, this almost ideal transparency and uniformity of the Circassian nominal
paradigm is disrupted in the closely related Adyghe (e.g. Paris 1989: 166–167). In this language, both the structure of the paradigm and the inflectional material are identical to those
shown in Table 1, but for one exception: the combination of values {plural, oblique} can be
expressed, in addition to the agglutinative combination of the plural suffix -xe and the oblique
suffix -m, by the cumulative suffix -me, or by the combination of the latter with the plural suffix, cf. table 2.
Table 2. The Adyghe nominal paradigm.
-xe-m, -me, -xe-me
Instrumental -m-č ̣ʼe
-xe-m-č ̣ʼe
From the point of view of morphological typology and morphological theory, the
Adyghe paradigm in table 2 poses several problems, primarily because different kinds of exponents for the oblique plural cell occur in free variation thus constituting a non-trivial example of the understudied phenomenon of overabundance (Thornton 2012). Though cumulative
exponents in otherwise “separatist” paradigms are well attested cross-linguistically (see e.g.
Plank 1986, 1999), instances where both the “separatist” and the cumulative expression of the
same bundle of morphosyntactic values coexist seem to be very rare or at least underdocumented. The situation in Adyghe is further complicated by the pleonastic combination -xe-me,
where the value “plural” is expressed twice, first by the “separatist” plural suffix -xe and then
by the cumulative oblique plural -me. Thus from the point of view of Canonical Typology
(e.g. Corbett 2011), the Adyghe paradigm is not just non-canonical in that it deviates from the
requirement of uniform composition of paradigm cells, but highly non-canonical in that it violates this requirement in several ways at once.
The Adyghe paradigm is problematic for realizational theories of morphology such as
Paradigm Function Morphology (Stump 2001), which crucially incorporate the so-called Panini’s Principle, whereby among morphological rules competing for the exponence of the same
bundle of morphosyntactic features the one whose domain of application is more narrowly
specified wins. Indeed, the description of the Adyghe declension should incorporate at least
the following realization rules (the rules format is simplified for expository reasons):
{number:plural} → -xe
{case:oblique} → -m
{number:plural; case:oblique} → -me
{case:instrumental} → {case:oblique}-č ̣ʼe
According to Panini’s Principle, rule (3) should always override rules (1) and (2), implying that the oblique plural forms in -xe-m would never be derived, unless the application of
(3) is specified as optional. Even in this case, it remains unclear how the pleonastic exponence
-xe-me can be derived, since the rule introducing -me should arguably belong to the same
block of rules as -xe. Further problems are posed by the instrumental taking oblique as its
stem: the instrumental plural does not inherit the overabundance of the oblique plural, since
potential forms in *-me-č ̣ʼe and *-xe-me-č ̣ʼe are ungrammatical. This implies that the formulation of rule (4) should contain a stipulation that only the oblique suffix -m can serve as its input.
The Adyghe nominal paradigm clearly shows that just a single “deviant” cell in a
paradigm can render the whole paradigm highly non-canonical and that the phenomenon of
overabundance is especially problematic for morphological theory, first because it requires
special treatment in formal frameworks, and second because, as in the case of Adyghe, it may
bear the large part of responsibility for the non-canonicity of particular morphological paradigms.
COLARUSSO, J. (1992). A Grammar of the Kabardian Language. Calgary: University of Calgary Press.
CORBETT G.G. (2011). Higher order exceptionality in inflectional morphology. In: H.J. Simon,
H. Wiese (eds.), Expecting the Unexpected: Exceptions in Grammar. Berlin: Mouton de Gruyter, 107–126.
KUMAKHOV M. & K. VAMLING (2009). Circassian Clause Structure. Malmö: Malmö University.
PARIS C. (1989). Esquisse grammatical du dialecte abzakh (tcherkesse occidental). In: B.G. Hewitt
(ed.), The Indigenous Languages of the Caucasus. Vol. 2. The North West Caucasian Languages. Delmar, N.Y.:
Caravan, 154–260.
PLANK FR. (1986). Paradigm size, morphological typology, and universal economy. Folia Linguistica
20, 29–48.
PLANK FR. (1999). Split Morphology: How agglutination and flexion mix. Linguistic Typology 3.3,
STUMP G.T. (2001). Inflectional Morphology. A Theory of Paradigm Structure. Cambridge: Cambridge
University Press.
THORNTON A.M. (2012). Overabundance in Italian verb morphology and its interactions with other noncanonical phenomena. In: Th. Stolz et al. (eds.), Irregularity in Morphology (and beyond). Berlin: AkademieVerlag, 251–269.
The work has been supported by the Russian Foundation for the Humanities, grant No. 14-04-00580.
Multifunctionality in Icelandic morphology
The University of Iceland
[email protected]
Morphophonemics: stems, suffixes and ablauts
The morphophonemic structure of Icelandic words distinguishes between stems and endings.
Thus nominal forms like hest-ur ‘horse-NOM’, hest-a ‘horse-ACC.PL’, hest-um ‘horseDAT.PL’ show the stem hest- with the suffixes –ur, –a and –um. Importantly, the same
structure of stems and endings is used in verbal conjugation: drep-ur ‘kills’ – drep-a ‘they
kill’ - drep-um ‘we kill’ etc. The permitted inflectional affixes form a phonologically
delimited set, and the forms are subject to systematic morphophonemic restrictions. Possible
vowels are: /a/, /ɪ/, /ʏ/; possible consonants: /r/, /n/, /m/, /s/, /t/, /ð/. The endings are maximally
disyllabic but sometimes segmentable by “fission” (cf. below). The morphophonemic
restrictions are part of the language specific word phonological system (the “inventory”, cf.
McCarthy 2002:70) of Modern Icelandic. Ablaut and umlaut forms are learned as (related)
word forms, and there is room for phonotactic principles, e.g. stating that certain endings call
for u-umlaut: land - lönd-um ‘land-DAT.PL’ (*land-um); fara ‘go’ för-um ‘we go’ (*far-um).
Sample monosyllabic affixes
/-a/: <A,N,GEN.PL>: dag-a ‘days-GEN’, systr-a ‘sisters-GEN’, góð-a ‘goodACC.FEM’;
<N,FEM,NOM,SG>: stúlk-a ‘a girl’
<V,PRES.3PL.> far-a ‘go’, kalla ‘(they) call’
<V,INF.> far-a ’to go’ tala ’to speak’
/-na/: <N,FEM,GEN,PL> kirk-na ’churches-GEN‘, sagn-a ‘stories-GEN’,
/-ra/: <A,FEM,GEN,PL>góð-ra ’good-GEN.PL’
/-ar/ <N.Nom,Pl> hest-ar ’horses’
<N,Gen,Sg> vin-ar ’friends’
<V,2./3.Sg, Pres> tal-ar ’speak(s)’
<N, NOM,MASC> han-i ’cock’
<A, NOM,SG,DEF> góð-i ‘good-DEF’
<V,1.SG,PRES> dæm-i ’I judge’
<V, 1./3.SG, PRES,SUBJ> tak-i ’take-SUBJ’
/-ɪn/: <PARTICIPLE>: kom-in ‘(have) come’
<N, MASC,ACC,DEF>: hest-inn ‘the horse-ACC’
/-ɪr/: <N,A, NOM,PL> gest-ir ‘guests’, góð-ir ‘good-NOM.PL’
<V,2.SG,PRES> dæm-ir ‘judges’
<V,2.SG,PRES, SUBJ> tak-ir ‘you take-SUBJ’
<V,2.SG,PRES> kallað-ir ‘called’
/-ʏ/: <N,NEUT, PL,-GEN> hjört-u ‘hears-PL’
<N,FEM,SG,-NOM> stúlk-u ‘girl-OBLIQUE’
<A, NEUT,SG,-NOM> góð-u ‘good-NEUT.OBLIQE’
<V, 3.PL,PAST> fóru ‘they went’
/-ʏm/: <N,A,DAT,PL> hest-um ‘horses-DAT’, ferð-um ‘travels-DAT’, kert-um
<A,Dat.Sg/Pl> góð-um ‘goodDAT’
<V,1.PL> ber-um ‘we carry’, fór-um ‘we went’, kölluð-um ‘we called’
/-s/: <N,GEN,SG,MASC/NEUT> hest-s ‘horse’
<A,NEUT,SG> gul-t ‘yellow-NEUT’, víst ‘certain-NEUT’
<V,PARICIPLE> fær-t ‘moved’
<N,NOM,NEUT,DEF> barn-ið ´the child’
<V,PARTICIPLE > far-ið ‘gone-NEUT’
/-að/ <V,PARICIPLE> kall-að ‘called-NEUT’
Sample polysyllabic (and segmentable) affixes
/ʏr-ɪn/ <N,NOM,SG,MASC> hest-ur-inn ‘the horse’
/ɪn-a/ <N,FEM,ACC,DEF> bók-in-a ‘the book’
/að-ir/ <V,PAST,2.SG> kall-að-ir ‘you called’
Sample ablaut patterns
/a/ – /ö/: land ‘land-SG’– lönd ‘land-PL’, svartur ‘black-MASC’ – svört ‘black-FEM’
Double exponence: fara ‘go’ – för-um ‘we go’
/a/ – /ɛ/: taka ‘to take’– tekur ‘takes’
/œ/ - /ɛ/: köttur ‘cat’ – kettir ‘cats’
Indo-European ablaut:
/i/- /ei/: líta ‘to look’– leit ´looked’, fara ‘go’ – fór ‘went’
Morphological markers and disambiguation
The facts summarised above are either not handled or clumsily by generative theories, where
endings are inserted into syntactically generated trees (e.g. Halle & Marantz 1993). Thus
Rögnvaldsson (1990) assumes 55 noun declensions, whereas Müller (2005) reduces the
number by allowing syncretism within certain “domains”, but not between unrelated
paradigms, leaving out verbal conjugation. A generalisation is obviously missed which is that
a limited set of forms and ablaut relations sampled above can be used across word classes as
exponents of abstract mophemes or morphomes (Aronoff 1994). I want to suggest that the
suffixes and vowel relations are diacritics added to stems, which means that their function
depends on the context, and they do not as such have meaning. The meaning is a
morphosyntactc/semantic property derived from the text. The disambiguation of the meanings
follows the same principles as any sort of resolution of polysemy, i.e. by reading the context.
Thus /-ir/ in the context of a noun stem: gest-ir ‘guests’ is interpreted as NOM.PL, and in the
context of a verb stem: dæm-ir ‘judges’, as 2.SG, much in the same way as the form á in the
context á hest is interpreted as a preposition governing the accusative case: ‘onto a horseACC’ (movement), or as part of a verb phrase ‘owns a horse’, but in á hesti ‘on horse-DAT’,
the meaning is stative. In certain cases the interpretation and disambiguation is segmented (cf.
“fission” in Distributed Morphology, see Müller 2005:247 and references), e.g. between PAST
and 2.SG in kall-að-ir ‘you called’ (Þú kallaðir ‘You called’). Here the derived past tense
stem kall-að (root+affix) takes the ending /-ir/, but the same form occurs in the inflected
participle kall-að-ir ‘called-PL.MASC’ (Þeir eru kallaðir ‘They are called’).
This approach to the analysis of morphological multifunctionality seems to fit into the
general framework of usage based grammar, according to which “constructions are the basic
units of morphosyntax”, and that grammar is “emergent from experience, ever coming into
being rather than static, categorical and fixed” (Bybee 2006: 714).
ARONOFF, M. 1994. Morphology by Itself. MIT Press, Cambridge, Ma. • BYBEE, J. 2006. From usage to
grammar: The mind‘s response to repetition. Language 82:711-733. • HALLE, M. & A. MARANTZ 1993.
Distributed Morphology and the pieces of inflection. In: K. HALE and S. J. KEYSER (EDS.) The View from
Building 20, Cambridge Ma: MIT Press. 111-176. • MÜLLER, G. 2005. Syncretism and Iconicity in Icelandic
noun declensions: a Distributed Morphology approach. Yearbook of Morphology 2004: 229-271. •
RÖGNVALDSSON, E. 1986. Íslensk orðhlutafræði. Reykjavík:Háskóli Íslands.
Motivation in the constructionist lexicon
University of Amsterdam,Tufts University
[email protected], [email protected]
Many models of morphology and the mental lexicon manage the amount of redundancy among
lexical items by means of inheritance. Typically, inheritance is associated with the "impoverished
entry" hypothesis, according to which lexical entries are specified for their idiosyncrasies only -predictable properties are inferred from nodes higher up in the lexical network. With the help of
inheritance, lexical storage is modelled as maximally parsimonious and redundancy-free.
While inheritance is proving useful in computational theories and implementations,
psycholinguistic evidence pleads against the assumption that parsimony is a priority of the mind.
For this and other reasons, a number of theories have embraced a "full entry" model of the
lexicon (first proposed by Jackendoff 1975), stipulating that any lexical entry can be represented
with its complete set of properties. A recent full entry theory of morphology is Construction
Morphology (Booij 2010).
In a full entry lexicon, inheritance does not serve the purpose of reducing redundancy.
Instead, inheritance relations are assumed to motivate complex words by capturing the nonarbitrariness of their properties and structure (Booij 2012). Thus, a word such as sleepless is
motivated by the [X-less]A schema as well as by its base noun sleep (cf. Booij 2010: 26).
In this talk, we suggest that in a full entry lexicon, the notion of inheritance can be
enriched to model a wider variety of lexical relations in a uniform way. Taking seriously the
assumption that inheritance links capture shared information among items, we wish to draw
attention to two theoretical limitations in the usual construal of inheritance:
- inheritance is asymmetric
- inheritance proceeds 'vertically' from a more general to a more specific node.
Instead, we propose that:
- inheritance is bi-directional
- inheritance can also work 'horizontally' between lexical items of equal complexity.
The clearest advantage of this approach lies in the modeling of sister-sister relationships in the
lexicon. In a full entry lexicon, several types of relations are recognized between lexical items.
The first is the general-to-specific relation that holds between schemas and words. The second is
the part-whole relation between a base word and a related complex word. The standard view of
inheritance models these two types of relation. However, it does not capture the third, which
subsumes the paradigmatic relations between members of inflectional paradigms and among
word families, in particular in paradigmatic word formation (Koefoed and Van Marle 1980). In
more recent work, paradigmatic relations have also been posited as obtaining between schemas,
under the rubric of “second-order schemas” (Nesset 2006, Kapatsinski 2013).
Standard inheritance falls short in this domain of paradigmatic relations between items of
equal complexity, since such relations are typically symmetrical. There is often no principled way
to tell which of a pair of sisters inherits from which. This point is particularly important for root
derivations such as English ambitious or amputate, which lack a lexical base but nevertheless
form (small) families with other words such as ambition and amputee. Such clusters are
notoriously challenging, in particular for word-based morphological theories. Since ambit- and
amput- are neither base words nor seem to be candidates for a schema, the relation among the
words in their families is difficult to state.
A similar problem recurs in second-order schemas such as <[X-ism]N ≈ [X ist]N > (Booij
and Masini in preparation), instantiated by words such as socialism and socialist. While the
schema expresses the fact that many -ism nouns have -ist counterparts and vice versa, the relation
is nondirectional and non-hierarchical, and therefore not covered by traditional inheritance.
Here we suggest that this variety of lexical relations can be subsumed under an enriched
notion of inheritance based on a domain-general cognitive relation proposed in Culicover &
Jackendoff 2012. This relation is called SAME-EXCEPT and expresses the notion ‘item X is the
same as item Y except in respect Z.’ It is manifested in a wide range of perceptual and conceptual
phenomena, from vision to music to language. SAME-EXCEPT has two variants. In the first,
ELABORATION, item X has an extra component that item Y lacks. In the second, termed
CONTRAST, item X and item Y differ in a single property.
'Top-down' relations in morphology, such as between the schema [X-less]A and its
instantiation sleepless can be characterized as SAME-EXCEPT ELABORATION: the schema is
enriched by replacing the variable with a constant. Moreover, we can state that sleepless is an
elaboration of its base word sleep. Thereby, we can avoid having to say that sleep is more general
or situated 'higher up' in the lexicon than sleepless, a slight embarrassment that arises in a
traditional inheritance-type explanation.
For cases of 'horizontal inheritance', the notion SAME-EXCEPT CONTRAST offers a
precise characterization. Paradigm sisters, root family members, and the parts of second-order
schemas are best treated as non-directional relations consisting of shared and contrasting
properties. Here, SAME-EXCEPT CONTRAST captures the intuition that sister items in the
lexicon motivate each other semantically and structurally, without the existence of a higher-level
mother schema.
Thus SAME-EXCEPT enables us to capture the enriched notion of inheritance, unifying
various types of lexical relation. Moreover, since it is a domain-general relation, it incurs no extra
cost to morphological theory per se.
BOOIJ, GEERT (2010). Construction Morphology. Oxford: Oxford University Press.
BOOIJ, GEERT (2012). Inheritance in Construction Morphology. Paper presented at the workshop on Default
Inheritance, U Kentucky, Lexington Ky, May 21-22.
BOOIJ GEERT, and MASINI, FRANCESCA (in preparation). The role of second order schemas in the construction of
complex words. ms.
CULICOVER, PETER W., and JACKENDOFF, RAY (2012). Same-Except: A domain-general cognitive relation and how
language expresses it. Language 88, 305-340.
JACKENDOFF, RAY (1975). Semantic and morphological regularities in the lexicon. Language 51, 639-671.
KAPATSINSKI, VSEVOLOD (2013). Conspiring to Mean: Experimental and Computational Evidence for a UsageBased Harmonic Approach to Morphophonology. Language 89,110-48.
KOEFOED, GEERT and VAN MARLE, JAAP (1980). Over Humboldtiaanse taalveranderingen, morfologie en de
creativiteit van taal. Spektator 10: 111–147.
NESSET, TORE (2006). Second-Order Schemas and Active Subschemas: Vowel Reduction in Cognitive Grammar.
Poljarnyj Vestnik 9.
Floating Morphological Paradigms in Seri
Surrey Morphology Group, University of Surrey
[email protected]
The inflectional class system of verbs in Seri (a language isolate of Sonora State, Mexico;
Marlett 2009-11) displays a degree of complexity that rivals or surpasses any that has ever
been described. This complexity is concentrated in the four-cell suffix paradigm that
distinguishes subject number (singular ~ plural) and aspect (perfective ~ imperfective, or
effectively single event ~ multiple event). Each cell displays considerable allomorphy (1),
with the result that in a corpus of 1001 lexemes taken from Moser & Marlett’s (2010)
dictionary, there c. 244 distinct paradigm types. There is very little internal predictability
between the allomorphs (2). Semantic, syntactic, phonological and morphological classes
play little or no observable role in determining the form of the suffixes. Compounding this
cross-lexemic unpredictability, few of the suffixes have a fixed function; e.g. in (3a) the
distribution of the suffixes -(t)ox and -(t)am is the reverse of (3b) (The presence or absence of
suffix-initial /t/ is phonologically predictable; Marlett 2009-11). On the face of it, it looks as
if the four basic forms must be lexically listed.
The nature of the system is revealed by comparing the verbal paradigm to the nominal
paradigm. The verbal suffixes are nearly all drawn from the repertoire of nominal plurals (4).
The reason for this is that both subject plurality and event plurality (imperfective aspect) are
treated morphologically as simply as plural, their markers drawn from a common set of plural
markers, so that there is no overarching morphological distinction between three of the four
cells of the paradigm. Nevertheless, a certain systematicity can be observed. Given the
morphosyntactic hierarchy:
there are more or less rigid suffixal hierarchies that map onto this, but differ in their entry
point. E.g. the suffixal hierarchy -t ~ -(t)ox ~ -(t)oɬka maps onto both (5a) and (5b), but the
entry point for (5b) is ‘later’, and thus the paradigm only gets as far as -(t)ox. These differing
entry points account for a portion of the complexity of the system. Since the hierarchy has an
endpoint, some local implications emerge. E.g. the suffix -(t)oɬka is the end of the line, thus
whatever point it enters the paradigm, it must be retained to the end, so that it appears in the
last (6a), last two (6b) or last three cells (6c) of the paradigm.
This shows that in order to find the underlying organization of the Seri paradigm, we
need to completely separate the morphological paradigm from the morphosyntactic paradigm
(Sadler & Spencer 2001, Stump 2006). The morphological paradigm can then be described as
a series of ordered forms without reference to the morphosyntactic values. The system is still
complex, but within a typologically familiar range. The real complexity emerges when the
morphological paradigm is mapped onto the morphosyntactic paradigm, an operation which
itself falls into various subtypes. The resulting mismatch, though allowed by a realizational
framework, is quite remarkable typologically.
Examples of suffixal allomorphy
‘make small’
‘make into a fence’
‘be pregnant
‘be connected’
‘cover o.s.’
Seri compared to the languages discussed by Ackerman & Malouf (2013)
average conditional entropy
1.96 (with frequency factored in)
a. ‘shoot’
b. ‘lap up’
‘go away’
a. ‘go away’
b. ‘be used up’
a. ‘go away’
b. ‘die’
c. ‘make soft’
Nominals with the suffix –ač in Croatian
Institute of Croatian Language and Linguistics
[email protected]
Although the interplay between nominal and verbal properties in deverbal and similar nouns
has been widely studied and recognized as an important morphological issue in the literature
and within different theoretical research, I will try to approach this interesting topic by
investigating the derivation of agentive/instrumental nouns with the suffix -ač in the Croatian
The suffix -ač is mostly a deverbal suffix from which nouns with agentive and/or instrumental
interpretation are derived. Three main properties of the suffix -ač are commonly known
(Babić 1991). Firstly, -ač mainly selects deverbal bases; secondly, it is mostly attached to
imperfective bases and thirdly, its interpretation varies between agentive and instrumental.
Birtić (2008) describes a fourth property of the suffix -ač: similar to the English suffix -er, it
can attach to unergative intransitive bases without restrictions, and mostly not to unaccusative
intransitive bases (although it can attach to all transitive bases):
1. pliv-ač ‘swimmer’
2. *ton-ač ‘one who sinks’
It has been claimed that English -er nominals correspond to the external argument of the base
verb (Fabb 1992, Rappaport Hovav & Levin 1992, Marantz 2001), but other research shows
that this is not completely accurate (Lieber 2004).
Firstly, I will try to determine, on the basis of a large corpus (Croatian Language Repository),
the correctness of the claim that the suffix -ač selects only verbs with the external arguments.
Apparently, there are counterexamples, such as the noun padač 'one who falls down', and I
will attempt to identify other meanings of nouns with this suffix.
Secondly, my aim is to investigate whether the aspect of the verbal base (perfective or
imperfective) has any impact on the interpretation of nouns with -ač (agentive, instrumental
or other). Although the suffix -ač mostly selects imperfective verbal bases, there are some
exceptions to the rule:
3. osigurač ‘fuse’ < osigurati (perf.) ‘secure, insure’
The fact that the lexical aspect of the base verb is recognizable in the derived noun (but does
not have the meaning 'perfective' or 'imperfective' as in the case of -nje nouns) and that the
categories of perfectivity/imperfectivity play a role in the selectional restrictions of this suffix
will serve as guidelines for the coming discussion of the syntactic vs. the lexicalist approach
to morphology.
BABIĆ, S. 1991. Tvorba riječi u hrvatskom književnom jeziku. Zagreb: Hrvatska akademija znanosti i umjetnosti
– Globus • BIRTIĆ, M. 2008. Unutarnja struktura odglagolskih imenica u hrvatskome jeziku. Zagreb: Institut za
hrvatski jezik i jezikoslovlje. • FABB, N. 1988. English suffixation is constrained only by selectional restrictions.
Natural Language and Linguistic Theory, 6, 527–539. • LIEBER, R. 2004. Morphology and Lexical Semantics.
Cambridge: Cambridge University Press. · MARANTZ, A. 2001. Words, abstract, WCCFL 20, USCL, 2/2001.
•RAPPAPORT HOVAV, M. & B. LEVIN 1992. -ER Nominals: Implications for the theory of Argument structure.
Syntax and Semantics 26, 127–153.
The role of second order schemas in word formation
Leiden University
[email protected]
Word formation can be characterized in terms of schemas that specify the systematic
relationships between form and meaning of complex words. These schemas are based on
paradigmatic relationships between sets of words. For a proper characterization of word
formation we also need second order schemas, that is, schemas that are paradigmatically
related. For example, the relation between English nouns of the form [x-ism] and those of the
form [x-ist] is such that the meaning of the words in –ist is often a compositional function of
the words in –ism, even thought the morpheme –ism is lacking in the words in –ist. Hence we
need a second order schema of the following type (Booij 2010):
<[x-ism]Ni ↔ SEMi > ≈ < [x-ist]j ↔ [person involved in SEMi]j>
The necessity of second order schemas has also been argued for by Kapatsinski (2013) and
Booij (2010, to appear) for inflection, and by Nesset (2008) for phonology.
In my presentation, I will present various types of evidence for the necessity of second order
schemas for a proper account of word formation processes., with a focus on Dutch word
formation processes such as the formation of elative compounds and the nominalization of
particle verbs.
BOOIJ, GEERT. 2010. Construction morphology. Oxford: Oxford University Press.
BOOIJ, GEERT. to appear. Construction Morphology. In: A. Spencer and A. Zwicky (eds.), The Handbook of
Morphology, second edition. London: Blackwell [downloadable from http://geertbooij.word.press.com].
KAPATSINSKI, VSEVOLOD. 2013. Conspiring to mean: experimental and computational evidence for a usagebased harmonic approach to morphophonology. Language 89.110-48.
NESSET, TORE. 2008. Abstract phonology in a concrete model: Cognitive linguistics and the morphologyphonology interface. Berlin: Mouton de Gruyter.
Incidental acquisition of (ir)regularity from syntactically simple and complex texts in
L1 and L2 German
Leipzig University
[email protected]
Though L2 grammar acquisition has been in the focus of researchers for several decades,
surprisingly little is known about its incidental acquisition during reading. It has been claimed
that incidental acquisition substantially contributes to the vocabulary growth both in L1 and
L2 (Grabe, 2009). However, despite the general consensus that "knowing a word involves
more than being able to recall the meaning (or L1 translation) of a presented word form" (Nation,
1994, p. 121), little research has gone beyond the acquisition of the meaning of new words. In
the present study, we use a novel experimental paradigm to explore the incidental
acquisition of (ir)regularity status of novel verbs in syntactically simple and complex contexts.
The results are compared to those on acquisition of subcategorization (SC) - transitive vs.
intransitive verbs. Differences in the result patterns of L1 and L2 participants as well as the
influence of syntactic complexity are discussed.
Method: 72 advanced learners of German and 80 native Germans read 20 short texts (plus
fillers), each with one occurrence of a novel verb in infinitive and two in preterit (e.g. belfen balf). The novel verb (constructed according to German phonotactic rules) was replacing a low
frequency existing German verb and its meaning could be derived from the context. Half of the
novel verbs were conjugated regularly, the other half irregularly. The syntactic complexity of the
texts was manipulated yielding two different versions of otherwise identical texts: one
syntactically simple and the other syntactically complex (the latter containing longer
sentences with more embedded structures). After each text, several additional sentences were
presented. One of them contained the novel verb in perfect tense, conjugated either consistently
with the forms occurred in the text (e.g. irregular hat gebolfen, plausible condition), or differently
(e.g. regular hat gebelft, implausible condition). Participants read these sentences in a self-paced
manner with a moving window. The L1 and L2 experiments on SC were constructed
analogically: The texts contained intransitively used novel verbs and in the implausible self-paced
reading condition these verbs were then presented as transitive with the direct object “es” (it).
Results and Discussion: In the SC experiment, both L1 and L2 participants acquired the SC
frame from the texts. Reading times were slower in the implausible condition reflecting
participants’ sensitivity to the SC violation. The implausibility effect in the L2 experiment
appeared later (at the spill-over region) and only if the novel verb was presented in a syntactically
complex context. In the Irregularity experiment, the results of the L2 participants are
similar to those on SC: The implausibility effect was observed only in the complex condition
when the novel word was presented as irregular. We assume that when syntactic complexity leads
to difficulties in reading comprehension (as in the case of the L2 learners), it triggers
“conscious registration of attended specific instances of language”, which corresponds to
Schmidt’s (2012:32) definition of noticing. We further assume that noticing can be triggered also
by irregular or otherwise phonologically or morphologically conspicuous word forms. For
advanced learners who are familiar with the German verb system, morphological irregularity can
trigger noticing, since advanced learners know that acquisition of irregular verbs requires
additional effort including memorization. Being thus sensitized to irregular verbs as a
“learning problem”, learners attend to them more closely when encountering them during
reading. Along these lines, both the claim that irregular forms are salient and are not salient can
be valid: They are salient in the sense that they are morphologically conspicuous to
learners on advanced levels and thus attract their attention (cf. Godfroid & Uggen, 2013 for
beginning learners), but they are possibly not salient with respect to the particular
phonological changes that occur on their stems. The pattern of the results of the L1 participants
gave no indication that they acquired the regularity status of the novel verbs in the texts since
reading times were not slower in the implausible condition in general. Instead, they responded
with slower reading times and higher outlier rates whenever the presented novel verb form in the
self-paced sentence was an irregular one – irrespective of the regularity status of the verb in the
previous text. This sharply contrasts with the results on SC, when the L1 participants reacted with
slower reading times in the spillover regions when the SC frame of the novel verb was violated
in the implausible condition compared to how the novel verb was presented in the text. We
assume that this difference is due to the differences in the linguistic characteristics of the two
grammatical features: Different theoretical frameworks converge on the assumption that regular
and irregular verbs are stored and processed differently and assume that the regular conjugation is
the default type. So called “dual-mechanism models” implement this idea in form of two
principally different systems: The default system splits regular forms into their stem
morphemes and affixes (rule based route) and the memory-based system stores and retrieves all
exceptions to the default as whole forms (lexical route) (e.g. Dual-Route Model: Clahsen, 1999).
On the other hand, most of the studies on SC do not propose that one type of verbs (transitive vs.
intransitive) would be expected as the default representation. If they make such assumptions, then
transitive verbs are considered the default class: Syntax always provides a complement
position for verbs, but the verb’s selectional properties determine whether it is filled by
an explicit argument (Cummins & Roberge 2005). In our study we saw that both L1 and L2
participants readily acquired the intransitive feature of the novel verbs and perceived the
transitive usage as a violation, supporting the claims that intransitive and monotranisitive verbs
do not essentially differ in their prominence. We further observed that while the L2 learners can
acquire the (ir)regularity status of novel irregular verbs and perceive their regularization as
violation, native German speakers perceive as violation any novel irregular form and seem to
classify all new verbs as regular irrespective of the evidence that they receive through input. We
interpret this finding in terms of a “learning by unlearning” effect: The L1 participants learnt that
regular conjugation is a productive, default type and that the set of irregular verbs is a rather
small, closed group of verbs whose all members they already know. Having learnt this, they stop
acquiring regularity from texts and instead assume per default that all new verbs are
regular. If an unknown irregular form appears, they consider it implausible. Seemingly
paradoxically, the reduced amount of knowledge and experience that the L2 learners have, keeps
them more open to turning “input into intake” and to learning new forms of all types. On the
other hand, the more profound previously acquired knowledge of L1 native speakers makes them
much more reluctant to acquire idiosyncratic morphosyntactic features of novel words.
CLAHSEN, H. (1999). Lexical entries and rules of language: A multidisciplinary study of German inflection.
Behavioral and brain sciences, 22(6), 991–1013. • CUMMINS, S., & ROBERGE, Y. (2005). A modular account of null
objects in French. Syntax, 8(1), 44–64. • GODFROID, A., & UGGEN, M. S. (2013). Attention to irregular verbs by
beginning learners of German. Studies in Second Language Acquisition, 35(02), 291–322. • GRABE, W. (2009).
Reading in a Second Language. Cambridge, New York:Cambridge University Press. • NATION, P. (1994). New
Ways in Teaching Vocabulary. New Ways in TESOL Series: Innovative Classroom Techniques. ERIC. •
SCHMIDT, R. (2012). Attention, awareness, and individual differences in language learning. Perspectives on
Individual Characteristics and Foreign Language Education, 6, 27-50
Tel-Aviv University
Bar-Ilan University
[email protected]
[email protected]
This talk examines the strategies of the morphological adaptation of English adjectives that
are borrowed into Modern Hebrew (MH). Borrowed words can undergo both morphological
and phonological adaptation. In morphological adaptation, the borrowed word receives
morphological features of the target language, i.e. one of the native word formation strategies
is applied (Schwarzwald 1998, 2013). We argue that borrowed adjectives are an intermediate
category, applying morphological adaptation strategies of both verbs and nouns.
Borrowed adjectives demonstrate three main patterns of adaptation (Ravid 1992).
Linear formation: Borrowed adjectives tend to end with the suffix –i. This is done by three
main strategies.
1. Suffixation: Borrowed adjectives receive the native MH suffix –i, although some of
them already end with an English adjectival suffix (efectivi 'effective', noʁmali
2. Truncation: Adjectives that end with the English suffix –ic or –ate undergo deletion of
this suffix or part of it, resulting in an adjective that ends with –i (stati 'static').
3. Zero conversion: Adjectives that end with the English suffix –i do not undergo any
morphological process (elementaʁi 'elementary', kitʃi 'kitchy').
These three patterns conspire towards the same goal: having an adjective that ends with –i,
making it appear as a MH adjective. This is because the suffix –i is one of the most
productive word processes in MH adjective formation (e.g. jaldut 'childhood' jalduti
'childish' (Ravid & Shlezinger 1987)).
Non-concatenative formation: The participle meCuCaC form is used both for the formation
of adjectives and the present forms of CuCaC passive verbs. Some foreign adjectives are
formed directly in this template, e.g. medupʁas 'depressed' and meʃnutsal 'having the shape of
a schnitzel', mefuksal 'pixellated'. Such formations are independent of verb formation; for
example, the adjective meʃnutsal has no verbal counterpart (*ʃnitsel 'give X the form of a
schnitzel'). This is also true for MH adjectives that have no verbal counterpart, e.g. menumas
'polite' but *numas, *nimes.
Zero conversion: There are adjectives that are borrowed as is without any morphological
adaptation. These are mostly adjectives that end with –ing or –ed (e.g. ameziŋg 'amazing',
dedikeited 'dedicated') but also other forms like laʁdʒ 'large' and kul 'cool'. Note that unlike
the forms in (1c), these do not end with –i like MH adjectives. In addition, unlike foreign
adjectives that undergo morphological adaptation, these adjectives are also not inflected for
gender and number, but remain morphologically frozen. The adjective dedikeited, for
example, is used for feminine and does not take the suffix –it (*dedikeitedit), the feminine
suffix typical of the inflection of borrowed words.
Conclusions: Borrowed adjectives function as an intermediate category between nouns and
verbs, as they are the only group where such a wide array of formation strategies are found.
Verbs are formed exclusively by non-concatenative morphology, as every verb in the
language must conform to one of the existing verbal templates (Berman 1978, Bolozky 1978),
e.g. fikses 'fax' in the pi'el template and hiklik 'click' in the hif'il template. Nouns are mostly
borrowed as is, undergoing only phonological adaptation, e.g. templet 'template' and neʁativ
'narrative' (Cohen 2010). Adjectives, as we have shown, demonstrate different patterns: taking
a MH suffix, similarly to MH adjectives, templatic formation, similarly to verbs and
borrowing with no morphological adaptation, similarly to borrowed nouns.
This correlates with the behaviour of adjectives as an intermediate category in general (Smith
2001, 2011, Anttila 2002, Bat-El 2008). Previous studies have shown that adjectives tend to
share phonological and morphological properties and processes that are typical to both verbs
and nouns. Morphological adaptation appears to be scalar with respect to adjectives, from a
highly typical MH formation to no morphological adaptation whatsoever.
Selected References
ANTTILA, A. 2002. Morphologically conditioned phonological alternations. Natural Language and Linguistic
Theory 20:1-42.
BAT-EL, O. 2008. Morphologically conditioned V–Ø alternation in Hebrew: Distinction among nouns, adjectives
and participles, and verbs. In S. Armon-Lotem, G. Danon, and S. Rothstein (eds.) Current Issues in
Generative Hebrew Linguistics. Amsterdam: John Benjamins. pp. 27-60.
BERMAN, R. A. 1978. Modern Hebrew Structure. Tel-Aviv: University Publishing Projects.
BOLOZKY, S. 1978. Word formation strategies in MH verb system: denominative verbs. Afroasiatic Linguistics
COHEN, E. G. 2010. The Role of Similarity in Phonology: Evidence from Loanword Adaptation in Hebrew.
Doctoral dissertation, Tel-Aviv University.
RAVID, D. 1992. The absorption of foreign words in the nominal and verbal systems. In M. Muchnik (ed) The
foreign Influences on Contemporary Hebrew. Tel Aviv: The Open University of Israel. pp. 11-20.. [in
RAVID, D. AND Y. SHLESINGER. 1987. On the classification and derivation of adjectives with the suffix -i in
Modern Hebrew. Hebrew Computational Linguistics 25:59-70.
SCHWARZWALD, O. R. 1998. Word foreignness in Modern Hebrew. Hebrew Studies 39:115-142.
SCHWARZWALD, O.R. 2013. The Typology of Nonintegrated Words in Hebrew. SKASE Journal of Theoretical
Linguistics 10(1):41-53.
SMITH, J. L. 2001. Lexical category and phonological contrast. In R. Kirchner, J. Pater and W. Wikely (eds.),
Papers in Experimental and Theoretical Linguistics 6: Workshop on the Lexicon in Phonetics and
Phonology. Edmonton: University of Alberta. pp. 61-72.
SMITH, J. L. 2011. Category-specific effects. In M. van Oostendorp, C. Ewen, B. Hume and K. Rice (eds.), The
Blackwell Companion to Phonology. Malden: Wiley-Blackwell. pp. 2439-2463.
Categorial multifunctionality and stress-(in)sensitivity of affixes
Universiteit van Amsterdam
[email protected], [email protected], [email protected]
In the literature derivational affixes have been viewed in different ways. In lexical models of
morphology affixes are analyzed as elements on a par with stems: they are phonologically
and semantically specified and have a syntactic category (a.o. Lieber 1980, Strauss 1982,
Kiparsky 1982). In syntactic approaches of morphology (e.g. Distributed Morphology) the
syntactic information of affixes is separated from their phonological realization. Affixes are
thus seen as the mere spell-out of morphosyntactic information. Within this theoretical
framework it has been proposed quite recently that derivational affixes have no syntactic
information whatsoever, reducing them to categoriless ‘roots’ (Lowenstamm 2010, De Belder
2011). We claim that these latter approaches are not precise enough and consequently,
empirically insufficient; instead, we propose that only a subset of the derivational affixes is
indeed not specified for categorial information (i.e. roots) - implying that these affixes are
syntactically multifunctional - whereas the second type of affixes is the realization of a
syntactic head. This distinction correlates with stress-behavior of these affixes and their
distributional properties.
Lexical models explain the distinction between stress-sensitive affixes and stressneutral affixes (SPE), as well as the so-called Affix Ordering Generalization (Siegel 1974) in
terms of level-ordering (Kiparsky 1982). However, such models have received critique in the
late eighties (e.g. Fabb 1988, Gussmann 1988), pointing at both empirical and theoretical
problems. Moreover, it turns out that these models are problematic for the syntactically
multifunctional affixes that seem to realize different syntactic heads. For if affixes are
specified for a syntactic category, it is not expected that they can be multifunctional in the
category they denote (cf.(1a) as observed by Lowenstamm 2010 for English and (1b) by De
Belder 2011 for Dutch).
Syntactic models of morphology can more easily account for the observation in (1),
because in such approaches affixes are roots, i.e. underspecified elements not realizing
syntactic categorial information. Therefore, affixes occur wherever we may expect roots to
occur. However, in such syntactic approaches it is not clear how a difference can be made
between stress-sensitive and stress-neutral affixes. Lowenstamm (2010) has tried to account
for this difference in a distributed approach; a key-element in his analysis is that all affixes
are roots not denoting any categorial information. The different stress-behavior follows
from a difference in feature-specification on these affixal roots.
This paper aims to show that an interesting improvement of Lowenstamm’s
analysis is possible. We provide a syntactic analysis (couched in terms of Distributed
Morphology) for both the (non-)flexibility and the stress-behavior of affixes and propose
that there are in fact two different types of derivational affixes: affixes that are unspecified
for a category, thus roots (‘root-affixes’) and affixes that spell out a categorial head (‘headaffixes’). The present hypothesis predicts a correlation between the flexibility of affixes and
the stress-sensitivity of affixes: root-affixes can be multifunctional and are always stresssensitive, whereas head-affixes are always non-multifunctional and stress-neutral. Root-
affixes are in the same, first phase as other root elements and therefore are predicted to be
stress-sensitive. Head-affixes are in a different phase, because they are the realization of a
categorial head, thus in a phase determining node (Marantz 1997), and cannot influence the
stress-pattern of the word they attach to, see (2).
We show that this hypothesis makes the correct predictions with regard to both the English
and Dutch derivational affixes. This paper further shows that stress-sensitive affixes (being
roots) can only attach to (complex) roots, and stress-neutral affixes (being the spell-out of
categorial heads) cannot occur inside a root-phrase. Finally, it turns out that there is a third
type of affix that does contain categorial information ánd is stress-sensitive. At first sight this
type of affixes seems to contradict our hypothesis since it breaks down the correlation
between stress-sensitivity and flexibility. However, a closer look at the interaction between
phasal heads and the interpretative components in fact predicts the existence of such elements:
these affixes (‘first heads’) are always the first categorial head above the root phrase and
precisely this structural position has a somewhat special status in a particular theory of phasal
spell-out (Embick 2010) that we adopt.
Our analysis shows that there are three types of derivational affixes, whereas previous
approaches (lexical approaches following SPE) only recognized two types of affixes.
Moreover, these approaches failed to see a correlation between stress-sensitivity and multicategoriality. The division in root-affixes, head-affixes and ‘first heads’ gives a better insight
in the different behavior of derivational affixes. This three-way distinction is problematic both
for level-ordering approaches and for the recent proposals within Distributed Morphology. In
our proposal the existence of three types of affixes comes out as the only possible option,
given the interaction between syntax and spell-out.
BELDER, MARIJKE DE (2011) Roots and Affixes, eliminating Lexical Categories from Syntax. LOT: Utrecht.
CHOMSKY, NOAM AND MORRIS HALLE (1968) The Sound Pattern of English. New York: Harper and Row.
EMBICK, DAVID (2010) Localism versus globalism in morphology and phonology, Cambridge: MIT Press.
FABB, NIGEL (1988) ‘English suffixation is constrained only by selectional restrictions’, Natural
Language and Linguistic Theory 6, p. 527 – 539.
GUSSMANN, EDMUND (1988) Review of Mohanan (1986). Journal of Linguistics 24: 232-239.
KIPARSKY, PAUL (1982) ‘From Cyclic Phonology to Lexical Phonology’, in: Hulst, H. van der and N. Smith (eds.)
The Structure of Phonological Representations (I), p. 131-175.
LIEBER, ROCHELLE (1980) On the Organization of the Lexicon, PhD diss. Univ. of New Hampshire, rep.
by IULC.
LOWENSTAMM, JEAN (2010) Derivational affixes as roots (phasal spellout meets English stress
shift). Ms.,Univeristé Paris-Diderot & CNRS.
MARANTZ, ALEC (1997) ‘No Escape from Syntax: Don’t try morphological analysis in the privacy of your
own lexicon’, in: Dimitriadis, A. et al. (eds.) Proceedings of the 21st Annual Penn Linguistics
Colloquium: Penn Working Papers in Linguistics 4, p. 201-225.
STRAUSS, STEVEN (1982) Lexicalist Phonology of English and German, Foris: Dordrecht.
Deconstructing exuberant exponence in Batsbi
CNRS, Laboratoire de linguistique formelle, U Paris-Diderot
[email protected]
Extended exponence provides one of the core arguments in favour of word-and-paradigm approaches (Matthews, 1974). Harris (2009) has placed the case of Batsbi exuberant exponence
on the research agenda, a phenomenon which constitutes a particularly challenging case of multiple extended exponence where one and the same marker expressing the same morphosyntactic
property may wind up being realised up to 4 times in a morphological word. According to Harris,
class markers in Batsbi define a paradigm with 16 cells, expressing combinations of 8 genders
in 2 numbers (singular/plural). Choice of class markers is controlled by an ergative pattern, i.e.
they mark gender/number of intransitive subjects and transitive direct objects. Furthermore, the
properties this marker expresses may overlap person/number agreement in 1st and 2nd person.
Morphotactically, these class markers can surface in several positions within a complex word:
immediately preceding (i) the root (or roots in the case of compounding), (ii) the transitive/intransitive suffixes -i, -al, -is, and (iii) evidential suffixes. Presence vs. absence of the marker on the root
is morphologically conditioned: while all roots featuring the (consonantal) markers are essentially
vowel-initial, there are still roots that match the phonological condition yet fail to take the marker.
Similarly, presence of post-root class markers is licensed exclusively by the suffixes listed above,
to which they are left adjacent. Thus, presence of class markers is morphotactically conditioned
by the class of their right-adjacent morph, a suitable root or suffix. As a consequence of which
suffixes and stems appear in a complex word, one may find one of four logically possible situations
(see Harris, 2009, tab. 2, p. 276): both pre-stem and pre-suffix markers (1), pre-stem markers and
no suffix markers (2), no pre-stem markers but pre-suffix markers (3), as well as no class markers
at all (4).
horse(b/d)-OBL-ERG saddlebags(/d)-PL.ABS CM-fall-CM-TR-AOR
‘The horse threw off the saddlebags.’
(Harris, 2009, p. 274)
(2) xen-go-ħ
tree-ALL-LOC leaf(d/d)-PL.ABS CM-fall-AOR
‘The leaves of the tree were falling.’
(Harris, 2009, p. 274)
(3) kuyrc’l-e-x
qečqečnayre daq’r-i
wedding-OBL-CON various
food(d/d)-PL.ABS go-CM-TR-AOR above
‘At the wedding [they] passed around various foods.’
(Harris, 2009, p. 274)
(4) k’alam
pen(d/d).ABS change-AOR-1SG.ERG
‘I change my pen.’
(Harris, 2009, p. 275)
Based on the systematic application of a series of diagnostic tests (word-final reduction, choice
of agreement trigger, intervention of particles, coordination and gapping), Harris (2009) shows
convincingly that exuberant marking does indeed occur within a single morphological word. She
further emphasises the problems exuberant extended exponence presents for lexical and incremental theories of inflection. Harris (2009) therefore proposes a word-and paradigm approach to
Batsbi exuberant exponence that crucially relies on the notion of exemplars and schemata. Under her approach, combinations of class markers with stems and affixes are stored as paradigms,
and abstracted into schemata such as [d] for morphs with a dependent class marker and
] for those without. In order to capture the relatedness between cells in these paradigms,
Harris builds on the concept of analogy to derive new forms from existing listed paradigms of
inflected words. However, under this exemplar-based solution the shared morphotactic properties,
as well as the identity of shapes are pictured as coincidental, placing these generalisations outside
the scope of the grammar.
In this talk, I shall investigate how a constructional approach to the introduction of exhuberant
marking can be formalised in such a way, as to explicitly integrate into the grammar the description
of shared properties of exponence and morphotactics. Moreover, under the approach I shall adopt
no paradigms of fully inflected words must be listed, but they can rather be inferred on the fly
by means of systematic combination of underspecified descriptions, thereby constructing complex
expressions from more primitive ones. To this end, I shall build on recent proposals by Bonami
and Crysmann (2013) who propose an inferential-realisational model of morphology that crucially
makes use of online type construction (Koenig, 1999) by means of cross-classification between
orthogonal dimensions.
 
gend 
,… 
num  
⇌] [
 
1  
,... 
sg 
⇌ ] [ ]⟩
{ }
[ ]⟩
 
1  
,... 
pl 
⇌ ] [ ]⟩
] }
aor ,...
⇌ 
en 
{[ ] }
tr ,...
⟨[ ] 
, 
{ } ⟨[ ]⟩
... ⇌
⟨ 
] }
,... ⇌ ...
1 
{ } ⟨[ ] [ ]⟩
... ⇌ ,
ek’ · · ·
The main intuition behind the present analysis is that class markers are syntagmatically dependent: as illustrated by the type hierarchy of realisation rules, rules of exponence are distinguished
as to whether they introduce a single morph (as with the aorist), or rather two morphs (as with the
transitive), where only the shape of the second morph is constrained. Similarly, stem introduction
rules are equally segregated into an open (default) class that exclusively introduces stem phonology ( ) and position class ( ), and a closed class that requires the presence of a pre-stem marker,
again, without constraining that morph’s identity. In order to form a well-formed realisation rule
(cf. Koenig, 1999; Bonami and Crysmann, 2013), types from the EXPONENCE dimension must
be combined, under unification, with a compatible type from the MORPHOTACTICS dimension,
which provides a default type for single morph introduction, as well as types for combinations
of multiple morphs. As shown in (5), these latter types constrain the shape of the initial morph,
together with an adjacency requirement, couched in terms of position class ( ) indices. Breaking down morph combinations into syntagmatically dependent realisation rules directly expresses
generalisations about shape and position of individual morphs, whereas their combinatorial aspect
is handled by the logic of orthogonal dimensions. Finally, given the inferential-realisational character of the underlying morphological theory, extended, or even exhuberant exponence can be
accounted for straightforwardly.
To conclude, I have shown how the organisation of realisation rules into multi-dimensional type
hierarchies lends itself quite naturally to capture the dependent nature of exuberant exponence
in Batsbi: by abstracting out the introduction of class markers into a dimension of its own, we
can account for both morphotactics, i.e. the dependent nature of the markers, and the pairing
of morphosyntactic features with exponents. Thus, online construction of realisational rules from
underspecified rule types makes it possible to deconstruct the class marking construction into its
constituting parts, providing a formal grammar-internal interpretation of analogy.
Bonami, O. and Crysmann, B. (2013). ‘Morphotactics in an information-based model of realisational morphology’. In S. Müller (ed.), Proc. of the 20th Intl’ Conference on HPSG, Berlin. 27–47.
• Harris, A. C. (2009). ‘Exuberant exponence in Batsbi’. Natural Language and Linguistic Theory,
27:267–303. • Koenig, J.-P. (1999). Lexical Relations. Stanford: CSLI publications. • Matthews,
P. H. (1974). Morphology: An Introduction to the Theory of Word-Structure. Cambridge: Cambridge University Press.
Opacity, lexicalisation, recomposition and phonological adjustment in prefixation
Pázmány Péter Catholic University & Research Institute for Linguistics, H.A.S.
[email protected]
This paper discusses certain morphophonological phenomena displayed by prefixed forms in
Latin, in particular the role of internal factors (e.g. sonority) vs. external factors (e.g.
frequency) in assimilations at morphological boundaries, the scope of regularities such as the
Place Condition with respect to the same boundaries, and specific dissimilatory tendencies
(e.g. that of [r]) that determine the possibility of certain morphologically complex forms.
In terms of productivity, transparency and phonological interference, prefixed forms
can be arranged on a cline with strongly lexicalised and opaque forms at one end and
transparent formations at the other. The difference can be exemplified with dēgere ‘to live’
(usu. with an object like ‘life’ or ‘time’), which is composed of dē ‘from’ + agere ‘to do,
drive’, vs. perpolitus ‘highly polished’, in which the adjective politus ‘polished’ is combined
with per, which has two meanings, ‘through’ and ‘highly, very’. The transparent nature of
perpolitus is not only seen in its semantics and in the lack of phonological interference
between the two component parts, but also in some instances of its use, where it is used next
to other per-prefixed adjectives with similar meanings. This cline is related, though certainly
not in an isomorphic fashion, to the diachronic emergence of these forms (for data see Prinz
1949–50 and 1953, Leumann 1977:181–219, Buck 1899, García González 1996).
Prefixation led in many cases to lexicalisation, which in turn resulted in drastic
phonological modifications at the prefix–stem boundary as well as within the stem. The pace
and the extent of lexicalisation, however, was highly variable. Furthermore, prefixation also
involved recomposition in all periods of the documented history of Latin. An early case of
recomposition is seen in perjūrare ‘to forswear’, which is the recomposed variant of the older
form peierare ([pejjera˘re], same meaning). Later recompositions can be reconstructed on the
evidence of Romance languages; it is well known that reflexes of forms like rétinet ‘he keeps’
(< re + tenet) often derive not from the inherited Classical Latin forms but from recomposed
variants such as *reténet (> Fr retient etc.).
The varying pace of lexicalisation and the varying degree of transparency coupled
with the phonological processes that took place at prefix–stem boundaries resulted in an often
indirect relation between written forms and probable phonological variants. This is further
complicated by the fact that beginning with the 1st century AD etymologically oriented habits
of spelling began to gain ground, but did not affect all words of a similar composition to the
same extent. Since, however, many of the characteristic traits of the variation that appear in
the texts are clearly phonologically based, it can be assumed that the picture is not badly
distorted — that is, with all the necessary provisos.
The analytical goal of the paper is to flesh out the observations introduced above (the
role of frequency, place of articulation and sonority in morphophonological processes at
boundaries) and the scope of regularities involving assimilation and dissimilation with respect
to morphological boundaries. The conceptual goal of the paper is to explore the relation
between lexicalisation, recomposition and phonological adjustment in the history of prefixed
forms in Latin.
BUCK, CARL DARLING (1899) Notes on Latin orthography. The Classical Review 13:116–9, 156–167.
GARCÍA GONZÁLEZ, JUAN JOSÉ (1996) Asimilación de prefijos en inscripciones latinas. In: A. Bammesberger, F.
Heberlein (eds) Akten des VIII. Internationalen Kolloquiums zur Lateinischen Linguistik. Heidelberg: Winter.
LEUMANN, MANU (1977) Lateinische Laut- und Formenlehre. München: Beck.
PRINZ, OTTO (1949–50) Zur Präfixassimilation im antiken und im frühmittelalterlichen latein. Archivum Latinitatis
Mediae Aetatis 21.87–115.
PRINZ, OTTO (1953) Zur Präfixassimilation im antiken und im frühmittelalterlichen latein. Archivum Latinitatis
Mediae Aetatis 23.35–60.
On present subjunctive inflection, between morphosyntax and pure morphology:
some clues from Ibero-Romance dialects
Università degli Studi di Padova
[email protected]
The affixal inflection of the present subjunctive displays some unexpected developments in
some dialectal varieties of Ibero-Romance. In providing a possible interpretation of data, this
contribution casts light on the kind of information, both morphosyntactic and purely
morphological in nature, which should be assumed to be associated with present subjunctive
affixes in the verbal systems of the observed varieties.
In the analysis of Spanish verbs, present subjunctive forms are traditionally regarded as
involving a Tense/(Aspect)/Mood marker, whose actual shape may be influenced by
inflectional class so as to give rise to a binary opposition. Thus first conjugation verbs,
characterised by the affix –e, are distinguished from second and third conjugation verbs,
sharing the affix –a (cf., inter alia, Alcoba 1999, Aguirre and Dressler 2001): Table 1.
In Carstairs-McCarthy’s 1994 terms, the marker –e can therefore be analysed as a ‘classidentifier’, while –a would be the ‘class-default’.
Dialectal varieties may be found to depart from this binary opposition according to either of
two almost opposite types.
The so-called ‘Type-A’ varieties show a tendency towards neutralization of the binary
opposition, such that some present subjunctive cells end up displaying one affix for the
inflection of all verbs, regardless of conjugation (a ‘superstable marker’, Wurzel 1989). In the
variety of San Ciprián de Sanabria (Krüger 1923), for instance, the affixes –jes and –jen are
invariably used for the inflection of 2nd singular and 3rd pl(ural) present subjunctive forms,
respectively: e.g. first conj(ugation) cánt-jes, cánt-jen ‘that you (sg.)/they sing’, second conj.
vénd-jes, vénd-jen ‘that you (sg.)/they sell’, third conj. dúrm-jes, dúrm-jen ‘that you (sg.)/they
sleep’, as opposed to, for instance, 1st singular cánt-i versus vénd-a, dúrm-a.
In ‘Type-B’ varieties, on the other hand, not only is the binary opposition described above
basically retained throughout the present subjunctive, but, within this basic distinction, the
inflectional material used for inflecting second and third conjugation verbs is found to
multiply. Thus in upper Aller (Rodriguez-Castellano 1952) slightly different endings compete
for the inflection of both 1st pl. and 2nd pl. forms, e.g. dem-jamos, dem-jais ‘that we/you (pl.)
harvest’, tus-jamos, tus-jais ‘that we/you cough’, versus diš-ámos, diš-áes ‘that we/you say’.
Notice that this may also occur in the inflection of one and the same verb: ker-jámos, ker-jáis
beside kiš-ámos, kiš-áis ‘that we/you (pl.) want’. In one of the most striking cases (the
Asturian dialect of Cabranes, as described by Canellada 1944), some second and third
conjugation verbs display two competing sets of endings for the whole present subjunctive:
Table 2.
The existence of the so-called ‘Type-A’ and ‘Type-B’ inflectional patterns can be accounted
for by acknowledging, as suggested above, that present subjunctive affixes convey some
morphosyntactic information fused with some purely morphological information (actually,
‘Present Subjunctive + Class Identifier / Class Default’). The idea, at this point, is that in
Type-A varieties the morphosyntactic information takes over, though at the cost of sacrificing
the morphological information related to inflectional class. In Type-B varieties, by contrast,
the purely morphological information strengthens, as it manifests itself not only on the
paradigmatic dimension (the relevant distinction being between ‘class-identifier’ and ‘classdefault’ markers, as already observed) but also on the syntagmatic dimension, with ‘classdefault’ markers being used as indices of different co-occurring stems (Carstairs-McCarthy
and Cameron-Faulkner 2000). What differentiates the stems at issue, in fact, is the pattern of
paradigmatic distribution (the ‘morphome’, in Aronoff’s 1994 terms) with which they are
associated. Returning to the data of Table 2, for instance, it is worth noting that the stem of
the forms in I is a ‘special’ stem, i.e. a stem with a diphthong, normally marking the so-called
‘N-pattern’ (Maiden 2011), while the stem in II is the ‘default stem’, i.e. the stem which is not
associated with any morphomic pattern, and normally appears in the infinitive (cf. Malkiel
The idea that inflection-class membership should be counted as part of the information
possibly conveyed by inflectional affixes (Carstairs-McCarthy 1994) proves useful in
providing an explanation of the desinential allomorphy characterising the present subjunctive
in several Romance varieties. Starting from this assumption, the analysis of the IberoRomance data carried out above suggests two things: first, the information about inflectional
class is something that fusional affixes may, but need not, convey, as proven by the cases in
which the inflection-class distinction is neutralized. Second, when there is, actually, a formal
opposition between ‘class-identifier’ and ‘class-default’ affixes, it is possible for the latter to
be differentiated on the syntagmatic dimension, as indices of (distributionally) different cooccurring stems.
Table 1 – Present Subjunctive (Standard Spanish)
First Conjugation Second Conjugation Third Conjugation
CANTAR ‘sing’
TEMER ‘fear’
PARTIR ‘leave’
1sg. cánt-e
2sg. cánt-es
Table 2 – Present subjunctive, competing endings (Cabranes)
1sg. arrespuend-a
2sg. arrespuend-as
3sg. arrespuend-a
1pl. arrespuénd-amos arrespond-iámos
2pl. arrespuend-aes
arrespond-iáes / arrespond-iés
3pl. arrespuend-an
AGUIRRE, C. & W.U. DRESSLER. 2001. ‘On Spanish Verb Inflection’. Folia Linguistica XL/1-2, pp.75-91.
ALCOBA, S. 1999. ‘La flexión verbal’. In: Bosque I. and V. Damonte (eds.), Gramática Descriptiva de la Lengua
Española. Madrid: Espasa Calpe, pp. 4915-91.
ARONOFF, M. 1994. Morphology by Itself. Stems and Inflection Classes. Cambridge (MA): The MIT Press.
CANELLADA, M.J. 1944. El bable de Cabranes. Madrid: RFE, Anejo XXXI
CARSTAIRS-MCCARTHY, A. 1994. ‘Inflection Classes, Gender and the Principle of Contrast’. Language70:737-88.
CARSTAIRS-MCCARTHY, A. & T. CAMERON-FAULKNER. 2000. ‘Stem alternants as morphological signata:
evidence from blur avoidance in Polish nouns’. Natural Language and Linguistic Theory 18: 813-35.
KRÜGER, F. 1923. El dialecto de San Ciprián de Sanabria. Madrid: RFE, Anejo IV.
MAIDEN, M. 2011. ‘Morphophonological innovation’. In: Ledgeway, A. et al. (eds.), The Cambridge History of
the Romance Languages. Cambridge: Cambridge University Press, pp. 216-67.
MALKIEL, Y. 1969. ‘The five sources of epenthetic /j/ in Western Hispano-Romance’. Hispanic Review 37: 23975.
RODRIGUEZ-CASTELLANO, L. 1952. La variedad dialectal del alto Aller. Oviedo: La Cruz.
WURZEL, W.U. 1989. Inflectional morphology and naturalness. Dordrecht – Boston – London: Kluwer
Academic Publishers.
Economy and Multifunctional Affixes
Université du Québec à Montréal
[email protected]
I take multifunctional affixes to be part of functional structures legible at the interfaces with
the external systems, the semantic and the sensorimotor systems. I discuss the properties of
Italian and English Numerals, Adjectives, Demonstratives and Determiners including
multifunctional affixes. I argue that these functional elements are derived by the operations of
the language faculty and subject to principles reducing complexity. I bring theoretical,
empirical and diachronic evidence in support of my hypothesis.
1. Focus. A multifunctional affix has a basic phonetic form and more than one formal and
semantic properties. I discuss the properties of multifunctional affixes in the nominal
extended projection, (1), adapted from Cinque (2005). I focus on -esimo in Italian and -th in
English. I argue that these functional items are derived by morphological merger, as defined
in Asymmetry Theory (Di Sciullo 2005, 2014), and subject to economy principles, including
Minimal Search, (2a), and Pronounce the Minimum, (2b), Chomsky (2013).
(1) [Quniv . . . [Dem . . . [Numord . . . [RC . . . [Numcard . . . [Cl . . . [A . . .N]]]]]]]
(2) a. Minimal Search: Limit the search space.
b. Pronounce the Minimum: Limit the externalization.
2. Determiners, Ordinal numbers and Adjectives. 2.1 Ordinal numbers and Adjectives. I
argue that in Italian, the merger of the multifunctional affix -esimo, with a complex cardinal
yields an ordinal number, (3). The merger of -esimo with the identity affix med yields an
emphatic identity adjective in (4), (5), an emphatic reflexive pronoun in (6) and an anaphoric
pronoun in (7). The derivation of complex functional elements with the multifunctional affix
-esimo is restricted to the extended projection of nominal constituents.
(3) undicesimo, ….ventesimo , … millesimo, … millionesimo, …. (It)
(4) medesimo, medesima, medesimi (It)
(5) Mi ha detto le medesime cose. (It)
‘same sg.’ ‘same fem sing’, same pl’
‘He told me the same things.’
(6) Il presidente medesimo era presente. (It) (7) Il presidente era il medesimo. (It)
‘The president himself was present.’
‘The president was the same one.’
Thus, according to this analysis, cardinal numbers and emphatic reflexives are derived on the
basis of the merger of the multifunctional affix -esimo with complex cardinal numbers on the
one hand and the adjectival identity affix med on the other. Both the complex numeral and
the pronoun have a binary branching structure.
2.2 Ordinal numbers and Determiners. In English, an ordinal number is formed of a cardinal
number and the multifunctional affix -th, (8), apart for first, second and third. I extend the
granular approach to the multifunctional th- affix to the definite determiner and
demonstrative pronouns, (9). I argue further that the definite determiners and demonstratives
are the result of the merger of the th- multifunctional affix and a restricting vocalic
constituent, including a spatial proximate or distal component (Di Sciullo 2005).
(8) eleventh, … twentieth, ... one hundredth, … one thousandth, …
(9) a. th-e book on the table b. th-is book on the table c. th-at book on the table
3. Principles reducing derivational complexity. The derivation of functional elements is the
effect of the interaction of Merge and economy principles. While morphological merger
derives sets, economy principles ensure efficient derivations by limiting the search space. It is
predicted that functional affixes may not merge with categories that are not in their
immediate local domain. This can be seen by the difference in acceptability of (10a,b) versus
(10c,d). The locality restriction is represented in (11).
(10) a. #omni-th (En)
b. #omni-esimo (It)
c. one hundredth (En)
d. cento-unesimo (It)
[Quniv . . . [Dem . . . [Numord . . . [Numcard . . . [Cl . . . [A . . .N]]]]]]]
4. Multifunctional affixes at the interfaces. I argue further that principles reducing complexity
also ensure that the externalization of the functional structure is reduced to the minimum at
the sensorimotor interface, while preserving the fine-grained conceptual structure, including
the configurations in (12) at the semantic interface.
(12) a. (RelationRANK (SpaceNumords ))
b. (RelationIDENTIY(SpaceObjects))
c. ( RelationPROXIMITY (SpacePlaces ))
The externalization of linguistic expressions and the variation between languages is a
function of the properties of the language faculty and laws reducing complexity (Hauser,
Chomsky and Fitch 2002, Chomsky 2005, Di Sciullo 2011, 2012).
I provide diachronic evidence showing the effect of the principles reducing
complexity in the development of multifunctional affixes. For example, -esimo subsumes
multiple diachronic developments, including the following. Latin -esimo (identical form in
masc. and fem.) derives complex nouns, e.g. protestante, protestantesimo. Latin -esimus
(masc.) gives rise to -esimo (It) -ième (Fr) in ordinals, e.g. ventesimo (It), vingtième (Fr). In
Vulgar Latin metipsimus is formed of met (ego), ipse (self) and -issimus (superlative). The
Romance-Latin medisme gives rise to medesimo (It), meïsme and mesme in (OFr), to mismo
in (Sp) and to mesmo in (Po).
5. Consequences for the model of the language faculty. The minimalist Asymmetry Theory
approach to multifunctional affixes offers both a principled and a fine-grained account of the
very existence of these affixes, their derivation and their interface legibility.
CHOMSKY, N. 2013. Problems of Projection. Lingua 130: 33-49.
CHOMSKY, N. 2005. Three factors in language design. Linguistic Inquiry 36:1-22.
CINQUE, G. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36:315-332.
DI SCIULLO, A.M. 2014. Minimalism and I-Morphology. In P. Kosta, S. Franks and L. Schürcks (eds),
Minimalism and Beyond: Radicalizing the interfaces, 267-286. Amsterdam: John Benjamins.
DI SCIULLO, A.M. 2012 Perspectives on Morphological Complexity. In F, Kiefer, M. Ladanyi & P, Siptar (eds),
Morphology. (Ir)regularity, Frequency, Typology, 105-135. Amsterdam: John Benjamins.
DI SCIULLO, A.M. 2011. A Biolinguistic Approach to Variation. In A.M. Di Sciullo & C. Boeckx (eds), The
biolinguistic Entreprise: New Perspectives on the Evolution and Nature of the Human Language Faculty,
305-328. Oxford University Press.
DI SCIULLO, A. M. 2005. Asymmetry in Morphology. Cambridge, Mass: The MIT Press.
HAUSER, M., N. CHOMSKY, AND T. FITCH. 2002. The faculty of language: What is it, who has it and how did it
evolve? Science 298:1569‐1579.
Exploring the meaning and productivity of a polysemous prefix: The case of the Modern
Greek prepositional prefix paraANGELIKI EFTHYMIOU
Democritus University of
Thrace, Greece
University of Athens,
Democritus University of
Thrace, Greece
[email protected]
[email protected]
[email protected]
This paper aims at exploring the meaning and productivity of the Modern Greek prepositional
prefix para-, which creates adjectives (e.g. para-kratikós ‘extra-governmental’), nouns (e.g.
par-onímio ‘byname’), adjectival (e.g. para-zalizménos ‘bemused’) and adverbial participles
(para-pléondas ‘sailing by’), verbs (e.g. para-káno ‘to overdo’) and adverbs (e.g. para-ékso
‘further out’). Para- derives mainly from the ancient Greek preposition pará ‘close to (but
falling short of)’ (Bortone 2010: 291, cf. Triandafyllidis Dictionary 1998). Para- is also used
in neological loan translations (e.g. para-stratiotikos ‘paramilitary’). Moreover,
Triandafyllidis Dictionary (1998) has two different homonymous lemmas, one for para-1, and
another for para-2; the latter denotes the meaning of excess and derives from the adverb pára,
which in turn originates from the ancient Greek preposition pará.
Given that para- derives from Ancient Greek pará, its morphological status is often described
as ambiguous and the formations in which it participates can be considered as either
compounds or derivatives (for discussion, see e.g. Ralli 2004, 2005). On the other hand,
according to Ralli (2004), prepositional prefixes like para- should not be considered as
autonomous words, but rather as bound morphemes which are similar to derivational prefixes.
As regards verbal formations with para-, according to Ralli (2004) para- displays a dual
character, since it behaves like an internal prefix (in the sense of Di Sciullo 1997) when it
affects the meaning of the base and like an external prefix when it brings only external
specifications to it. More specifically, she claims that para- is internal when it expresses the
basic idea of proximity or parallelism to the meaning of the base (e.g. méno ‘to stay, to live’ para-méno ‘to stay on, to remain’) and external in its excessive meaning (tróo ‘to eat’ - paratróo ‘to over-eat’). Moreover, Ralli (2004) claims that the ‘excessive’ para-, which was not
present in Ancient Greek, is highly productive in Modern Greek.
Our aim in this paper is to study the meaning and productivity of the prefix para- in a large
corpus of Modern Greek and in particular: (a) to identify the different meanings of the prefix
on the basis of authentic data, (b) to examine which of these meanings are more or less
productive, (c) to investigate if the excessive meaning is productive in Modern Greek, (d) to
compare the productivity of the prefix across spoken and written registers, and especially
across the 14 text types of the corpus, (e) to measure the productivity of para- across the
grammatical categories of the items derived by it, and (f) to find correlations between the
grammatical categories of the derived items and the meanings of para-.
The data used for the measurement of the productivity of para- comes from the Corpus of
Greek Texts (CGT), a synchronic corpus of Modern Greek, including approximately 28
million words from a variety of spoken and written text types (see Goutsos 2010). We follow
our previous work (Efthymiou, Fragaki & Markos 2012), which has investigated the
frequency and morphological productivity of Greek verb-forming suffixes with a corpusbased methodology, similarly to earlier research on other languages (e.g. Baayen & Lieber
1991, Baayen 1992, Plag, Dalton-Puffer & Baayen 1999, Gaeta & Ricca 2003). Token
frequency, type frequency and hapax legomena were used to calculate the potential
productivity of the prefix, both in the corpus as a whole and within each subcorpus. In order
to account for the differences in size of subcorpora, we resorted to the family of LargeNumber-of-Rare-Events (LNRE) models, which are appropriate for modelling word
frequency distributions (see Baayen 2001). An LNRE model attempts to estimate the
expected number of types (the vocabulary size) both at smaller sample sizes (interpolation)
and at larger sample sizes (extrapolation), based on the counts of low frequency types in the
corpus (the frequency spectrum). Currently, three major models are available: Generalized
Inverse Gauss-Poisson (GIGP; Baayen 2001), finite Zipf-Mandelbrot and Zipf-Mandelbrot
(fZM and ZM; Baroni & Evert 2006). These models are implemented in the package zipfR
(Baroni & Evert 2006), a tool for lexical statistics in the R language, which is used in this
In our data para- expresses several meanings such as: (a) proximity (parathalásios ‘seaside’),
(b) parallel, subsidiary or accessory role (parádromos ‘side road’, paramána ‘nanny’), (c)
divergence, error or violation (e.g. paravlépo ‘to overlook’, parerminévo ‘to misread’,
parátipos ‘irregular’), (d) excess (e.g. paracimáme ‘to oversleep’, paraxaidévo ‘to
overindulge’, parefthís ‘straight away’), (e) resemblance (e.g. parafrázo ‘to paraphrase’), (f)
temporal continuity or duration (e.g. paraméno ‘to stay on, to remain’, paratherízo ‘to spend
the summer’), (g) transmittal (e.g. paradído ‘to deliver’). The most productive meanings are
divergence, error or violation and proximity, followed by excess and parallel, subsidiary or
accessory role. The least productive meanings are those of transmittal, resemblance and
temporal continuity or duration.
The prefix para- is more productive in written than spoken registers, following the pattern
found with other Greek prefixes (Efthymiou, Fragaki & Markos 2013). In terms of particular
text types, para- is more productive in opinion articles and popularized non-fiction texts,
followed by academic texts, news and literature. The prefix is least productive in public
speech and law and administration. It is also most productive in nouns, followed by verbs,
adjectives and adjectival participles. Our results show that both locational (e.g. proximity) and
non-locational meanings (e.g. divergence, error or violation) are productive in Modern Greek
prefixation with para-. It is interesting that the new meaning of excess is also very productive,
especially in verbs. There also seems to be a correlation between productivity in meaning and
productivity in grammatical categories, a finding which will be further explored.
EFTHYMIOU, A., FRAGAKI G. & A. MARKOS. 2012. Productivity of verb forming suffixes in Modern Greek: A
corpus-based study. Morphology 22 (4): 515-545. ● EFTHYMIOU, A., FRAGAKI G. & A. MARKOS. 2013. Aspects
of productivity of the Modern Greek prefix iper-: A corpus-based study. Paper presented at the 11th
International Conference of Greek Linguistics, 26-29 September 2013, University of the Aegean, Rhodes. ●
BAAYEN, H. 1992. “A quantitative approach to morphological productivity.” In Geert Booij, and Jaap van Marle
(eds), Yearbook of Morphology 1991, 109-149. Dordrecht: Kluwer. ● BAAYEN, H. & R. LIEBER. 1991.
Productivity and English derivation: A corpus-based study. Linguistics 29: 801-843. ● BAAYEN, R. H. 2001.
Word Frequency Distributions. Dordrecht: Kluwer. ● Baroni, M. & S. Evert. 2006. The zipfR package for
lexical statistics: A tutorial introduction. http://zipfr.r-forge.r-project.org/materials/zipfr-tutorial.pdf. ●
BORTONE, P. 2010. Greek Prepositions. From Antiquity to the Present. Oxford: Oxford University Press. ● DI
SCIULLO, A. M. 1997. Prefixed-verbs and adjunct identification. In A. M. Di Sciullo (ed.), Projections and
Interface Conditions. New York: Oxford University Press, 52-73. ● GAETA, LIVIO & D. RICCA. 2003. Frequency
and productivity in Italian derivation: A comparison between corpus-based and lexicographical data. Rivista di
Linguistica 15 (1): 63-98. ● Goutsos, D. 2010. The Corpus of Greek Texts: A reference corpus for Modern
Greek. Corpora 5 (1): 29-44. ● PLAG, I., DALTON-PUFFER, CH. & H. BAAYEN. 1999. Morphological productivity
across speech and writing. English Language and Linguistics 3 (2): 209-228. ● RALLI, A. 2004. Stem-based
versus word-based morphological configurations: The case of Modern Greek preverbs. Lingue e Linguaggio
2004/2: 241-275 ● RALLI, A. 2005. Morphology. Athens: Patakis. [In Greek]. ● TRIANDAFYLLIDIS DICTIONARY.
1998. Standard Modern Greek dictionary. Thessaloniki: Institute of Hellenic Studies, Aristotle University of
Thessaloniki. [In Greek].
On How Fusional and Agglutinative Patterns Emerge from a Learning Constraint
Universitat Autònoma de Barcelona
[email protected]
The aim of this study is to argue that a learning constraint independently motivated
for acquisition on how the learner builds morphological paradigms is responsible for
the appearance of fusional and agglutinative patterns. Given the representation and
acquisition of morphological knowledge in paradigms as in Pinker (1984), the
following constraint, active during the process of language acquisition, operating as a
simplifying complexity device (Chomsky 2005, Roberts 2007), is proposed:
Minimise Paradigms Constraint (MPC). The learner hypothesises just one general
paradigm when affixes in cells show a systematic syntagmatic relationship.
A more informal way of reformulating the MPC is stating that the learner prefers to
hypothesise the fewest possible paradigms when affixes show a concrete pattern
detected by the Language Acquisition Device, that of being in a systematic
syntagmatic relationship. Two morphs show that configuration when they always
appear adjacently, one cannot appear without the other and no other material can
appear in between. In Pinker’s original model, if the learner encounters a specific
paradigm (with the stem, X, included) like the one in (1), he builds, after extracting the
phonetic material in common, the following two agglutinative general paradigms (2),
(3). However, once we introduced the modification in Pinker’s model in order to
capture the effects of the MPC, when the learner encounters two morphs that show a
systematic syntagmatic relationship through a paradigm, he builds just one general
paradigm, that is, he takes all affixal information in cells and let it the same as a
general paradigm (4).
Due to the effects of the MPC, two relevant predictions about the degree of fusion
(Sapir 1921) in verbal morphology can be made. Prediction 1 states that when some
morphs show a systematic syntagmatic relationship in a paradigm and consequently
the learner builds just one general paradigm for all the affixes, these affixes will
(eventually) show a fusional pattern in subsequent intances of the language. The
logic behind prediction 1 is that, once affixes are put together in the same paradigm,
they will begin to show inconsistencies among forms, suffer morphophonological
erosion and finally fusionalisation, due to the systematic contiguity of. For that
reason, where on a first stage there are two different morphs instantiating two
morphemes, in subsequent stages of the language there will be just one fusionalised
morph. Prediction 2 states that when two morphs do not show such a systematic
relationship and consequently the learner builds as many general paradigms as
needed, the affixes will show an agglutinative pattern in subsequent intances of the
language. Prediction 2 captures the observation that potentially discontiguity
between morphs block morphophonological erosion and favours agglutination. The
logic behind this proposal is that the learner's analysis during language acquisition
can influence the I-language that at the end of the process he will end up acquiring. If
learners' analyses of a given generation are consistent, their (modified with respect to
the previous one) language will serve as input to the following generation of
learners, who will acquire an already modified languagel (Roberts 2007).
In order to illustrate the emergence of fusional patterns we will focus on Latin verbal
paradigms and their Romance counterparts in Catalan, Spanish, Italian and French.
We will argue that, because of some independent changes in the Latin passive voice
system, two morphs, the Tense-Aspectual-Mood (TAM) marker and the personal
desinence, became always adjacent in Latin and, as an effect of the MPC, verbal
paradigms underwent an important fusionalisation, observable in different degrees in
Romance. We will pay attention to morphological paradigms of different tenses in all
conjugations in order to show the increasing difficulty in distinguishing TAM
markers from personal desinences as two different units in Romance languages.
Regarding agglutinative patterns, we will pay attention to the structure of verbal
complexes in Bantu languages. We will argue that the agglutinative nature of Bantu
morphs is due to the non-adjacency of mandatory elements, as it can be observed in
the traditional schema of the full structure of the verbal pieces in Bantu languages, as
in Meeussen (1967): (preinitial) initial (postinitial) (preradical) radical (prefinal)
final (postfinal). Given that distribution, the MPC cannot be used by the learner in
these contexts and the agglutinative patterns are derived. We will concentrate on data
in Chichewa from Mchombo (2001) and other Bantu languages, and on some
Turkish data (Korn 1997).
The present approach can be considered a step forward in predicting how
morphological change happens inasmuch as it establishes specific morphological
contexts where the learner’s analises are going to change input representations. As
far as the author is aware, this is the first attempt in the literature to relate the use of
distributional properties, which are extensevely managed by language learners
(Redington et al. 1998), with the discovery of concrete semantic aspects of morphs,
namely, if they encode only one semantic distinction or more.
Selected References
Chomsky (2005) Three Factors in Language Design. Linguistic Inquiry; Korn (1997) Turkish.
Routledge; Mchombo (2001) Chichewa (Bantu). The Handbook of Morphology. Oxford
University Press; Meeussen (1967) Bantu grammatical reconstructions. Africana linguistica;
Pinker (1984) Language Learnability and Language Development. Harvard University Press;
Redington et al. (1998) Distributional information: A powerful cue for acquiring syntactic
categories. Cognitive Science; Roberts (2007) Diachronic Syntax. Oxford University Press;
Sapir (1921) Language. Harcourt, Brace and World.
Combining gender and classifiers: a typological perspective
University of Surrey
[email protected] [email protected]
Many languages systematically categorize their nominal vocabulary. This may involve a
gender system, as in Italian, where nouns are assigned to either the masculine or the feminine
gender. Another possibility is a system of classifiers, as in the Austronesian language
Kilivila, which distinguishes at least 177 categories (Senft 1996), based on semantic
properties, for example long and flexible objects, pots, or wooden objects. In the main a
language has only one system or the other. However, there are languages which combine two
systems. The interaction and functioning of the different systems shed light on the typology
of noun categorization. We here present new data on a language with two categorization
systems, and use it to refine the typology of such systems.
The data come from Mian, a Papuan language of Papua New Guinea (Fedden 2011).
Mian has a gender system with four values: masculine, feminine, neuter 1 and neuter 2. In
addition there are six classifiers: the M-classifier (for male referents and some inanimates),
the F-classifier (for female referents and many inanimates), the bundle classifier, the
classifier for objects which serve as covers, and a residue classifier. Agreement of the clitic
article (=e) in gender and number with the controller fút ‘tobacco (N1)’ and the use of the
classifier tob- ‘long object (SG)’ are both illustrated in example (1):
1SG tobacco=ART.SG.N1 3SG.LONG.OBJ-take-SS-1SG.SBJ=and
‘I take the long tobacco leaf and then I …’
The classifiers are prefixes on the verb. They make a singular-plural distinction and they are
restricted to occurring on verbs of object handling or movement, such as ‘give’, ‘take’,
‘throw’ and ‘fall’.
Languages like Mian require us to further develop the typology of categorization
systems. Earlier work often treated the properties of gender systems and classifier systems as
opposed to each other. Dixon (1982; 1986) gives helpful criteria for contrasting the two, and
he was followed by Aikhenvald (2000) and Grinevald (2000). Since then, further instances of
intermediary systems have been found, for instance, Miraña (Seifart 2005). Data from these
languages suggest that factors which are grouped together in the more familiar languages can
vary independently of each other, and that we need to look at these variables individually.
Given this situation, it is appropriate to take a Canonical Typological approach (Brown,
Chumakina & Corbett 2013). This approach involves examining the variability within and
across languages to establish the parameters of variation which are relevant to setting up a
theoretical space of possibilities. Only then do we examine how real examples are distributed
in the space. We find that there is a clear point of convergence in this space, and this points
us to defining a canonical gender system. We can do this largely on the basis of previous
work on canonical morphosyntactic features (Corbett 2013). Like other canonical
morphosyntactic features, a canonical gender system (and its gender values) is clearly
distinguished by formal means, its use is determined by simple syntactic rules and it is
realized by canonical inflectional morphology. Where gender differs from the other
morphosyntactic features is in the availability of values: in the canonical system, controllers
have one gender value, while targets have all values. From previous work on canonical
agreement (Corbett 2006), we take over the criterion that a canonical agreement system
involves lexical features, and gender is the key type of lexical feature.
Given this idealized picture provided by the canon, we can demonstrate its usefulness
by calibrating the facts of Mian against it. We find that the Mian gender system is close to
canonical in most respects. On the other hand, the classifier-like elements have some
properties close to a canonical gender system (they are realized as verbal prefixes reminiscent
of the agreement affixes found in gender systems) and other properties which are distant from
the canonical ideal (they are highly restricted in the lexical items involved).
We conclude that:
• languages like Mian are crucial to the understanding of the typology of noun
• categorization systems can be understood better by calibrating them from a
canonical ideal
• adopting Canonical Typology allows us to integrate into our typology the systems
close to canonical (Italian) as well as the intermediate cases we find both in single
systems and in combined systems (as in Mian)
2000. Classifiers: A typology of noun categorization devices. Oxford: Oxford University
Press.● BROWN, D., M. CHUMAKINA & G.G.CORBETT. 2013. Canonical morphology and syntax. Oxford: Oxford
University Press.● CORBETT, G.G. 2006. Agreement. Cambridge: Cambridge University Press. ● CORBETT, G.G.
2013. Canonical morphosyntactic features. In: Dunstan Brown, Marina Chumakina & Greville G. Corbett
(eds.), Canonical morphology and syntax, 48-65. Oxford: Oxford University Press. ● DIXON, R.M.W. 1982.
Where have all the adjectives gone? and other essays in semantics and syntax. Berlin: Mouton. ● DIXON, R.M.W
1986. Noun classes and noun classification in typological perspective. In Colette Craig (ed.), Noun classes and
categorization, 105-112. Amsterdam: Benjamins. ● FEDDEN, S. 2011. A grammar of Mian. Berlin: De Gruyter
Mouton. ● GRINEVALD, C. 2000. A morphosyntactic typology of classifiers. In Gunter Senft (ed.), Systems of
nominal classification, 50-92. Cambridge: Cambridge University Press. ● SEIFART, F. 2005. The structure and
use of shape-based noun classes in Miraña (North West Amazon). Radboud University PhD dissertation. ●
SENFT, G. 1996. Classificatory particles in Kilivila. Oxford: Oxford University Press.
Deverbal nominalizations and the ‘means’ interpretation
Laboratoire de Linguistique Formelle, CNRS & U Paris Diderot
[email protected]
1. Introduction Deverbal nominalizations (NZNs) that express a ‘means’ have repeatedly
been overlooked in the literature (e.g. Grimshaw 1990) or confounded with NZNs having an
Instrument interpretation (Vendler 1968). Recently, Melloni distinguished them from other
nominalizations and proposed to group them into a subclass dubbed ‘entities in state’, which
includes derived nouns denoting concrete entities associated with a specific state viz. result
e.g. translation and psych nominals e.g. preference (Melloni 2011: 115). This talk aims to
make explicit the conditions that determine when a given NZN may denote a means. It will
discuss Melloni’s proposal and show that means NZNs have to be set apart from Instrument
NZNs and are orthogonal to Result NZNs. The data will mainly be taken from English,
French and Italian.
2. Properties A distinctive feature of means nominalizations is their capacity to head an NP
occurring as the subject of their base-V, as heating does in (1). From this feature comes the
formula and paraphrase (2) regularly associated with them:
Our new central heating heats the whole basement.
λx. Bse-V’(e, x, y) ‘X such that X base-V Y’ e.g. heating = ‘X such that X heats Y’.
What makes Means NZNs special among nominalizations is the fact that they share formula
(2) with deverbal nouns that denote an Agent e.g. seller = ‘X who sells Y’, although their
exponent is primarily used to express the eventive meaning e.g. discuss-ion, beat-ing, spillage. It is so because they denote extensional entities (they are "R nominals"), which are either
concrete e.g. wrapping or abstract e.g. fra empêchement ‘hindrance’. As expected, NZNs
denoting an event never satisfy (2), hence the ungrammaticality of (3)
*Our thorough examination examines the data.
This situation follows from the fact that Means NZNs are derived from Vs heading a stative
construction, which involves either a spatial e.g. (4) or a causative relation e.g. fra équipement
in (5), whence (6) (Fradin 2012)(the representations given here are rough approximations).
surroundings’(x) ≡ surround’(e,x,y) ∧ LOC(s, x, CIRCUM(y)) ∧ FIG(x) ∧ GRND(y)
L’équipement des monocoques surpasse celui des catamarans.
‘the equipment of mono-hulls overpasses that of catamarans’
équipement’(x) ≡ equip’(e1,x,y) ∧ CAUSE(e1, e2) ∧ function_normally’(e2, y), where
e = event, s = state
The stativity of the constructions in question is shown by the fact that their base-V is odd with
the progressive and gets a non-habitual reading in the present tense (cf. (1))(Dowty 1979).
??Our new central heating is heating the whole basement.
3. Distinctions The following criteria allow us to distinguish Means (MNS) from Instruments
(INS): (i) INSs always involve a dynamic situation, (ii) INSs are not licenced as subjects of a
sentence denoting an actual eventuality that is not agent driven e.g. (8)(Schlesinger 1989),
(iii) INSs denotes countable objects which must exist as a separate entity, before and after the
event in which they are used has been completed, which is not necessarily the case of Means
cf. (9), (iv) INSs need not comply with (2).
?*This plane planed the floor.
These junction obstructions can even obstruct tiny blood vessels.
4. Independency The occurrence of the Means interpretation hinges on the possibility to
derive a NZN from a V exhibiting properties (2) and (4)/(5). But this NZN may itself result
from nominalizing a creation verb, the outcome of which is a noun denoting a means e.g. ita
prolungamento ‘continuation’ = ‘part of an X (road) which has been extended (but did not
exist before)’ as in (10).
Un prolungamento di 5 chilometri prolunga la stada.
‘a 5 km extension continues the road’
Hence the three way interpretation for prolungamento: eventive, result and means. On the
other hand, there are NZNs which only have a means interpretation e.g. fra renseignement
‘piece of information’← renseigner ‘give some information’. This NZN cannot denote an
event e.g. (11), let alone a result, since renseigner heads a verbal construction denoting an
activity or more precisely a so-called ‘degree achievement’ (Beavers 2009) and only telic
(creation / duplication / modification) verbs can be the base of Result nominalizations (Bisetto
& Melloni 2007).
*Le renseignement du touriste a eu lieu a midi.
‘The tourist has been informed at noon’
Many NZNs with a means interpretation pattern like renseignement e.g. fra consolation
‘consolation’, garantie ‘guarantee’.
5. References
BEAVERS John. 2009. Aspectual Classes and Scales of Change. Austin: The University of Texas at Austin.
BISETTO Antonietta & Chiara MELLONI. 2007. "Result Nominals: a Lexical-Semantic Investigation". In On-line
proceedings of the Fith Mediterranean Morphology Meeting (MMM5) Fréjus 15-18 September 2005,
Booij G., L. Ducceschi, B. Fradin, E. Guevara, S. Scalise & A. Ralli (eds). http://mmm.lingue.unibo.it/.
Bologna: University of Bologna.
DOWTY David R. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel.
FRADIN Bernard. 2012. "Les nominalisations et la lecture 'moyen'". Lexique 20:129-156.
GRIMSHAW Jane. 1990. Argument Structure. Cambridge (Mass.) / London: The MIT Press.
MELLONI Chiara. 2011. Event and Result Nominals. Bern: Peter Lang.
SCHLESINGER I. M. 1989. "Instruments as Agents: On the Nature of Semantic Relations". Journal of Linguistics
25 1:189-210.
VENDLER Zeno. 1968. Nominalizations and Adjectives. The Hague / Paris: Mouton.
Mass and count: a gradatum resulting from two combining features
Università degli Studi di Padova
[email protected]; [email protected]
Introduction. The mass property and the count property have been traditionally described as
inherent properties of the lexeme that trigger some grammatical constraints (i.a.: Renzi et al.,
2001). With respect to morphology, a mass noun is said not to have a plural form. As a
consequence, in the literature the phenomenon of mass nouns admitting plural is treated as
some kind of semantic shift (Renzi et al., 2001). This description does not really explain why
the plural is acceptable for a mass noun and is not predictive for what concerns the contexts
in which such a possibility takes place. Moreover, the traditional mass-count dichotomy is
not exhaustive about the range of possibilities that happen to surface, for example, in Italian.
With respect to nouns behavior, it seems that we face more than two categories: A.
Sangue(blood)-like nouns: they refer to substances. They can occur only at the singular form:
the plural is not grammatical; B. Anello(ring)-like nouns: they refer to individuated entities: at
the singular they refer to one, at the plural to more than one; C. Vino(wine)-like nouns: they
are usually singulars. When they surface as plurals, they can get only a “kind” interpretation;
D. Mela(apple)-like nouns: they refer to individuated entities: at the singular they refer to one,
at the plural to more than one. At the singular -under certain conditions- they can refer to the
substance of which the entities they refer to are made (e.g.: c’era molta mela nel frullato, lit:
‘there was a lot of apple in the smoothie’). At the plural -under certain conditions- they can
get a “kind” interpretation; E. Pizza-like nouns: they may realize either the possibilities of the
mela(apple) or of the vino(wine)-like noun type. Jackendoff (1991) and Chierchia (2010) have
explained this range of possibilities by assuming a complex system of features intrinsic on the
lexemes and operations performed on them.
Main proposal. Differently from the previous interpretations, we propose that there is no a
priori label attached to the lexeme to define a noun as mass or count. Thus, with respect to
Jackendoff (1991) and Chierchia (2010), there is no need to perform operations starting from
on a fixed status to change the position of a noun in a classification. We propose that a
nominal expression can be said to refer to a mass or count entity only when it has undergone a
computation through the levels that interface between the perceptual reality and the lexicon,
the morphology, the syntax. To give account of this, we will show: (i) the necessary and
sufficient features whose combinatory possibilities are needed to individuate the mass
interpretation or the count interpretation - as well as the other possibilities listed above - of a
nominal expression; (ii) that a mass or count interpretation is not a matter of a lexical
specification but the offspring of a computation at the levels of perceptual, morphologic and
syntactic interfaces as well; (iii) that a nominal expression can surface alternatively with more
than one feature configuration. In this presentation we will focus on the interface between the
referential world and lexical morphology.
[±u; ±k]. At a semantic level, an entity may have –or not – a boundary. If an entity is defined
within its boundaries, it is recognizable as an individual. This semantic property is linked with
the perceptual level: if an entity has a physical and natural boundary, the noun referring to it
will likely be associated with a semantic [UNIT] interpretation. However, it doesn’t follow
from this that the symmetric condition is as likely: the semantic property must not be
confused with a physical property of the entity to which the noun refers to. An entity that has
no physical boundaries per se can be associated with a semantic [UNIT] interpretation. For
example, the noun birra ‘beer’ may designate the substance: in this case, there is no possible
[UNIT] interpretation. Rather, it may designate a portion of the substance as well: this portion
comes in precise boundaries (e.g., a glass or a bottle) and thus the interpretation [UNIT] is
available. There is another possibility to get an individuality interpretation: it could not refer
to the entity defined by the boundary, instead by its kind (e.g., a particular kind of beer). ‘A
kind of beer’ shares all the properties of the substance ‘beer’, but at the same time it is
recognizable as an individual as it shows some properties of its own that allow to distinguish
it among the other kinds of beer. Here, the individuality property takes from the [KIND]
interpretation. The semantic level must interface on the one side to the perceptual, referential
level and on the other side with the morphological level. At the morphology-semantic
interface the individuality in terms of unit or kind (or, as a consequence, the nonindividuality) is codified at the level of this interface into binary features [± unit], from now
on [± u], and [± kind], from now on [± k]. Every entity linked (or not linked) to an
individuality property, [UNIT] or [KIND], at the semantic level puts into availability the
feature system, [±u; ±k], that make semantically interpretable the number morphology at the
semantic-morphologic interface.
Not possible
Singular only
Behaves like:
sangue/*sangui (blood/*bloods)
vino/vini (wine/wines)
anello/anelli (ring/rings)
A, C, E, D
C, E, D
B, D, E
From what we have stated and illustrated above, we can draw the following generalizations:
(i) in the computation, at least one of the two features must assume a negative value; (ii) only
one of the two features may assume a positive value. Roughly speaking, the two
generalizations state that if an individuality property is available, it is possible to activate a
counting/quantification processing focusing on the individual or on the kind; but it is never
possible to count on the unit and on the kind at the same time.
Conclusions. The different behavior of the nouns does not actually reflect a categorization
among five different groups A, B, C, D, E. Instead, it is the direct consequence of the fact
that a noun is linked to a particular combination of features. This links basically depends on
the speaker’s “encyclopedic” knowledge of the referent and on the communicative context in
which this latter is referred to. The consequence of this is the fact that a noun can be
associated to different combinations of features; each of them is equally well-formed if placed
in the appropriate communicative context. The fact that a noun has been traditionally
associated with a particular class depends on the frequency with which it is associated to a
specific configuration of features. In other words, a noun as ‘burro’ (butter) has been
traditionally classified as a mass noun since the contexts in which it surfaces with the
configuration [-u; -k] are more frequent than those in which it surfaces with the configuration
[-u; +k] or [+u; -k]. But, at least in theory, this does not mean that such a noun cannot be
linked to the configuration [-u; +k] or [+u; -k], i.e., cannot be a count noun at all.
Potentially, a lexeme can have more than one configuration of features at the morphologylexical interface and then undergo different syntactic computations. In this view, the
unmarked condition for a nominal expression is the one which allows the larger number of
combinations in the larger number of contexts. In the present study we considered a language
(Italian) that bears an overtly marked distinction singular vs. plural in number morphology.
Nevertheless, our analysis is based on a feature system that, at least in theory, should be
applicable cross-linguistically. We leave these surveys open to further examinations.
Selected references
CHIERCHIA, G. (2010). Mass nouns, vagueness and semantic variation. Synthese 174, 99-149. • CORBETT, G.
(2000). Number. Cambridge: Cambridge University Press. • STUMP, G. (2001). Inflectional morphology: A
theory of paradigm structure. Cambridge: Cambridge University Press. • JACKENDOFF, R. (1991). Parts and
Boundaries. Cognition 41, 9-45. • RENZI, L., SALVI, G. & CARDINALETTI, A. (2001). Grande grammatica
italiana di consultazione. Bologna. Il Mulino.
Word formation in a balanced diachronic corpus of Italian
Livio Gaeta
Claudio Iacobini
Università di Torino
Università di Salerno
[email protected]
[email protected]
A characteristic of recent morphological research is the ever growing importance of the use of
corpora. The broadening of the empirical basis for morphological analysis has allowed
scholars to provide descriptions and explanations of phenomena which would not have been
possible some years ago. Within a large research project funded by the Italian government
and aimed at investigating Italian word formation, a new research tool is being developed,
namely MIDIA (Morfologia dell’Italiano in DIAcronia). This diachronic corpus of Italian
fills an important lacuna: although a considerable amount of texts are available electronically
with special regard to the earlier stages of Italian, a balanced diachronic corpus was until now
a desideratum. MIDIA was explicitly designed to cover all periods of the linguistic history of
Italian, divided in five homogeneous time spans and collecting texts for seven different text
types, going from prose and poems to chronicles, scientific essays, etc. For each time span
and text genre 25 text chunks of about 8000 tokens have been taken and acquired into a
database. This amounts to a final corpus of about 7 million tokens, which is hopefully large
enough to perform a number of investigations on productivity of single word formation
patterns (cf. Cowie & Dalton-Puffer 2002, Gaeta & Ricca 2006, in press). Clearly, the
database is independently searchable for any of the parameters indicated, namely time span
and text genre.
The software utilized for building the database comes from the TreeTagger family as
already employed for the Repubblica corpus (cf. Baroni et al. 2004), enriched by more than
170,000 wordforms collecting all sort of allomorphs attested during the long history of Italian
(about 26,000 nominal, 11,500 adjectival, and 98,000 verbal forms plus about 15,600 forms
hosting clitics) and including most grammatical words (about 1,800) and proper names (about
21,000). All of this considerably enlarges the basic lexicon already present in the TreeTagger
version of the Repubblica corpus, and substantially improves its efficiency as a syntactic
parser. On the basis of a golden standard test, it was possible to verify a significant
improvement of the MIDIA-parser with regard to the Repubblica-parser increasing from
about 89% of correctness of PoS assignment to about 96% (cf. Gaeta et al. 2013). Moreover,
the user’s friendly searching mask allows any possible research on single as well as on
sequences of tokens or PoS-tagged lemmas (or parts of them), employing all common logical
symbols and extracting the data into any sort of tables and graphs.
On this basis, we want to exploit the empirical potential of MIDIA by carrying out two
different investigations sharing the same perspective, namely to explore two onomasiological
domains taking seriously into consideration the range of grammatical polysemy or
polyfunctionality, including homonymy, characterizing multifunctional affixes.
First, we will focus on the domain of spatial verbal prefixes. In the transition from
Latin to the Romance languages, and in the subsequent phase of development of the latter up
until the present day, there has been a reduction in the number of preverbal prefixes and the
types of meanings that can be expressed by these prefixes (cf. Iacobini 2004, Iacobini &
Masini 2006). Contemporary Italian is characterized by a low productivity of spatial verbal
prefixes and the reduction of their number compared to Latin. Verbal prefixation is
productively used almost exclusively to express privative, reversative, and iterative values.
We intend to examine the reasons that determined this situation, moving in two converging
directions: a) the identification of the verbal prefixes used productively in the different
temporal phases of the Italian language and the description of the types of meaning they
express; b) the identification over time of the possible types of prefixed verbs (starting from
the verbal bases available in each period).
Second, we will have a look at the domain of agentive suffixes (cf. Gaeta 2010, 2012).
In particular, we will try to understand whether data extracted from dictionary-based
investigations are well supported on the textual basis provided by MIDIA, in particular with
regard to the competing productivity of masculine and feminine suffixes (-tore and its
allomorphs vs. -trice) for designating agents and instruments (cf. Rainer 2011). For this goal
MIDIA provides data from different text types, going from poems to legal texts, which allows
us to keep under control the textual variation relating to genre. Moreover, as with verbal
prefixes, a comparison will be made with different and competing coding strategies for
expressing agent / instrument meanings coming from syntax. This will give us significant
cues for understanding the way how the process of lexical entrenchment favoring complex
derived words over syntactic units concretely takes place in the course of time.
Baroni, Marco et al. (2004), Introducing the la Repubblica corpus: A large, annotated, TEI(XML)-compliant
corpus of newspaper Italian. Proceedings of LREC 2004, 1771-1774. Lisbon: ELDA.
Cowie, Claire S. & Christiane Dalton-Puffer (2002), Diachronic word-formation and studying changes in
productivity over time: Theoretical and methodological considerations. In: Javier E. Díaz Vera (ed.), A
Changing World of Words. Studies in English Historical Lexicography, Lexicology and Semantics, 410437. Amsterdam: Rodopi.
Gaeta, Livio (2010), On the viability of cognitive morphology for explaining language change. In: Alexander
Onysko, Sascha Michel (eds.), Word Formation from Cognitive Perspectives, 75-95. Berlin / New York,
Mouton de Gruyter.
Gaeta, Livio (2012), Strumenti come agenti: a metà tra natura e storia. In: Valentina Bambini, Irene Ricci, Pier
Marco Bertinetto (eds.), Linguaggio e cervello - Semantica / Language and the brain – Semantics. Atti
del XLII Congresso Internazionale di Studi della Società di Linguistica Italiana (Pisa, SNS, 2008), Vol. 2,
Paper II.A.1, 1-14. Roma: Bulzoni.
Gaeta, Livio & Davide Ricca (2006), Productivity in Italian word-formation: A variable-corpus approach.
Linguistics 44.1: 57-89.
Gaeta, Livio & Davide Ricca (in press), Productivity. In: Peter O. Müller, Ingeborg Ohnheiser, Susan Olsen,
Franz Rainer (eds.), Handbook of Word-Formation. An International Handbook of the Languages of
Europe. Berlin/New York: Mouton de Gruyter.
Gaeta, Livio, Claudio Iacobini, Davide Ricca, Marco Angster, Aurelio De Rosa & Giovanna Schirato (2013),
MIDIA: a balanced diachronic corpus of Italian. Paper presented at the 21st International Conference on
Historical Linguistics, Oslo 6.-9.8.2013.
Iacobini, Claudio (2004), Prefissazione. In: Maria Grossmann, Franz Rainer (eds.) La formazione delle parole in
italiano, 97-163.Tübingen: Niemeyer.
Iacobini, Claudio & Francesca Masini (2006), The emergence of verb-particle constructions in Italian: locative
and actional meanings. Morphology 16.2: 155-188.
Rainer, Franz (2011), The agent-instrument-place “polysemy” of the suffix -TOR in Romance. Sprachtypologie
und Universalienforschung 64.1: 8-32.
On the person agreement suffixes of gerunds in Udmurt
University of Szeged
[email protected]
In this talk I discuss the agreement suffixes of two gerunds in Udmurt, a Uralic language
spoken in the Russian Federation by about 350 000 speakers. I will focus on the gerunds,
formed with the suffixes -ku and -toź, both expressing temporal relations. The data comes
mainly from my own fieldwork with 27 native speakers of Udmurt.
The two gerunds can be inflected with agreement suffixes indicating the
person/number features of the subject (1), or can have an overtly expressed subject (2).
Edygarova (2010) claims that overt subjects and agreement suffixes are in complementary
distribution (3).
‘When I was building my house, nobody helped me.’
[Koľa otyn uly-toź],
Kaťa bazar-yś
there be-NMLZ
new dress[ACC] buy-PST-3SG
‘While Kolya was there [in the library], Katya bought a new dress at the market.’
(Edygarova 2010: 87)
*Ton [mon škola-je
2SG 1SG school-ILL go-NMLZ-1SG
friend-ACC.2SG meet-PST.2SG
‘While I was going to school, you met your friend.’
The type of agreement affixes that cannot co-occur with an overt subject have often been
considered to be pronominal in nature. Pronominal affixes are widely discussed topic in the
literature on agreement from both typological and theoretical perspectives (Siewierska 1999,
2004, Corbett 2006, Jelinek 1984, Bresnan&Mchombo 1987, Baker 1996, Fuß 2005 among
others). Several criteria have been proposed for distinguishing pronominal affixes from
grammatical agreement. The most important one is the co-occurrence of the agreement
marking with an overt DP. In contradiction to Edygarova’s claim, the results of my fieldwork
with 27 native speakers of Udmurt show the person agreement of the two gerunds in question
is subject to variation – speakers of the Northern (and some of the Southern) dialects do
accept examples like (3). Thus, for those speakers, the agreement marking is not strictly
pronominal since it can be doubled by an overt DP. In my talk I will also discuss other tests
for distinguishing pronominal affixes from grammatical agreement, e.g. the co-occurrence of
the agreement marking with indefinite, quantified and questioned DP-s and the degree of
referentiality of the agreement marking. I will demonstrate that the person agreement suffixes
on the gerunds have mixed properties: despite their similarity to the free pronouns in terms of
referentiality they can co-occur with different types of overt DP-s (for some speakers), which
indicates that they should not be considered pronominal suffixes. I will argue that there is an
ongoing morphosyntactic change in Udmurt, namely the former pronominal suffixes are being
reanalyzed as an instance of pro-drop (or ambiguous marking in Siewierska’s terms). The
change might have been induced by the fact that Udmurt is generally pro-drop language (see
Fuß 2005 for a recent theoretical proposal on how the pro-drop properties of a language might
enforce the reanalysis of the former pronominal affixes/clitics into grammatical agreement).
To sum up, in this talk I deal with two Udmurt gerunds and more specifically, with
their agreement. It has been stated in the literature that this agreement is in complementary
distribution with overt subjects. My fieldwork with native speakers of Udmurt demonstrates
that this is not always true. I suggest that there is an ongoing morphosyntactic change in some
varieties of Udmurt, according to which the former pronominal suffixes are turning into
grammatical agreement. In this way the talk sheds light on the some unexplored issues of the
Udmurt morphosyntax and contributes to our general knowledge on the Udmurt language.
BAKER, MARC C. 1996. The Polysynthesis Parameter. Oxford Studies in Comparative Syntax. New York:
Oxford University Press.
BRESNAN, JOAN & SAM A. MCHOMBO. 1987. Topic, pronoun and agreement in Chichewa. Language 63.
CORBETT, GREVILLE G. 2006. Agreement. Cambridge: CUP.
EDYGAROVA, SVETLANA. 2010. Kategoria posessivnosti v udmurtskom yazyke. Dissertationes Philologiae
Uralicae Universitatis Tartuensis 7. Tartu: University of Tartu.
FUß, ERIC. 2005. The Rise of Agreement. A formal approach to the syntax and grammaticalization of verbal
inflection. Amsterdam/Philadelphia: John Benjamins.
JELINEK, ELOISE. 1984. Empty Categories, Case, and Configurationality. Natural Language and Linguistic
Theory 2:39-76.
SIEWIERSKA, ANNA. 1999. From anaphoric pronoun to grammatical agreement marker: why objects don’t make
it. Folia Linguistica 33/2:225–51.
SIEWIERSKA, ANNA. 2004. Person. Cambridge: CUP.
German Nominalization Patterns in a Diachronic Perspective:
Constructional Change and Functional Reorganization
University of Mainz
[email protected]
This paper investigates the diachronic development of two highly productive German wordformation patterns, namely nominalization in the suffix -ung (e.g. Landung ‘landing’,
Behandlung ‘treatment’) and Infinitival Nominalization (e.g. das Singen ‘singing’, das Warten ‘waiting’). Both word-formation patterns have previously been studied in a variety of different frameworks ranging from Generative Grammar to Cognitive Linguistics and Construction Grammar. However, to date, no systematic quantitative corpus analysis has been conducted to address the key questions concerning their diachronic development. The present
study aims at filling this gap with an extensive corpus-based investigation of both patterns.
The results lend support to a usage-based and constructionist view of word-formation and
word-formation change.
Previous research suggests that ung-nominalization is subject to an increase in semantic wordformation constraints, which is mirrored in a decrease in morphological productivity (cf.
Demske 2000). Infinitival Nominalization comes in as a “replacement process” (Barz 1998,
cf. also Werner 2010). At the same time, ung-nominalization is ubiquitous in Present Day
German. In addition, a variety of cultural factors leads us to assume that the overall number of
nominalizations has increased, for example a) the need to name concepts previously conveyed
only in Latin (cf. e.g. Eichler 1996), b) the drive for “brevity” and “language economy” that
can be observed as a tendency in 17th and 18th century German word formation (cf. Schmidt
2007: 153).
Adopting Baayen’s (2009) measures of productivity, we can thus predict (1) an increase in
realized productivity (i.e. type frequency) for both ung-nominalization and Infinitival Nominalization over the past centuries, while we expect (2) a decrease in potential productivity, i.e.
the relation of hapax legomena to the sum total of tokens belonging to the word-formation
patterns in question, for ung-nominalization. For Infinitival Nominalization, by contrast, we
predict (3) an increase in potential productivity, since Nominalized Infinitives (NIs) have to
“replace” terms previously coined by means of ung-nominalization.
Apart from these overall developments, Demske (2000) has detected noticeable changes concerning the constructions in which ung-nominals tend to occur. For example, in Present Day
German, they occur more frequently with a determiner or in the pluralized form than in Early
New High German. By contrast, the construction in which ung-nominals serve as complements of a preposition (e.g. in Lesung des Gedichts ‘in reading the poem’) comes out of use.
This construction can be seen as profiling the progression of an event rather than its boundaries, hence evoking a more ‘verby’ construal, while pluralization and the determiner construction highlight the ‘nouny’ properties of nominalizations by evoking a ‘count noun’ construal.
Since Infinitival Nominalization, by and large, seems to have retained its ‘verbal’ properties
(cf. e.g. Ehrich 2002), we expect a corpus analysis investigating the three aforementioned
constructions to show significant changes only for ung-nominalization, but not for Infinitival
A corpus analysis using the Mainz Early New High German Corpus – an as yet unpublished
corpus covering the time span from 1500 to 1710 – and the GerManC Corpus (Durrell et al.
2007), which covers the years from 1650 to 1800, confirms the hypotheses outlined above.
Further examination of the data drawing on different variants of collostructional analysis (cf.
Stefanowitsch 2013) leads to a more fine-grained characterization of the word-formation pat-
terns in question, which in turn has important implications for properly describing the constraints affecting these patterns as well as their diachronic change.
I will argue that the diachronic development of ung-nominalization and Infinitival Nominalization can be accounted for in a usage-based and constructionist framework. Word-formation
patterns are seen as constructions, i.e. form-meaning pairings emerging via generalizations
over actual language use (cf. e.g. Traugott & Trousdale 2013). Since ung-nominals are subject
to lexicalization at the semantic level, their increasingly frequent use in lexicalized reading
variants brings about changes in the generalizations language users make over this input.
Hence, the constructional schema itself (i.e. the word-formation pattern) is modified. Thus,
more and more ung-nominals are replaced by NIs. Infinitival Nominalization becomes the
default word-formation pattern for deriving action nouns, while ung-nominalization is now
subject to considerable constraints. Hence, the diachronic development can be seen as a case
of functional reorganization in the German word-formation system caused by a series of interdependent constructional changes in the domains of semantics (lexicalization), morphology
(changes in word-formation constraints), and syntax (change in the prevalent syntactic patterns in which the respective word-formation products are used).
BAAYEN, R. Harald (2009): Corpus Linguistics in Morphology. Morphological Productivity. In: Lüdeling, Anke;
Kytö, Merja (eds.): Corpus Linguistics. An International Handbook. Vol. 2. Berlin, New York: De Gruyter (HSK, 29.2), 899–919.
BARZ, Irmhild (1998): Zur Lexikalisierungspotenz nominalisierter Infinitive. In: Barz, Irmhild; Öhlschläger,
Günther (eds.): Zwischen Grammatik und Lexikon. Tübingen: Niemeyer, 57–68.
DEMSKE, Ulrike (2000): Zur Geschichte der ung-Nominalisierung im Deutschen. Ein Wandel morphologischer
Produktivität. In: Beiträge zur Geschichte der deutschen Sprache und Literatur 122, 365–411.
DURRELL, Martin; Ensslin, Astrid; Bennett, Paul (2007): The GerManC Project. In: Sprache und Datenverarbeitung 31, 71–80.
EHRICH, Veronika (2002): On the verbal nature of certain nominal entities. In: Kaufmann, Ingrid; Stiebels, Barbara (eds.): More than Words. A Festschrift for Dieter Wunderlich. Berlin: Akademie-Verlag (studia
grammatica, 53), 69–89.
EICHLER, Birgit (1996): Fachlich konnotierte Sprachstrukturen in frühneuzeitlicher Wissensliteratur. In: Kalverkämper, Hartwig; Baumann, Klaus-Dieter (eds.): Fachliche Textsorten. Komponenten - Relationen - Strategien. Tübingen: Narr (Forum für Fachsprachen-Forschung, 25), 271–291.
SCHMIDT, Wilhelm (2007): Geschichte der deutschen Sprache. Ein Lehrbuch für das germanistische Studium.
10th ed. Stuttgart: Hirzel.
STEFANOWITSCH, Anatol (2013): Collostructional Analysis. In: Hoffmann, Thomas; Trousdale, Graeme (eds.):
The Oxford Handbook of Construction Grammar. Oxford: Oxford University Press, 290-302.
TRAUGOTT, Elizabeth Closs; Trousdale, Graeme (2013): Constructionalization and Constructional Changes.
Oxford: Oxford University Press (Oxford Studies in Diachronic and Historical Linguistics, 6).
WERNER, Martina (2010): Substantivierter Infinitiv statt Derivation. Ein ‚echter‘ Genuswechsel und ein Wechsel
der Kodierungstechnik innerhalb der deutschen Verbalabstraktbildung. In: Bittner, Dagmar; Gaeta, Livio
(eds.): Kodierungstechniken im Wandel. Das Zusammenspiel von Analytik und Synthese im Gegenwartsdeutschen. Berlin, New York: De Gruyter (Linguistik - Impulse und Tendenzen, 34), 159–178.
Overabundance in Zurich German
University of Zurich
[email protected]
THORNTON (2011) deals with overabundance, a kind of non-canonical inflectional
morphology. She distinguishes three types of overabundance: “forms built according to
different means of exponence in the same cell, […] forms built on two different stems in the
same cell, […] forms with different inflectional endings in the same cell.” Each type is a kind
of deviation from canonical inflection and by itself a class of canonical overabundance
(THORNTON 2011:361). Zurich German shows an example for the two latter deviations. The
indefinite article shows cell-mates in the dative in all three genders, cf. Table 1.
nom. / acc.
emene, eme
emene, eme
enere, ere
Table 1. The Zurich German indefinite article, cf. WEBER (1923:168)
Concerning the feminine paradigm, the emergence of the two forms can diachronically be
traced back to earlier stages of German. Synchronically, they do not show the same stem or
the same suffix either. In the masculine paradigm the overabundance is due to a suffix {ne}
which is not obligatory. This suffix is not yet attested in the inflection of the Early New High
German (ENHG) indefinite article. It is homophonous with the suffix in the dative plural of
the pronominal inflection and it builds a new form based on the already inflected form eme.
Generally, it seems possible that an inflected form is reanalyzed as a stem and another suffix
is added to the stem, as JOSEPH (2013) shows for Ancient Greek pronouns. I suggest a similar
process in the emergence of emene: the form eme has been expanded to emene. There is in
fact some dialectological evidence for this derivation. Still, the origin of this suffix {ne} in
emene remains a matter of dispute. The traditional view, however, that emene is a
phonological variant of eneme showing metathesis, has recently been regarded as not
convincing, cf. SEILER (2003).
Overabundance in the dative cells of the indefinite article is already evident in earlier stages
of German, cf. Table 2. In ENHG, there are two different types of dative forms of the
indefinite article in the feminine and masculine / neuter.
masc. /neutr. dat.
fem. dat.
einem, eineme, eime
einer, einere, eire
Table 2. The Early New High German indefinite article (dative), cf.
According to this, overabundance in this cell is diachronically well attested. It is, however,
very difficult to state how overabundance is organized in older periods of a language, cf.
THORNTON (2011:363). The mere fact that it appears in a grammar of ENHG does not mean
that the overabundant forms are not subject to diatopic, diaphasic, diachronic or other factors.
In modern languages further conditions can be found. Thus, overabundance can be
determined much more precise, and even intrapersonal variation is attested. In fact, the Zurich
German indefinite article does not only represent an instance of overabundance within a
paradigm, but is also subject to intrapersonal variation. The conditions for the variants to
occur are still unknown, though. Four kinds of factors have to be taken into account:
syntactical (e.g. syntactical position and function), phonological (e.g. rate of speaking, metre),
semantic-functional (e.g. specificity) and socio-linguistic conditions (e.g. age). In my previous
research I could not identify a determining factor for the usage of the forms. This makes
Zurich German overabundance a rather canonical type with regards to THORNTON’s (2011)
As far as we know, only the dative case is affected by overabundance. This, too, makes it a
highly canonical instance of overabundance since overabundance occurs in one cell only. The
question why it is restricted to the dative remains unanswered.
In my presentation, I will discuss the Zurich German data with respect to the factors
presumably affecting the occurrence of the variants. I will show that the occurrence of the
variants is unpredictable and therefore I can present a canonical case of overabundance.
Finally, I will discuss some hypotheses on how the dative overabundance could have
Selected references
JOSEPH, Brian (2013): Deixis and person in the development of Greek personal pronominal paradigms. In:
Kragh, Kirsten Jeppesen and Jan Lindschouw (eds.): Deixis and pronouns in Romance Languages. Amsterdam:
John Benjamins (= Studies in Language Companion Series 136), 19-32. • REICHMANN, Oskar / WEGERA, KlausPeter (eds.) (1993): Frühneuhochdeutsche Grammatik. Tübingen: Max Niemeyer. • SEILER, Guido (2003):
Präpositionale Dativmarkierung im Oberdeutschen. Stuttgart: Steiner (= Zeitschrift für Dialektologie und
Linguistik, Beihefte 124). • THORNTON, Anna M.: (2011): Overabundance (Multiple Forms Realizing the Same
Cell): A Non-canonical Phenomenon in Italian Verb Morphology. In: Maiden, Martin et. al. (eds.):
Morphological Autonomy. Perspectives from Romance Inflectional Morphology. New York: Oxford University
Press, 358-381. • WEBER, Albert (1923): Die Mundart des Zürcher Oberlandes. Frauenfeld: Huber & Co.
Ambivalent affixes: the -idz(o) verbalizer in Griko
University of Patras
[email protected], [email protected]
Morphemes are usually divided into groups according to their meaning/function. For
example, affixes which derive verbs out of nominal bases are usually called verbalizers and
are grouped together. However, there are affixes which have more than one function and thus
can be defined as multifunctional. Multifunctional affixes are formatives which occur in a
multitude of contexts conveying a number of different features (Himmelmann, 2006). These
affixes have been considered as an interesting topic in morphological theory since they pose a
number of problems such as the asymmetric relation between form and meaning/function
(Mel’cuk, 1973).
In this paper, we focus on the formative -idz(o) (a form variant of –iz(o) in Standard
Modern Greek) in Griko, a Greek based dialect of South Italy. This affix is primarily used as
a verbalizer, but it has developed some new functions in a number of contexts. The aim is to
explore the different functions of this affix, the motivation behind its multifunctionality and
the grammatical process underlying the correlations between the different functions. The
formative -idz(o) attaches to nominal bases in order to form verbs of a specific inflectional
class (see Ralli, 2013 for a description of the ICs in Modern Greek):
(1) [[[θer]-idz]-o] ‘to reap’
< [[θer]-os] ‘reaping’ (Griko)
However, in a significant number of cases this formative shows up in bases which are already
specified as to the verbal category and the inflectional class (IC) and forms verbal doublets:
Table (1)
A. Form with stem extension
[[akkoumb]-o] ‘to lean’
[[krat]-o] ‘hold’
[[lyp]-o] ‘to mourn, feel sad’
B. Form without stem extension
[[[akkoumb]-idz]-o] ‘to lean’
[[[krat]-idz]-o] ‘hold’
[[[lyp]-idz]-o] ‘to mourn, to feel
In these data we notice that -idz(o) does not keep its principal derivational character since it
attaches to verbal stems which are already specified as to both the verbal category and the IC.
In this paper we examine these cases and try to answer the following questions:
(a) How does this affix arise in the first place?
It has been claimed that similar elements may appear in order to show a semantic opposition
or an aspectual difference between the two members of the verbal pair. A classic example is
the inflectional element -isc- which appears in the 4th verbal conjugation in Italian and is a
continuation of the Latin derivational suffix -sc-. As Maiden (2003) points out, the broad
consensus about Latin -sc- is that it carried an ‘ingressive’ value, i.e. expressing the meaning
of ‘becoming/entering a state’, but nowadays it does not have any function or particular
semantic nuance, either syntagmatically or paradigmatically (see the discussion in Maiden
2003). Data in table 1 show that there is no semantic opposition or aspectual difference
between the forms with and without the formative -idz(o) (cf. Karanastasis, 1997). Thus, in
order to find the motivation behind this phenomenon we need to go beyond the classic
explanations and to examine the inflectional system of the dialect.
(b) What is the function of this affix in these particular formations?
In Griko, the inflectional classes of the verbs differ with respect to their productivity
(Karanastasis, 1997). This difference triggers cross-paradigmatic levelling which results in the
reshape of IC2 in favour of IC1 (most productive paradigm in Griko). Verbs of IC2 show a
phonological similarity in the Aorist with verbs which have the formative -idz(o) but belong
to IC1. We assume that this phonological similarity leads to paradigmatic interference and
through the process of proportional analogy -idz(o) has developed new properties, that is, it
can be added to formations as an explicit formal marker of the inflectional class. In this view,
we claim that it has acquired a rather stem-forming (morphomic, in the sense of Aronoff,
1994) status.
(c) One or two different affixes?
If looked from a broader perspective, the formative -idz(o) serves a dual function in the
system as both a verbalizer and a stem-forming suffix and thus is ambivalent between an
inflectional and a derivational status. Similar cases can be found in IE data where stemformatives may have the function of either aspect or mode of action which are separate
grammatical and lexical categories, respectively (Kastovsky, 2005: 39). The critical question
with respect to these data is whether one should assume that -idz(o) is one single affix or two
different affixes. We would like to propose that along with the old pattern, -idz(o) has
developed a new pattern which is used in different contexts.
This case study is important for two different reasons: first, it sheds light on the different
functions a multifunctional affix may have, and more importantly, it shows that an affix
which is purely derivational may develop into a formally explicit marker of the inflectional
class. Second, it shows that multifunctional affixes can emerge from reasons other than
semantic opposition or aspectual difference. We propose that paradigmatic levelling may be
seen as the driving force behind the rise of a multifunctional affix. Last but not least, it opens
the way to discuss the relation between inflection and derivation in the diachronic
development of affixes.
ARONOFF, M. (1994). Morphology by itself: Stems and Inflectional Classes (Linguistic Inquiry Monographs 22).
Cambridge, Massachusetts: MIT Press.
HIMMELMANN, N. (2006). How to miss a paradigm or two: multifunctional ma- in Tagalog. In: F. Ameka, A.
Dench & N. Evans (Eds.), Catching Language (487-526). Berlin: Mouton de Gruyter.
KARANASTASIS, A. (1997). Grammar of the Greek dialects of Southern Italy (in Greek). Athens: Academy of
KASTOVKSY, D. (2005). Conversion and/or zero: word-formation theory, historical linguistics, and typology. In:
Bauer, L. and S. Valera (Eds.), Approaches to conversion/zero derivation (31-50). Münster: Waxmann.
MAIDEN, M. (2003). Verb augments and meaninglessness in Early Romance morphology. Studi di Grammatica
Italiana 22: 1-61.
MEL’CUK, I. A. (1973). The structure of linguistic signs and possible formal-semantic relations between them. In:
R-D. Josette (ed.), Recherches sur les systèmes signifiants. Symposium de Varsovie 1968 (103-135). The
Hague, Paris: Mouton.
RALLI, A. (2013). Compounding in Modern Greek. Dordrecht: Springer.
On a realistic LFG treatment of the periphrastic irrealis mood in Hungarian
Department of English Linguistics, University of Debrecen
[email protected]
1. Introduction. Consider the following table showing the 2SG indefinite and definite parts of
the conditional and irrealis mood paradigms in Hungarian.
‘would see’ + Oid
‘would see’ + Od
‘would have seen’ + Oid
‘would have seen’ + Od
see-PAST-2SG be-COND
see-PAST-2SG be-COND
As these examples demonstrate, conditional verb forms are synthetic and irrealis verb forms
are systematically analytic (= periphrastic). The latter use the following two-word pattern: the
first word is the conjugated past tense form of the lexical verb and the second verb is the
combination of one of the stems of the copula van ‘be’ (vol-) and the conditional marker (-na)
invariably. In other words, Hungarian encodes irrealis mood periphrastically via the
combination of two morphosyntactic features: PAST and CONDITIONAL.
Bartos (2000) shows that volna is an independent syntactic atom, see his examples in (1-4).
(1) %vár-t
wait-PAST.3SG too VOLNA
‘he would also have waited’
(2) %vár-t-ál
csak volna
wait-PAST-2SG only
(3) %vár-t-ál-e
‘you would only have waited’
(4) én megsüt-ött-em,
fry-PAST-1SG.DEF you by.contrast eat-PAST-2SG.DEF
‘I would have fried and you, in turn, would have eaten (it)’
‘would you have waited?’
For a considerable number of speakers (but not for all speakers, hence the % symbol), the two
verbal elements can be separated by unquestionably independent words (is ‘too’ and csak
‘only’, as in (1) and (2), respectively), and by the yes/no question marker (-e), as in (3),
which, under normal circumstances, attaches to finite, fully conjugated verb forms (e.g. vár-tál-e tegnap? wait-PAST-2SG.INDEF-QM yesterday ‘did you wait yesterday?). For the other
speakers, these three elements have to follow volna immediately. Moreover, these forms can
produce right-node-raising effects, as in (4). This construction is acceptable for all speakers.
The challenge then is the development of an appropriate and explicit treatment of this fully
and invariably periphrastic irrealis mood paradigm. My framework is Lexical-Functional
Grammar (LFG). In the paper, first I will outline an account using the technical apparatus of
classical LFG and then I will show how the inferential-realizational approach to paradigms
(also containing periphrastic forms) advocated by Ackerman & Stump (2004) and Ackerman
et al. (2011) can be formally accommodated in this model.
2. Two possible LFG analyses. It is a crucial and shared syntactic aspect of both accounts to
be presented below, that, inspired by Laczkó & Rákosi’s (2011) treatment of Hungarian noncompositional particle verb constructions, PVCs, I assume a non-projecting syntactic
category: PRT (particle) and claim that, in addition to preverbs (particles) in Hungarian PVCs,
volna is another PRT in its use in the paradigm under investigation.
2.1. A classical LFG solution. This account is morpheme-based. It employs two distinct
lexical entries for volna (PRT) and the finite, past-tense-marked verb form conjugated for
subject agreement and definiteness. Consider the following lexical and syntactic
representations in (5)-(7). (5) is the familiar lexical entry for this particular past tense verb
form except for the MOOD annotational disjunction: in the regular past tense use, this form
contributes the indicative value for the mood feature, or, alternatively, it constrains the mood
to be irrealis. The non-projecting word, volna, contributes the specification for irrealis mood,
at the same time constraining the tense form of the inflected lexical verb to encode past tense,
see (6). PRT, being non-projecting, is head-adjoined to the verbal head (and the two elements
are functional co-heads, each making its own contribution to the f-structure of the sentence),
see (7).
(5) láttad, V ‘see <(↑SUBJ) (↑OBJ)>’
(↑OBJ DEF)= +
(6) volna, PRT
2.2. An inferential-realizational paradigmatic LFG treatment. The basic idea here is that a
finite (lexical) verb form like (5) has two, more radically different lexical entries (contra the
previous approach), because it is involved in two distinct paradigms: in the regular past tense
paradigm and in the irrealis mood paradigm.
(8) láttad, V ‘see <(↑SUBJ) (↑OBJ)>’
(↑OBJ DEF)= +
(9) volna, PRT
(↑OBJ DEF)= +
(8a) is a single, synthetic form encoding a particular (finite, past tense) paradigmatic slot. (8b)
is the crucial lexical form from our current perspective. It is one of the two elements of a
periphrastic (analytic) mode of expressing a particular irrealis mood paradigmatic slot. Notice
that ALL the relevant features characterizing this slot are encoded in this lexical entry, see the
first four equations AND it also constrains that this form has to co-occur with the volna PRT.
At the same time, the lexical entry for volna in (9), the second analytic element of this
periphrastic expression, has been “impoverished”: it no longer contributes any
morphosyntactic features or constraints; instead, it only has a FORM feature with the VOLNA
value (just like idiom chunks in the classical LFG treatment). The functional co-head
annotations of V0 and PRT remain the same as in (7).
This approach has the following advantages. (1) It spells out the general and
programmatic paradigmatic (inferential-realizational) approach to periphrasis advocated by
Ackerman & Stump (2004) and Ackerman et al. (2011) in an LFG framework. (2) It leaves
the classical view of lexical encoding in LFG intact: by using an appropriate checking and
cross-referencing mechanism in the relevant lexical forms, it can avoid recourse to multiple
word lexical entries, which would pose rather severe problems for LFG’s general
morphological assumptions as well as for implementation). (3) The devices it employs can be
argued to be motivated and justified independently, again, see Laczkó & Rákosi (2011) for
the treatment of derivational processes in the case of non-compositional PVCs.
ACKERMAN, FARRELL; STUMP, GREGORY T. & WEBELHUTH, GERT. 2011. Lexicalism, periphrasis, and
implicative morphology. In Borsley, Robert D. & Börjars, Kersti. eds. Non-Transformational Syntax: Formal
and Explicit Models of Grammar. Oxford: Blackwell. 325-358. ● ACKERMAN, FARRELL & GREGORY STUMP.
2004. Paradigms and periphrastic expression. In Sadler, Louisa & Andrew Spencer. Projecting Morphology.
Stanford Studies in Morphology and the Lexicon. Stanford: CSLI Publications. 111-157. ● BARTOS, HUBA. 2000.
Az inflexiós jelenségek szintaktikai háttere. In Kiefer, Ferenc. ed. Strukturális magyar nyelvtan 3. Morfológia.
Budapest: Akadémiai Kiadó. 653-762. ● LACZKÓ, TIBOR & RÁKOSI, GYÖRGY. 2011. On particularly predicative
particles in Hungarian. In Butt, Miriam & King, Tracy H. eds. Proceedings of the LFG '11 Conference. On-line
publication. Stanford: CSLI Publications, ISSN 1098-6782. 299-319.
Bar-Ilan University
[email protected]
This study examines the formation of plural forms of borrowed nouns in Palestinian. Arabic.
Arabic has two types of plurals, suffix-based sound plural (e.g. mata:r – mata:ra:t ‘airport
sg.–pl.’) and template-based broken plural (e.g. maktu:b – maka:ti:b ‘letter sg.–pl.’). These
two types of plurals are found also in borrowed nouns, as exemplified in (1).
(1) Plural forms
Sound plural Broken plural
‘folder (in a computer)’
The nouns in (1a-b) take the sound plural and those in (1c-d) take the broken plural.
However, the noun in (1e) takes either plural. The question addressed in this study is: How
does the morphological mechanism select between the two formation strategies? I show that
the selection of plural forms can be partially predicted, based on morpho-phonological
constraints, as well as a semantic criterion.
The study is based on a collection of 153 examples of foreign plural nouns provided by native
speakers of Palestinian and Jordanian Arabic. While in most cases the -a:t suffix is selected as
the default plural marker, there is a noticeable number of cases where plural templates are
selected. From the morpho-phonological point of view, I argue that properties of the loan
word like stress, vowels and the number of syllables, as well as resemblance to existing
Arabic nouns, determine which word formation strategy is selected and within the templatic
plural forms, which ones are favored over others. On the semantic dimension, I show that
when the foreign noun denotes human being, broken plural are almost exclusively selected.
Plural formation in Arabic. There are two main plural formation strategies in Arabic:
Suffixed-based ‘sound plural’ (SP) and template-based ‘broken plural’ (BP). There are two
SP suffixes in Arabic. The suffix -i:n is attached to masculine nouns that denote humans only.
It is also restricted with regard to the form of the singular base. This suffix is attached to
participle forms like maCCu:C as well as other templates like CaCCa:C. Feminine formation
in Arabic is performed by attaching the suffix -e to the masculine form, e.g. mallemmalleme ‘teacher’. The plural formation of such nouns is performed by attaching the SP
suffix -a:t (mallema:t ‘teachers fem.’). The broken plural involves internal modification of
the singular base, e.g. maktab-maka:tib ‘office’. BP formation manifest diversity, but at the
same time there is a great extent of regularity that allows partial prediction about the
selection of one template and not another (Levy 1971, McCarthy 1983, Hammoud 1988).
McCarthy & Prince (1990) show that there are structural properties that are drawn from the
singular base, which cannot be attributed to the root or the template. This is attested when
properties like vowel length are transferred from the singular stem to the plural and when
derivational morphemes survive derivational processes. This study also supports their claim,
showing that most cases where BP templates are selected for foreign words can be explained
based on properties on the loan singular base.
Plural formation of loan words in Arabic. Out of 153 foreign plural nouns such forms, 108
(71%) take the -a:t suffix, 34 (22%) are formed is a template and 11 (7%) have two possible
plural forms. While Arabic loan words prefer the sound plural -a:t as their plural marker
(71%), the cases where BP is used does not seem accidental and can be accounted for by
faithfulness constraints and a semantic constraint.
Faithfulness constraints. Faithfulness constraints require identity between a base and its
derived form. I show that they play an important role in the selection of a plural form.
Selecting the SP suffix -a:t is the default strategy, as it allows keeping the structure of the
base as is without any internal modifications. Syllable structure, number of syllables and the
vowels and consonants of the base all remain intact (kombyuter- kombyutera:t ‘computer’).
The selection of BP templates occurs almost exclusively in cases where the foreign base is
mono-syllabic (film-afla:m ‘film’) or bi-syllabic (banšer-bana:šer ‘puncture’). In such cases
the loan word has a syllabic structure of Arabic words that take themselves BP template. The
word bank, for example (CVCC) has the same structure as the Arabic noun bayt ‘house’. The
latter has the BP form byu:t (CCV:C) and the plural of bank takes the same template (bnu:k).
The existence of a CVCC- CCV:C paradigm in plural formation allows the implementing the
same strategy on loan words. When the loan nouns exceed the minimal word size, none of the
BP are selected. This is because the morphological mechanism cannot adjust such words into
any of the Arabic templates without deletion of syllable and some of the consonants. A
quarto-syllabic word like karburater ‘carburetor’ cannot fit any of the existing BP templates.
Stress preservation also plays a role in the selection of BP forms. The suffix -a:t always
takes stress and therefore changes the location of stress. Selecting some of the BP forms
allows keeping the same type of stress. The selection of one of the BP forms CaCa:CiC and
CaCa:Ci:C is not random. CaCa:Ci:C is selected when the base has ultimate stress (karto:nkara:ti:n ‘cartoon’). The type of stress in the plural form remains in tact, as well as original
location of stress in the loan. When stress is penultimate, CaCa:CiC is selected (tankertana:kir ‘tanker’). The location of stress does not remain intact, as it is on the first syllable in
the loan. However, the type of stress remains the same. Adding the SP suffix would alter the
stress pattern completely as it fall on the suffix (e.g. folder -*foldera:t, instead of fala:dir).
Such cases demonstrate the central role of stress in word formation. Generally, the
findings provide further support to the claim that properties of the base are taken into account
in word formation. The morphological mechanism tries to keep the derived form as faithful as
possible, and in the case of plural formation of loan words, the competition is between two
faithfulness constraints: keeping the base structure in tact or keeping the same stress pattern,
while taking into considerations templatic restrictions on word formation in Arabic. This is
supported by previous studies, where faithfulness constraints have been shown to play a role
in adapting foreign words (Bolozky 1978, 1999, Bat-El 1994, 2002, Ussishkin 1999, 2005).
The [+/- human] distinction. The SP -i:n suffix is rather unproductive within Arabic existing
words. Its use is restricted to nouns and adjective of certain templates like mCaCCeC. Since
the -a:t suffix is never used for plural of [+human] masculine nouns, it cannot be used for the
foreign nouns in (2). The only option that is left for plural formation is the use of the BP
forms that are also used in Arabic words and that can also denote human being. This provides
further evidence that the +/- human (or animate) distinction is relevant for the grammar, as
have been shown in various studies.
(2) Plural forms of [+human] nouns
Variation in plural formation. Plural formation of Arabic loan nouns is subject to variation,
where the same noun takes both a SP and a broken plural form. The noun ballo:n ‘balloon’,
for example, takes both the -a:t SP suffix (ballo:na:t) and the BP CaCa:CiC template
(bala:li:in). This variation is found between dialects of Arabic and in some cases within the
same dialect. This happens mainly in nouns with ultimate stress. The existence of few plural
forms result from the two competing faithfulness constraints discussed above.
Examining the morphological behavior of loan nouns provides direct access to the process of
word formation and shows how different types of constraints are taken into consideration.
Specifically, this study provides further support for a word-based approach, as it demonstrates
another case where elements from the base, in addition to the consonants, are transferred to
the derived plural form.
What are you afraid of?
A semantic analysis of the French neoclassical element -(o)phobie
CLLE-ERSS & University of Toulouse
[email protected]
Neoclassical compounds contain elements which were originally lexemes in Ancient Greek or
Latin but are bound forms in modern languages (for a description, see among others Amiot
and Dal (2007)). It is tempting to classify them as an homogenous class, but this hypothesis
turns out to be difficult at a purely synchronic analysis. Rather, we consider that neoclassical
elements which can form long set of lexemes (for example in French -(o)logue, -(i)cide,
-(o)crate, ...) are closer to affixes than elements like anthrop- or pyr- which can be analyzed
as the suppletive stems of lexemes HOMME, ‘man’ and FEU, ‘fire’ (see Amiot and Dal (2007)
for an analysis of these latter elements). These productive final elements are analyzed here as
exponents of constructions (Booij, 2010): they are the formal manifestation of semantic and
syntactic operations applied to lexemes, just as affixes are. These operations can be more or
less close to the original meaning of the neoclassical element. The final neoclassical element
-(o)phobie is different from the others in that it can be used as an autonomous lexeme and
therefore might be interpreted as the second element of a canonical compound. Nevertheless,
we argue here that its meaning is more restricted than when it is used as an autonomous
lexeme. In fact, the latter only displays the original meaning of the Greek noun φoβóς, ‘panic
fear’, already used in Ancient Greek as two bound forms, -φoβóς to form adjectives with the
meaning of ‘afraid of’ and, more rarely, the suffixed bound form -φoβια, with the meaning
‘fear of’. The semantic treatment of the French bound form -(o)phobie is quite different from
both Greek bound forms and French autonomous form. We indeed inventoried at least two
distinct meanings that the X-phobie construction can take.
(1) X-phobie constructions: a.
‘fear/dread of X’
‘aversion/hostility to X’
Meaning (1-a) refers to a psychopathological condition and meaning (1-b) to an inclination to
combat. Lexemes which were coined first and/or are the most frequent ones (e.g. xénophobie,
agoraphobie, claustrophobie, homophobie) play the role of leader words and behave as
attracting poles for the construction of the meaning of new lexemes (concerning the notion of
leader word, see, among others, Roché (2011)). Hence, the meaning of each new lexeme is
built in connection with the meaning of one or more actual lexemes. The question here is
about the way in which this connection is realized. In order to answer to this question, we
built a corpus based on both dictionaries (Le Grand Robert (online) and Le Trésor de la
Langue Française informatisé (TLFi)) and Google. It is constituted of 529 lexemes, of which
73 (13.8%) come from dictionaries.
A first way to determine which construction we are dealing with is to observe the bases of the
constructed words. We can indeed predict that when the base is [+ human], the whole lexeme
will contain the feature ‘hostility’ instead of ‘fear’, by analogy with leader words xénophobie
or homophobie. However, an analysis only built on the observation of bases is not sufficient:
a same lexeme can display different interpretations depending on its context.
The distributional hypothesis of Harris or Wittgenstein (see Sahlgren (2008) or Lenci (2008)
for a description and an application to semantics and computational linguistics) states that
there is a correlation between the meaning of a word and its distribution. If we implement this
for derivational morphology, we can consider that it is necessary to take into account the
context in order to determine the meaning or meanings of constructed lexemes. We applied
this hypothesis on our corpus in two ways. First, we investigated the distribution of the ten
most frequenti lexemes of our corpus in different resources: a Web sample represented by the
two first pages of Google results, all the distributions given in Frantext (a corpus of French
based on literary texts) and the results given in Les Voisins de Le Monde (http://redac.univtlse2.fr/voisinsdelemonde/) and Les Voisins de Wikipédia (http://redac.univtlse2.fr/voisinsdewikipedia/), two resources based respectively on ten years of the French
newspaper Le Monde and the French Wikipedia. The two latter are tagged for syntactic
contexts and are presented via an interface which allows searching the syntactic relations a
word enters into. After this first step, the most relevant contexts were applied to the whole
corpus by Google searches; for example, a context like "souffrir de", ‘suffer from’, which is
used in several resources with claustrophobie ‘claustrophobia’, was searched with all the
lexemes of our corpus, by typing each entry of our corpus preceding by the sequences
"souffrir de", "souffre de" and "souffres de" in Google, in order to have a representative
sample of different verbal forms.
‘fear, dread’
Google searches
X= christianophobie
< chrétien ‘Christian’
X = arachnophobie <
araignée ‘spider’
‘aversion, hostility’
contre * X”
“X ambiante”
‘protest against *
‘prevailing X’
“souffrir de X”
‘suffer from X’
“soigner * X”
‘treat * X’
Table 1 Example of results for a distributional analysis of -phobie
Obvious results, as those presented in Table 1, were highlighted. However, not every lexeme
could be interpreted in this way and a second analysis was necessary. To do so, we constituted
a sample of contexts for each lexeme in the corpus by listing its distribution in the two first
pages of results provided by Google. When it was possible, these contexts were indexed
according to semantic features (e.g. “traitement de X”, ‘treatment of X’ contains obviously
the feature ‘medicine, pathology’, whereas “lutter contre”, ‘fight against’ can be applied
either to a disease or an inclination) and all these distributions were then cross-referenced in
order to interpret each lexeme relative to each other. This method is representative of how we
view the organization of the lexicon and the creation of new words: a new word is coined in
connection with those already present in the lexicon and its meaning is constructed thanks to
several factors, including its distribution. The distributional hypothesis applied here allowed
us to distinguish different meanings for the outputs of an apparently unique construction. This
implies that distribution is as relevant as the formal construction itself for the formation of
new lexemes. Analogy takes into account not only the form but also the possible contexts of
occurrence. Distribution can also distinguish the metaphorical uses of an unique output.
AMIOT, D. and DAL, G. (2007). Intergrating neoclassical combining forms into e lexeme-based morphology. Online Proceedings of Fifht Mediterranean Morphology Meeting, ed. Geert Booij and Luca Ducceschi and Bernard
Fradin and Emiliano Guevara and Angela Ralli and Sergio Scalise:323-336.
BOOIJ, G. (2010). Construction Morphology. Oxford: Oxford University Press.
LENCI, A. (2008). Distributional semantics in linguistic and cognitive research. Italian Journal of Linguistics,
ROCHÉ, M. (2011). Des unités morphologiques au lexique. In Roché, M., Boyé, G., Hathout, N., Lignon, S., and
Plénat, M. editors, Des unités morphologiques au lexique, pages 69-143. Hermès Science.
SAHLGREN, M. (2008). The distributional hypothesis. Italian Journal of Linguistics, 20(1):33-54.
Frequency was calculated on the basis of the pages containing each lexeme indexed by Google
The rise and fall of a multifunctional affix: the case of -hAtÓ (-able) in Hungarian
Leiden University / Research Institute for Linguistics, H.A.S.
[email protected] / [email protected]
The issue
The received opinion on the Hungarian –hAtÓ (-able) affix holds that it is an
unifunctional derivational affix that always attaches to passivized transitive verbal roots and
derives adjectives from them (Kiefer & Ladányi 2000, Kiefer 2008, see also Wasow 1977 on
English -able): [V Vpass] + –hAtÓ = [AdjVpass–hAtÓ ]. Contrary to received wisdom, we will show
that this view is not tenable for the reason that the category of ható-affixed words is in many
cases demonstrably not adjectival: applying a battery of diagnostic tests for the adjectival
category only gives positive results for some but not all cases of ható expressions. The
availability of the comparative morpheme (-bb) for example singles out érthető 'understandable'
but not leültethető 'seatable' as an adjective, while both forms are perfectly acceptable as
prenominal modifiers, as the following data illustrate.
a. az
b. a
understand-ABLE-(COMP) texts
seat-ABLE-(COMP) texts
adjectival ható
non-adjectival ható
We will argue that the difference observed is that of syntactic complexity. While the adjectival
forms are indeed lexical adjectives, the non-adjectival forms are syntactic objects corresponding
to non-finite clauses. Although this might suggest that ható is a multifunctional affix – in one of
its uses a derivational affix (attaching to verbal roots in the lexicon), in another an inflectional
one (deriving non-finite clauses in the syntax) – appearances are deceptive here. Ható is in fact an
unifunctional affix and inflectional in all its occurrences: it always derives non-finite clauses,
some of whose verbal heads are lexicalized as adjectives.
Arguments for distinguishing adjectival vs. clausal ható forms
The two groups of ható
forms referred above can be distinguished on the basis of syntactic, semantic and
morphophonological evidence of various sorts. To cherry-pick a handful of these, next to their
compatibility with the comparative morpheme (cf. 1), the two forms also differ in (i) their
compatibility with intensifiers (cf. 2); (ii) their serving as input for further derivational processes,
like nominalizations (cf. 3); (iii) the retention of the event and argument structure of the
underlying verbal predicate (cf. 4), and (iv) the morphophonological realization of plural
morpheme they appear with (cf. 5).
a. egy {kifejezetten / nagyon} hosszú / ért-hető
understandable text
b. a problémára
{ * kifejezetten / jól}
alkalmaz-ható megoldás
the problem.ONTO
easily apply-ABLE solution
/ * leültet-hető-ség
a. a
*(gyerekeknek) ad-ható
the children.DAT
give-ABLE books
'the books that can be given to children'
b. a
(*Mari által) beszámít-ható emberek
the Mari
by count-ABLE people
'people who are sane to/for/by Mari'
a. A
piros színű gombák
nem e-hető{–k/–ek}.
the red coloured mushroom-PL not eat-ABLE-PL
'Red mushrooms are not edible.'
b. A legfiatalabb asszisztensek nem bíz-ható{–k/?*–ak} meg nehezebb feladatokkal.
the youngest assistants
not trustable-PL
difficult.COMP tasks.WITH
'The youngest assistants cannot be trusted with more difficult tasks.'
Further evidence for the syntactic clausehood of non-adjectival ható forms and the lexical
wordhood of ható adjectives will be provided from the realm of preverb placement, where ható
clauses show the same possibilities as finite (and some non-finite) clauses in Hungarian. After
focus and negation (and in some elliptical contexts not illustrated here), the preverb (PV)
belonging to the base verb is necessarily split off the verb and may appear at a distance from it
(see verb and preverb underlined), cf. (6). With adjectival ható expressions, such splitting is
never possible, which is predicted if these forms are lexical words and syntax does not have
access to their subparts (cf. Lexical Integrity; Selkirk 1982, Booij 1985).
János nem bíz-ható ma
meg ezzel
a feladattal.
not trust-ABLE today PV this.WITH the task.WITH
'János cannot be trusted today with the task.'
* János nem számít-ható ma
not count-ABLE today PV
(intended meaning) 'János is not sane' (lit. ‘countable on’)
clausal ható
adjectival ható
Ható as a unifunctional inflectional affix
To explain the above noticed differences, we will
argue that ható clauses are non-finite clauses with passive and modal semantics, and that ható is
in fact bimorphemic, containing -hAt and -Ó. -Ó spells out the [−finite] T head, while -hAt
lexicalizes Modality, which in turn selects a VoiceP >vP projection. The derivation of ható
clauses is thus akin to that of non-finite clauses and explains why ható clauses necessarily retain
their base verb's event and argument structure (cf. 4), and exhibit syntactic independence between
the verb and its preverb (cf. 6).
It will be shown that adjectival ható forms originate from syntactically derived ható clauses,
too, but they acquire their lexical status/adjectival category via the process of lexicalization.
Supportive evidence for lexicalization comes from the fact that ható adjectives (unlike ható
clauses) cannot be productively formed and that such a lexicalization process is also attested with
other non-finite verb forms that give rise to adjactives.
This novel account allows us to define ható as an inflectional affix in all its occurrences, and
this in turn leads to the conclusion that even though it gives the impression of an affix that can
attach to both word-level and phrase-level entities, it is in fact a unifunctional morpheme.
BOOIJ, G. E. 1985. Coordination reduction in complex words: A case for prosodic phonology. In: H. van der Hulst, N.
Smith (eds), Advances in nonlinear phonology, 143–60. Foris, Dordrecht. • KIEFER, F. 2008. A melléknevek szótári
ábrázolásáról. Strukturális magyar nyelvtan 4. 505-538. • KIEFER F. & M. LADÁNYI. 2000. A szóképzés. In: F.
Kiefer (ed.) Strukturális magyar nyelvtan 3, Morfológia, Akadémiai Kiadó, Budapest, 137-164. • SELKIRK, E. 1982.
The Syntax of Words. Cambridge MA, MIT Press. • WASOW, T. 1977. Ttransformations and the lexicon. In P.
Culicover, A. Akmajian and T. Wasow (eds), Formal syntax. NY: Academic Press.
Polyfunctional pronominal markers in South-Western Mande
Langage, Langues et Cultures d’Afrique Noire, CNRS
[email protected]
The paper deals with the syntactic functions and morphosyntactic features of polyfunctional
(PF) pronominal markers in South-Western Mande languages (SWM). This group includes 6
languages spoken in Guinea, Liberia and Sierra Leone. The central purpose of the paper is to
provide a comparative analysis of PF pronominal markers in SWM. This analysis is based on
the synchronous grammar descriptions and manuals on SWM languages as well as on the data
collected during the author’s fieldwork with Woi-Bhalaga, a Guinean dialect of Looma. This
work is supported by a public grant overseen by the French National Research Agency (ANR)
as part of the “Investissements d’Avenir” program (reference: ANR-10-LABX-0083)”.
It is typical for Mande languages to have several sets of pronominal markers opposed according to their syntactic functions and/or morphological status. One of them is often PF: for example, in Goo (Southern Mande) several sets of subject pronouns are opposed to a set of PF
pronouns occurring in non-subject positions (Vydrin 2013). Traditionally, PF markers in
SWM are regarded as personal pronouns, too. However, here they are not autonomous word
forms but rather inseparable and intransposable bound morphemes. Moreover, the paradigms
of PF markers include two suprasegmental morphemes, these expressing the 1st and the 3rd
person singular (except for Mende, where the 1st person singular is expressed by means of a
segmental morpheme ɲá-); they have no segment exponent and manifest themselves on the
following morpheme through the change of its initial consonant and tonal alternations. On this
evidence, V. Vydrin in (2011) proposes to name such elements in SWM personal prefixes.
Here, this term will be avoided in order not to refer to their morphological status.
The specificity of these morphemes consists in their transcategorial nature: they can be attached to a verb (1), a noun (2) or a postposition (3):
1sg.base 3sg.pi\see-aor
‘I saw him’.
kɛ̀ ɛ̀
‘my father’
Fòlòmò ɣà
níkɛ́ -y
Folomo 3sg.npst ref\cow-def sell-ipfv
‘Folomo is selling the cow to him’.
Particular functions which can be fulfilled by PF pronominal markers in SWM vary. The standard
set includes a direct object (4), an argument of a qualitative verb (5), a possessor in the situation of
inalienable possession (6) and an indirect object with a postposition (7):
1sg.ipfv 3sf.pf-do cop ref\day every
‘I do it everyday’. (Innes 1971: 150)
lè .́
2sg.pf-short-qual cop
‘You are not tall’. (Bandi, ms.: II, 17)
‘his hand’ (Babaev 2011: 60)
kɛ ̀ ɓɛ,́ èì
pù (`)-máà.
ref\chief.def 3sg.cond2 be here 3sg.cond2 ref\law put 1sg.pf-on
‘If only the chief was here, he’d punish me’. (Leidenfrost & McKay 2005: 77)
At the same time, there are a lot of smaller differences in the use of PF pronominal markers
in SWM. Thus, for example, SWM languages vary according to whether PF markers are necessary even with a full NP present or not. For example, in Looma in an intransitive resultative
construction, verbal agreement with a plural subject is facultative (8), while in Zialo it is
obligatory (9):
tí-ɓɛ́ tɛ́ -ʋɛ́
ref\man-def-pl 3pl.pi-be.wealthy-res /ref\man-def-pl
‘The men are wealthy’.
kpɛ́ tɛ́ -ʋɛ́ .
fèlè-gɔ̀ -y-tì
tí-w̃ ɛ́-yàà
ref\road two-def-def-pl 3pl.pf-meet-res exi
‘The two roads have been linked together’. (Babaev 2011: 129)
Especially interesting cases represents the use of PF markers for marking inalienable possession on the head noun. In this case their use may depend on a whole range of factors, such
as the type of a noun representing possessum, the meaning of the category of number expressed by a NP in the possessor’s position, its referential status, etc. In the report, a detailed
analysis of such peculiarities of PF pronominal markers in SWM will be proposed.
(A Bandi Grammar). Unpublished Manuscript. Monrovia: The Institute for Liberian languages vol. I-II.
BABAEV, K. 2011. Zialo Language: Grammar Sketch and Dictionary. Moscow: HSSU. (in Russian)
INNES, G. 1971. A Practical Introduction to Mende. London: School of Oriental and Classical Studies.
LEIDENFROST, T., MCKAY, J. 2005. Kpelle–English Dictionary with a grammar sketch and English-Kpelle
finderlist. Moscow (USA): Palaverhut Press.
VYDRIN, V. 2011. Ergative/Absolutive and Active/Stative alignment in West Africa: The case of Southwestern
Mande. In: Studies in Language 35:2, 409-443.
VYDRIN, V. 2013. Verbal construction with a progressive meaning in Goo: Transformation of non-subject pronouns to subject ones. Handout, The Third conference “Typology of Morphosyntactic Parameters”, Moscow, 1618 October 2013. (in Russian)
Subjective perception of affixation: A test case from Spanish
University of Wisconsin, Milwaukee
[email protected]
of noise
Level_4, Level_5
Affixation changes our perception of a word, but in ways that we do not fully
understand. Typological studies point to the prevalence of suffixation over prefixation as
evidence that suffixed words provide perceptual advantages, because they place the root in
initial position (Hawkins & Cutler 1988). But this idea is at odds with the experimental
literature, which shows that listeners “strip” prefixes from multimorphemic words (Taft &
Forster 1975), and provides evidence for a prefixation advantage: the root pay and the
prefixed word prepay facilitate recognition of payment, but the suffixed word payable inhibits
it (Marslen-Wilson et al. 1994; Zwitserlood, & Roelofs 1991, Feldman & Larabee 2001).
Sorting out the perceptual consequences of affixation is important for models of word
recognition, and could help explain problems in phonology, such as the lack of prefixtriggered alternations on roots (Hyman 2008).
The current study investigates this issue using a noise-rating task in Spanish.
Participants heard Spanish phrases that had been partially overlaid with white noise, and
assigned a rating indicating how loud the noise sounded, on a scale from 1 (softest) to 5
(loudest). The phrases consisted of verbs plus pronouns; the noise, indicated by strikethrough,
coincided with either the prefix (me patea ‘she kicks me’, se me pisa ‘I am stepped on’), the
root, or the suffix (patéame, ‘kick me’, písamelo ‘step on it for me’). Spanish offers a good
test case for our question because the same personal pronoun clitics, which behave
phonologically like affixes (Hualde 2005, 2012), can precede or follow the verb root, and
need not occur at word edges – properties that allow us to focus on affixhood per se. The
noise-rating task permits a measurement of listeners’ subjective experience of the word, with
the premise that listeners will experience the overlaid noise as softer when the spoken word
itself seems perceptually clearer (Jacoby et al. 1988).
A speaker of Colombian Spanish recorded the words. The intensity of each root and
affix was calculated separately and white noise was added accordingly at one of three signalto-noise ratios: +24 dB, +17 dB, or +10 dB. Forty-four native Spanish-speaking participants
each heard 36 target phrases, plus 80 fillers, and were asked to attend to phrase meanings
during the rating task. Results, analyzed with proportional odds logistic regression, showed an
interaction between Noise location andNoiseLocation*Position
Affix type (β = -1.37,
t = -7.10, p <0.01).
effect plot
Prefixed Words
Position : Preposed
Suffixed Words
Position : Postposed
L_3 - L_4
L_3 - L_4 L_3 - L_4
L_3 - L_4
L_2 - L_3
L_2 - L_3 L_2 - L_3
L_2 - L_3
L_1 - L_2
Affix Root
L_1 - L_2 L_1 - L_2
L_1 - L_2
Participants rated noise on suffixed roots as very loud, suggesting diminished perceptual
clarity, but rated noise on prefixed roots as relatively soft, suggesting perceptual
enhancement. These findings confirm the previously reported prefixation advantage, and
extend it by suggesting that prefixes facilitate priming of related words precisely because they
provide listeners with enhanced perception of the root. Interpreted in light of listener-based
theories of diachronic change (Ohala 1993), this finding could help to explain the lack of
prefix-root phonological alternations cross-linguistically: just as certain sequences of vowels
and consonants encourage misperception more than others, so do certain sequences of
FELDMAN, L. B., & LARABEE, J. (2001). Morphological facilitation following prefixed but not suffixed primes:
Lexical architecture or modality-specific processes?. Journal of Experimental Psychology: Human
Perception and Performance, 27(3), 680.
HAWKINS, J. A., & CUTLER, A. (1988). Psycholinguistic factors in morphological asymmetry. Explaining
language universals, 280-317.
HUALDE, J. I. (2005). The sounds of Spanish. Cambridge University Press.
HUALDE, J. I. (2012). Stress and Rhythm. The Handbook of Hispanic Linguistics. In Hualde, J. I., Olarrea, A., &
O'Rourke, E. (Eds.). The Handbook of Hispanic Linguistics (Vol. 69). (pp. 153-172). Wiley.
HYMAN, L. M. (2008). Directional asymmetries in the morphology and phonology of words, with special
reference to Bantu. Linguistics, 46(2), 309-350.
JACOBY, L. L., ALLAN, L. G., COLLINS, J. C., & LARWILL, L. K. (1988). Memory influences subjective
experience: Noise judgments. Journal of Experimental Psychology: Learning, Memory, and Cognition,
14(2), 240.
MARSLEN-WILSON, W., TYLER, L. K., WAKSLER, R., & OLDER, L. (1994). Morphology and meaning in the
English mental lexicon. Psychological review, 101(1), 3.
OHALA, J. J. (1993). The phonetics of sound change. In Charles Jones (ed.), Historical Linguistics: Problems and
Perspectives (pp. 237-278). London: Longman.
SCHRIEFERS, H., ZWITSERLOOD, P., & ROELOFS, A. (1991). The identification of morphologically complex
spoken words: Continuous processing or decomposition?. Journal of Memory and Language, 30(1), 26-47.
TAFT, M., & FORSTER, K. I. (1975). Lexical storage and retrieval of prefixed words. Journal of verbal learning
and verbal behavior, 14(6), 638-647.
Gender neuterisation in the light of language contact: the case of Pontic & Heptanesian
University of Patras
[email protected], [email protected] & [email protected]
From a theoretical perspective, Corbett (1991) argues that gender is an inherent property of
nouns and its assignment depends on semantic, pragmatic, phonological and morphological
factors. Ralli (2002) maintains that gender constitutes a lexically-specified feature, actively
involved in the word-formation process of Standard Modern Greek (hereafter SMG), where a
tripartite gender distinction into masculine, feminine and neuter characterizes nouns,
adjectives and determiners.
In this paper, we examine grammatical gender assignment in loanwords of two
different contact-induced dialectal systems of Modern Greek (hereafter Greek), Pontic and
Heptanesian. Pontic has been affected by the agglutinative Turkish, while Heptanesian has
been influenced by the semi-analytic Romance (Italian and Venetian). Data drawn from these
dialects reveal similarities as well as deviations in gender assignment compared to that in
SMG, and confirm the significance of gender as an organizational feature of grammar
(Corbett 1991, Ralli 2002). More particularly, we show that in both Pontic and Heptanesian,
gender assignment in loanwords is primarily conditioned by language-internal properties of
the recipient system, that is, an old attested tendency for shifting to the neuter value of certain
types of nouns (Hatzidakis 1907) and the prevalence of the semantic feature of humanness,
while a phonological similarity between the inflectional endings of the donor and those of the
recipient language also play a role. We argue that paradigms of loanwords are formulated
accordingly, except for some rare but interesting cases, and try to interpret a number of
peculiarities that arise in gender agreement mechanisms involving a determiner and a noun,
by appealing to a language contact effect.
First in Pontic, although the tripartite grammatical gender distinction is maintained,
there is a dominant tendency to neutralize gender distinctions in the context of plural in favor
of the neuter value, affecting both native (1c-d) and loan words (1a-b), while the donor
language, i.e. Turkish, is gender-neutral, at least on the morphological level. Interestingly,
when gender leveling occurs, it may affect both the determiner and the noun (1a) or only the
determiner (1d) (Papadopoulos 1955). As a consequence, there are repercussions in the
agreement patterns of the particular dialect (occasional absence of agreement between the
determiner and the noun (1d)), as well as a reformulation of the inflectional paradigms.
(1)a. i giurulti ~ ta giurultia [the.FEM.NOM.SG noise.FEM.NOM.SG ~ the.NEU.NOM.PL
noise.NEU.NOM.PL] < Turkish gürültü : ‘the noise(s)’  b. i orospi ~ i orospiδes
[the.FEM.NOM.SG prostitute.FEM.NOM.SG ~ the.FEM.NOM.PL prostitute.FEM.NOM.PL] < Turkish
rospɪ : ‘the prostitute(s)’  c. i mana ~ i manaδes / ta manaδes [the.FEM.NOM.SG
mother.FEM.NOM.PL] (Ofis Pontic): ‘the mother(s)’  d. i kossara ~ ta kossaras
[the.FEM.NOM.SG chicken.FEM.NOM.SG ~ the.NEU.NOM.PL chicken.FEM.NOM.PL]: ‘the
chicken(s)’  e. o minas ~ ta minas [the MASC.NOM.SG month MASC.NOM.SG ~ the NEU.NOM.PL
month MASC.NOM.PL] ‘the month(s)
Second, the Heptanesian data also depicts a certain prominence of the neuter gender,
contrary to Romance, where there are only masculine and feminine gender values expanding
even on native examples (2c). This prominence is shown in both the determiner and the noun
and in both number values, singular and plural. Contrary to Pontic, however, agreement is
preserved between the determiner and the noun, as depicted in (2a, d).
(2)a. i belandza/to belandzi ~ i belandzes/ta belandzia [the.FEM.NOM.SG
weighing.scale.FEM.NOM.SG/ the.NEU.NOM.SG weighing.scale.NEU.NOM.SG ~ the.FEM.NOM.PL
wheighing.scale.FEM.NOM.PL / the.NEU.NOM.PL wheighing.scale .NEU.NOM.PL] < Italian la
bilancia : ‘the scale(s)’  b. to tsekini [the.NEU.NOM.SG venetian.gold.coin.NEU.NOM.SG] <
Venetian lo zéchin: ‘venetian gold coin’ c. to maγerjo ~ ta maγerja [the.NEU.NOM.SG
kitchen.NEU.NOM.SG ~ the.neu.NOM.PL kitchen.NEU.NOM.PL] : ‘the kitchen(s)’  d. to sokors-o
[the.NEU.NOM.SG assistance.NEU.NOM.SG] < Venetian il socόrs-o : ‘assistance’  e. i indovino
~ i indovines / ta indovina [the.FEM.NOM.SG clairvoyant.NEU.NOM.SG ~ the.FEM.NOM.PL
clairvoyant.FEM.NOM.PL/ the.NEU.NOM.PL clairvoyant.NEU.NOM.PL] < Italian l’indovino ~ gli
indovini : ‘the clairvoyant(s)’
In order to incorporate loanwords of Turkish origin, Pontic activates semantic,
phonological and morphological mechanisms, which do not differ from those employed by
SMG, Heptanesian and the majority of Greek dialects. For instance, there is an innate
tendency to assign masculine and feminine values to male and female nouns respectively,
which may be enhanced by the interference of the semantically-implied genderness (1b-c)
(covert gender according to Aronoff 1998) but overtly genderless Turkish. Generally,
grammatical gender assignment in Pontic confirms Ralli’s (2002) suggestion that the
lexically-specified gender is not part of the features of the inflectional suffixes but belongs to
those of stems, since gender in Pontic words is morphologically realized on the stem and/or
the determiner. Nevertheless, a question arises why the Pontic data conflict with the
corresponding SMG one as far as agreement and the predominance of the neuter value are
concerned, as well as why shift to neuter gender occurs only in the plural number.
A semantic effect also applies in Heptanesian, which bears witness to the effect of
humanness on gender, since -human loanwords are predominantly neuter (2b). It is also
corroborated by a tendency to shift to the neuter value loanwords which belong to a different
gender, e.g. feminine, as in (2a), for phonological reasons, that is, because of a form similarity
between the endings of the donor (Italian/Venetian) and the recipient language
(Greek/Heptanesian). This tendency seems to be stronger than in Pontic, since it may alter
even +human nouns (2e), affects both the determiner and the noun, and is spread throughout
the entire paradigm of singular and plural. In addition, like Pontic, Heptanesian brings to the
forefront the linguistic property of Greek to have the stem and/or the determiner lexically
specified for gender. Again, for loanwords, this property seems to be enhanced by a similar
tendency in the donor language, i.e. Italian/Romance, as stated by Thornton (2003).
Interestingly in Pontic, the shift of feminine and masculine nouns to the neuter gender
seems to be related to the semantic +/-human feature, and has an impact on the agreement
between the determiner and the noun. In certain areas of Pontus (e.g. Ofis), an innovative
absence of agreement is systematically attested in the plural of +/-human feminine nouns (1c),
while in other areas, this absence is sporadic, related to -human masculine and feminine ones
(1d,e). We believe that contact with the genderless Turkish must have played a role into it. In
contrast, Heptanesian, which has been in contact with a language with gender distinctions,
shows a systematic agreement in both singular and plural, and whenever gender shifts to
neuter, the noun is combined with a determiner marked for neuter.
To sum up, this paper demonstrates that the integration of loans in a recipient
linguistic system which bears an overtly marked gender may offer insights into grammatical
gender assignment. It also provides evidence that in contact morphology, this feature results
from the interplay of language-internal factors of both the recipient and the donor languages.
ARONOFF, M. 1998. Gender Agreement as Morphology. Proceedings of the 1stMediterranean Morphology
Meeting, ed. by G. Booij, A. Ralli & S. Scalise. Mytilene 1997, 7-18.  HATZIDAKIS, G. 1907. Meseonika ke nea
Elinika [Medieval and Modern Greek], v. 2. Athens: Sakellariou.  CORBETT, G. 1991. Gender. Cambridge:
Cambridge University Press. PAPADOPOULOS, A. 1955. Istoriki Grammatiki tis Pontikis Dialectou [Historical
Grammar of the Pontic Dialect]. Athens: Epitropi Pontiakon Meleton. RALLI, A. 2002. The Role of Morphology
in Gender Determination: Evidence from Modern Greek. Linguistics 40: 519-551. THORNTON, ANNA M. 2003.
L’assegnazione del genere ai prestiti inglesi in italiano [Gender assignment in English borrowings in Italian].
Italiano e inglese a confronto, ed. by Anna-Vera Sullam Calimani. Firenze: Cesati, 57-86.
Approaching case polysemy from the lexicon:
The case of the Kabardian instrumental
NRU Higher School of Economics
[email protected]
NRU Higher School of Economics
[email protected]
Institute of Slavic Studies (Russian Academy of Sciences), Russian State University for the
Humanities, Sholokhov Moscow State University for the Humanities
[email protected]
The instrumental case in Circassian languages Adyghe and Kabardian (North-West Caucasian
language family) is highly polysemous, its uses ranging from typical instrument and means to
the encoding of spatial and temporal relations and even such functions as iudicantis and stimulus (Kuznetsova & Serdobolskaya 2009). Though it is fairly obvious that not all functions of
the instrumental arise when the case marker combines with any lexical noun, semantic interaction of the instrumental case with nominal lexemes has not been studied in any detail.
Likewise, this issue has not been raised in the detailed cross-linguistic studies of the polysemy
in the domain of the instrumental such as Narrog & Ito (2007) and Lehmann & Shin (2005),
and general works on case polysemy, e.g. Malchukov & Narrog don’t treat it either.
We follow Aristar (1997), Rakhilina (2000) and Luraghi (2003) in assuming a crucial role of
nominal lexical semantics in the functioning and interpretation of case, and empirically test
the hypothesis that the functions of the highly polysemous Circassian instrumental case are
non-randomly distributed according to the lexical-semantic classes of nominals. The data for
our research come from the Besleney dialect of Kabardian spoken in the village Ulyap in Republic of Adygheya (Russia), collected during the field-trip July 2013.
Our investigation is based on a questionnaire containing fifty nominal lexemes; we asked native speakers to compose naturally sounding sentences containing the target noun in the instrumental case. The results of the experiment show that Besleney nouns fall into several classes according to their co-occurrence with the meanings of the instrumental case.
1) Nouns which resist co-occurrence with the instrumental altogether, i.e. those for which our
consultants experienced difficulties of providing naturally sounding contexts; these are e.g.
nouns denoting animals (‘fly’, ‘fox’).
2) Nouns co-occurring with a wide range of the meanings of the instrumental, e.g. those denoting humans. Thus, the instrumental form of ‘father’ can have the functions of iudicantis
(‘it’s good for father’), spatial direction (‘go towards father’), stimulus (‘be content with father’), specification (‘paternal uncle’, lit. ‘uncle father-INS’).
3) Nouns naturally co-occurring with just one meaning of the instrumental, like nouns denoting prototypical instruments (‘hammer’) or food (‘to feed with porridge’).
4) Nouns regularly combining with several meanings of the instrumental. These fall into two
4.1) Nouns combining with semantically similar meanings of the case. E.g. with nouns denoting spatial objects the instrumental can, depending on the type of verb, express motion towards, from or via the landmark, while the combination of the instrumental with temporal
nouns express temporal extent (‘run home in half an hour’), temporal distance (‘return in a
year’) and temporal location (‘at night’).
4.2) Nouns which due to their lexical semantics are compatible with heterogeneous semantic
roles. E.g. nouns denoting cattle when combined with the instrumental can express exchange
equivalent (‘exchanged a cow for a horse’) and means of transport (‘ride a horse’), while
body-part nouns can be used as instruments (‘take with the hand’), point of application (‘hit
one’s head’) or beneficiary (‘good for the eyes’).
On the basis of this study we draw the following conclusions:
1) Individual functions of the instrumental case are distributed across lexical classes of nominals in a principled manner, which suggests that each function selects those lexical classes
which are most semantically compatible with it. This closely resembles the interaction between verbs and argument structure constructions (Goldberg 1995) or prefixes (Janda et al.
2013), which form radial networks where particular types of verbs combine with particular
constructional or prefixal meanings.
2) According to their compatibility with the instrumental, nominals can be divided into “core”
and “peripheral” (cf. Aristar 1997). The “core” are those which are robustly associated with a
single function or with a family of related functions of the case. To the “periphery”, in addition to those nominals which do not readily combine with the instrumental, belong those
whose instrumental form can yield a wide range of meanings, usually more generalized and
less “selective” than the core meanings of the instrumental, cf. such functions as beneficiary
or stimulus. Both the wide range of possible interpretations of such nouns in the instrumental
and inter-speaker variation in their interpretation are probably due to the low frequency of instrumental forms of the nouns of these lexical classes.
We conclude that explicitly taking into account and controlling for lexical semantics is necessary for an adequate characterization of case polysemy and that looking at the meanings of
case from the perspective of lexical semantic classes of nominals it combines with is a promising direction of empirical and cross-linguistic research of morphological case and related
functional domains.
ARISTAR A.R. (1997). Marking and hierarchy types and the grammaticalization of case-markers. Studies in Language 21, 313–368.
GOLDBERG A. (1995). Constructions. A Construction Grammar Approach to Argument Structure. Chicago, London: The University of Chicago Press.
Why Russian Aspectual Prefixes Aren’t Empty. Prefixes as Verb Classifiers. Bloomington: Slavica.
KUZNETSOVA J. & N. SERDOBOLSKAYA (2009). Dvojnoe padežnoe markirovanie: unikal’nyj slučaj adygejskogo
jazyka [Double case marking: A unique case in Adyghe]. In: Ya.G. Testelets et al. (eds.), Aspekty polisintetizma:
Očerki po grammatike adygejskogo jazyka. [Aspects of Polysynthesis: Studies in the Grammar of Adyghe] Moscow: RGGU, 166–200.
LEHMANN CHR., SHIN Y.-M. (2005). The Functional Domain of Concomitance. A Typological Study of Instrumental and Comitative Relations. Berlin: Akademie Verlag.
LURAGHI S. (2003). On the Meaning of Prepositions and Cases. The Expression of Semantic Roles in Ancient
Greek. Amsterdam, Philadelphia: John Benjamins.
MALCHUKOV A., H. NARROG (2009). Case polysemy. In: A. Malchukov & A. Spencer (eds.) The Oxford Handbook of Case. Oxford: Oxford University Press.: 518–534.
NARROG H., ITO SH. (2007). Re-constructing semantic maps — The Comitative-Instrumental area. Sprachtypologie und Universalienforschung 60, 273–292.
RAKHILINA E.V. (2000). Semantika russkogo tvoritel’nogo i taksonomija [The semantics of the Russian instrumental and nominal taxonomy]. In: E.V. Rakhilina. Kognitivnyj analiz predmetnyx imёn: semantika i sočetaemost’. [Cognitive Analysis of Object-denoting Nouns: Semantics and Co-occurrence] Moscow: Russkie slovari,
A new semantic map approach to account for the development of polysemy in affixes
Bielefeld University
[email protected]
This paper presents a new approach to account for the polyfunctionality of derivational
affixes, namely a significantly adapted version of semantic maps. Semantic maps have until
now mainly been used to account for the semantics of inflectional categories across a variety
of languages (e.g. Haspelmath 2003). In this paper, an adapted version of the semantic map
approach will be introduced. The advantages of this adapted approach will be illustrated by a
diachronic study of the semantics of two English derivational suffixes, -age and -ery.
Semantic maps represent the different functions of grammatical categories and the
connections between these on a two-dimensional map. This approach has a number of
advantages over other ways of representing affix semantics: it is straightforward to compare
the semantics of different grammatical categories, and it is easy to account for language
change. Another important advantage is the fact that semantic maps do not make a priori
assumptions regarding the structure of a morphological category. For example, if a category
has a single core sense and sense extensions of this primary sense, a semantic map can depict
this easily. But if a category shows a different structure this can also be represented on a
semantic map. Spatial prepositions, for example, are often expected to exhibit a prototypical
structure (e.g. Tyler & Evans 2001). Semantic maps do not make such a priori assumptions as
they utilise a bottom-up approach, and are thus very open and flexible. However, traditional
maps such as Haspelmath's have been developed as a means for cross-linguistic comparison,
and are mainly used for inflectional categories. The adapted approach presented here can
account for the semantics of derivational affixes, and makes it possible to compare different
affixes within a single language as well as across different languages. It is also a particularly
suitable tool for representing the semantics of a single affix in different time periods of a
language, and can thus provide valuable information on the nature of language change
regarding the semantics of derivational affixes. As a major improvement over traditional
semantic maps, this approach incorporates frequency information in a new way (see also
Cysouw 2007 for an argument regarding the importance of this issue).
The study that provides the data to illustrate the adapted semantic map approach takes
both dictionary and corpus data into account. The Oxford English Dictionary (OED), an
extremely large historical dictionary, enables a large-scale diachronic investigation, which
compares the semantics of -age and -ery neologisms in Middle English (1100-1499) to
Present Day English (since 1900). The results of this investigation are then compared to -age
and -ery derivatives in the British National Corpus (BNC). This corpus represents language
use in the late 20th century.
Semantic maps are then created for each of the suffixes, periods, and data sources
investigated. Apart from being a good visual representation of the morphological categories
under investigation, the analysis of these maps provides a number of valuable insights into the
nature of affix semantics in general. The analysis of -age derivatives, for example, reveals that
this category has more than one core sense. In Middle English, the readings ACTION and
are core interpretations, but in Present Day English, this has changed to ACTION and
not only the core senses have shifted since Middle English, some readings that
were commonly used then have disappeared completely since. There has clearly been a
significant amount of change in the semantics of this suffix in the last 500 years. All of this is
illustrated on the semantic maps.
Another important question the maps help to answer regards the development of
polysemy in affixes. The connecting lines make polysemy visible, and change in the amount
or kind of polysemous derivatives can be represented in a straightforward manner. The maps
show some already well-documented cases of sense extension, e.g. ACTION → OBJECT, but
also reveal unexpectedly close links between the core senses and between other readings as
well. The diachronic analysis shows how the connections between these readings are
strengthened over time, which leads to a highly interconnected polysemous category with
many regular semantic overlaps and sense extensions. There are also significant differences in
the kind of polysemy shown by -age and -ery derivatives. Although the semantic contribution
of these two suffixes is often said to be very similar (e.g. Lieber 2004), the semantic maps
reveal significant differences in the connections between the various readings of the
derivatives. This shows that the adapted semantic map method is a good tool for bringing
subtle aspects of affix semantics to light.
CYSOUW, M. (2007). Building semantic maps: The case of person marking. In M. Miestamo & B. Wälchli (Eds.),
New challenges in typology. Berlin: Mouton de Gruyter, 225-247. • HASPELMATH, M. (2003). The geometry of
grammatical meaning: Semantic maps and cross-linguistic comparison. In M. Tomasello (Ed.), The new
psychology of language (Vol. 2). Mahwah, NJ: Erlbaum, 211-242. • LIEBER, R. (2004). Morphology and lexical
semantics. Cambridge: Cambridge UP. • TYLER, A. & EVANS, V. (2001). Reconsidering polysemy networks: The
case of over. Language, 77(4), 724-765.
Functions and typology of the compounding stem:
meaning-independent elements in compounds
University of Tsukuba
Tohoku University
[email protected]
[email protected]
University of Tsukuba
University of Tsukuba
[email protected]
[email protected]
It has long been observed that compounds may exhibit semantically-unmotivated
morphophonological elements (Bloomfield 1933), representative examples of which are
linking elements found in inflectional languages, e.g., (1a) from languages with stem-based
inflection and (1b) from languages with word-based inflection. Ralli (2008) argues that the
primary function of linking elements is to mark the process of compounding, i.e. they are
compound markers (CM henceforth). Using the framework of lexeme-based morphology,
we attempt to extend the CM analysis to various meaning-independent morphophonological
variations within compounds observed in non-inflectional languages or categories. Briefly, we
 Compounding employs special stems or compounding stems (Aronoff 1994) in order to
mark the wordhood of the entire construction.
 Compounding stems are morphomic, being formed by formal means available in each
language in the stem-formation component (Spencer 2012). Inflectional languages (or
categories) use inflectional stems or inflected forms as compounding stems, while
non-inflectional languages (or categories) use other formal marking options available in
their analytic grammars.
While Ralli mainly focuses on morphological CMs from inflectional languages like (1), we
take the view that CMs can be syntactic or phonological formatives because irrespective of its
morphological typology, any language needs to mark its compound unit in some way. In our
view, morphological typology concerns not whether but how a particular language marks the
First, the necessity of such a view is suggested by French data in (2). Given Ralli’s (2008)
description of French as inflectionally-rich in verbs but poor in nominal inflection, the CM
status she ascribes to P in the [N-P-N] compound in (2b) suggests that non-inflectional
categories use non-inflectional formatives to mark the compoundhood. Thus, in (1) and (2),
the non-head constituents including the underlined formative constitute compounding stems.
(1) and (2a) illustrate stem formation dependent on inflectional paradigms, while (2b) is a
case of stem formation by a syntactic functional morpheme.
Second, the stem analysis is preferable because the compound status is marked not only
syntagmatically but also via morphophonological alternation. Thus, the compound stress
exhibited by many English nominal compounds can be seen as (a component of) the
compounding stem that an analytic language like English can make use of in order to make
the word-phrase distinction. On the other hand, so-called neoclassical compounds (e.g.
anthropology) instantiate cases when the compounding stem is suppletive. The bound form
anthrop is a suppletive stem of the lexeme MAN (Amiot & Dal 2008). The use of the
suppletive bound stem can be seen as a marker of compounds with the stylistic feature
[+learned]. Hence, the stem analysis can encompass both major and minor types of English
compounds. Additionally, Nübling & Szczepaniak (2008: 3) note the existence of “subtractive
linking elements” in German (e.g. Wolle “wool” > Woll-kleid “woolen dress”), which is most
naturally treated in the lexeme-based stem analysis.
Like French, Japanese has paradigmatic inflection for verbs but not for nouns, so Ralli’s
(2008) parameters predict the existence of CM for Japanese verbs but not for nouns. Our
analysis, on the other hand, predicts that Japanese verbs and nouns show different types of
compounding stems. (3) supports the latter prediction. As in (3a), Japanese verbs always take
the so-called adverbial word-form in the compound non-head, independently of semantics. In
constrast, Japanese nouns exhibit syntactically or phonologically derived compounding stems;
(3b) shows that the particle no can appear in N-N compounds as a CM (Mukai 2008), while
(3c) shows that the second noun of N-N compounds can take an initially voiced allomorph
(Itô & Mester 1986). In addition, (3d) illustrates the suppletive compounding stem.
Chinese is a language extremely poor in overt inflection but very rich in compounding. If
the compound status is to be marked in grammar, this language should possess some formal
device to do so. (4) indicates that the word status of the whole combination is marked by the
bimorphemic morphophonological template. First of all, Modern Chinese vocabulary consists
mainly of bimorphemic words, which developed to compensate for the increased homophony
among (originally) monomorphemic words (Sampson 2013). The evidence for the
bimorphemic structure itself now functioning as a CM comes from the fact that when
compounded, bimorphemic input words can undergo a sort of morpheme reduction to produce
still bimorphemic output words. The data in (4) taken from Packard (2000) and Ceccagno &
Basciano (2009) may look like blending, but crucially differ from it in that the outputs in (i)
consist of two morphemes and also in that the reduction process is obligatory, hence the
ungrammaticality in (ii). Admittedly, it is difficult to view the bimorphemic word template
itself as a compounding stem. Under the stem analysis, the proposal above can be translated
as follows: each lexeme produces a monomorphemic stem for compounding. Bimorphemic
lexemes choose one of the constituent morphemes as their compounding stem based on the
informativity (Shaw et al. 2013).
(1) a. Greek: kukl-o-souto “doll house,”
Polish: ostr-o-słup sharp-LE-pillar “pyramid” (Szymanek 2009)
b. German: Wirt-s-haus host-LE-house “inn,” Jahr-es-zeit year-LE-time “season”
(2) a. French: essuie-mains “hand towel,” porte-drapeau “standard bearer” (Fradin 2009)
b. French: moulin à vent “wind mill,” étoile de mer “sea star”
(3) a. TATSU “stand” >
tachi-saru stand-leave “go off,”
tachi-sugata stand-figure “standing posture”
b. kitsune-no tebukuro fox-of glove “foxglove, digitalis,”
ama-no gawa heaven-of river “Milky Way”
c. KITSUNE “fox” >
no-gitsune field-fox “wild fox”
d. TATSU “stand” >
ritsu-an stand-idea “planning,”
KITSUNE “fox” >
byak-ko white-fox “white fox”
(4) a. căo “straw”
mào-zi hat-AFFIX “hat”
(i) căo-mào “straw hat”
vs. (ii) *căo-màozi
b. yì “righteous” +
pāi-mài “auction”
(i) yì-pāi “charitable auction” vs. (ii) *yì-pāimài
c. wèi-xīng “satellite” + diàn-shì “television”
(i) wèi-shì “satellite TV”
vs. (ii) *wèixīng-diànshì
Selected References
ARONOFF, M. (1994) Morphology by Itself, MIT Press. /MUKAI, M. (2008) “Recursive compounds,” Word
Structure 1. /NÜBLING, D. & R. SZCZEPANIAK (2008) “On the way from morphology to phonology: German
linking elements and the role of the phonological word,” Morphology 18. /RALLI, A. (2008) “Compounding
markers and parametric variation,” Language Typology and Universals 61. /SAMPSON, G. (2013) “A
counterexample of homonymy avoidance,” Diachronica 30. /SHAW, J, C. HAN & Y. MA (2013) “Surviving
truncation: Informativity at the interface of morphology and phonology,” a paper presented at the symposium
Morphology and Its Interfaces (Université Lille 3). / SPENCER, A. (2012) “Identifying stems,” Word Structure 5.
Paradigm leveling in non-standard Russian
HSE Moscow & St. Petersburg State University
St. Petersburg State University
[email protected]
[email protected]
In this paper, we look at a paradigm leveling process currently taking place in Russian that
affects historic consonant alternations. In standard Russian, these alternations are present in
some verb forms, in comparatives and after certain derivational suffixes (e.g. suxoj ‘dry’ /
suxo ‘dryly’ – suše ‘drier, more dryly’, ljubit’ ‘to love’ – ljublju ‘I love’, noga ‘leg’ – nožka
‘small leg, furniture leg’). But many non-standard forms lack alternations or have ‘incorrect’
alternations unattested in standard Russian. We will look at verb forms also analyzed in our
previous work (Slioussar & Kholodilova 2013) and, primarily, at comparatives.
Unfortunately, Russian corpora contain almost no non-standard forms that we were interested
in, so we had to look for them on the Internet. Estimating relative frequencies of different
forms found there is a challenge because the counts provided by search engines are extremely
unreliable. To circumvent this problem, we came up with several strategies. E.g. we
established what variants of a particular form are attested (see below), included all variants in
one search, i.e. asked the search engine to look for them simultaneously, sorted the results by
date, counted frequencies of different variants and then did the relevant statistical tests. We
developed a collection of Perl scripts Lingui-Pingui that greatly facilitates this task. This
program can automatically form queries from the list of morphs, send them to the search
engine, download the results, sort them according to specified criteria and do various counts.
Let us look at the non-standard innovation zafrendit' ‘to include in one’s friend list’ as an
example. The following 1SG future forms of this verb can be found on the Internet: zafrenžu
with the standard d // ž alternation, zafrenždu with a d // žd alternation originally coming from
Old Church Slavonic, zafrendju lacking alternation and several variants with alternations
unattested in standard Russian. Among them are zafrendlju, zafrendžu, zafrenču, zafrendču,
zafrendšu and zafrenšču. These alternations result from the inappropriate use of epenthetic l
(d // dl), adding the alternating consonant to the stem rather than replacing the final consonant
(d // dž), choosing a wrong alternating consonant (d // č, d // šč) or the combination of two
last strategies (d // dč, d // dš). Three of these forms allow the speaker to kill two birds with
one stone: to have an alternation and to keep the stem constant. Not to miss any variants, we
looked for all possible combinations of alternating consonants, stem-final d and epenthetic l.
1. We can show that very often, leveling simultaneously goes in two opposite directions.
However, some innovations are more frequent than the others. In particular, underapplication
of consonant alternations is more widespread than overapplication. We will discuss this
problem in detail for the verbs from the G-K verb class like žeč' ‘to burn’, where numerous
non-standard examples of both types can be found. Although the present/future paradigm of
G-K verbs includes four forms with alternations and two forms without them, innovative
present/future forms lacking alternations prevail dramatically.
2. Looking at various verbs from the I class (like the non-standard zafrendit' above) we can
demonstrate that leveling is influenced by the following factors. Firstly, less frequent words
lack alternations more often. Secondly, non-standard words have more forms without
alternations. Thirdly, the proportion of forms lacking alternations is the highest for stems
ending in obstruent clusters and the lowest for the stems ending in labials (where the
epenthetic l is standardly used).
3. For the study of alternations in comparatives two groups of adjectives were selected: ones
that have normative synthetic comparatives with alternations (this group is not productive)
and ones that do not, but native speakers still tend to generate such forms. In the first group,
the process of alternation loss is more sporadic than in the case of verbs: some adjectives like
ubogij ‘poky’ or uprugij ‘resilient’ have up to 30% of comparatives without alternations, but
no significant correlation between the number of non-standard forms and lemma frequency or
the last consonant of the stem could be found. The second group consisted primarily of compound adjectives ending in -gij, -kij, -xij. Analogous simplex adjectives have synthetic
comparatives with alternations, while these adjectives have only analytic standard forms. If
speakers nevertheless try to form synthetic comparatives, they lack alternations significantly
more often than in the first group. The most important factor is whether the second part of the
compound is used as an independent adjective. If it is not, as in the case of dlinnorukij ‘longarmed’ or dlinnonogij ‘long-legged', the majority of comparatives lack alternations. This is
notable because the relevant stems with alternations can be found in many highly frequent
words, e.g. ručka ‘small hand, handle’, nožka ‘small leg, furniture leg’. So it seems to be
crucial whether a particular form is listed in the mental lexicon, not whether the model is
available. This is similar to our results with the verbs: in the I class, the model is productive,
but it plays the most important role whether a standard form with alternations from a
particular verb is stored in the lexicon. The tendencies observed for comparatives were
replicated in a psycholinguistic experiment where participants (25 speakers of Russian, age
18-41) were asked to produce comparatives from various real and nonce adjectives
(embedded in short sentences they were asked to complete).
Many competing approaches to paradigm leveling exist. But we are cautious to interpret our
results in favor of any theory. For example, the fact that underapplication of alternations is
preferred to overapplication is problematic for McCarthy’s (2005) framework, being more
readily compatible with accounts like (Albright 2002, 2010). However, these and other
theories were primarily designed to work with different data, explaining why some groups of
words developed particular established forms rather than predicting different frequencies of
various non-standard innovations. We can gain access to such data only now, with the
development of Internet communication. We strongly believe that the general principles
underlying these data should be the same in both cases, but some adaptation is still needed.
To give another example, alternations that are unattested in standard Russian can be taken as
evidence that speakers, at least in some cases, rely on conditions on outputs (form X should
contain consonant A) rather than on input-output relations (stem-final consonant B becomes
A in form X) (e.g. Bybee 1995). However, this does not readily predict the vast diversity of
such alternations, as well as their very low frequency compared not only to “standard”
alternations, but also to the cases where alternations are missing.
Albright, A. 2002. The identification of bases in morphological paradigms. Doctoral dissertation, University of
California. • Albright, A. 2010. Base-driven leveling in Yiddish verb paradigms. Natural Language and
Linguistic Theory 28: 475–537. • Bybee, J. 1995. Regular morphology and the lexicon. Language and Cognitive
Processes 10: 425–455. • McCarthy, J. 2005. Optimal paradigms. In Paradigms in Phonological Theory, ed.
L.J. Downing, T.A. Hall & R. Raffelsiefen, 170–210. Oxford: Oxford University Press. • Slioussar, N., &
Kholodilova, M. 2013. Paradigm leveling in non-standard Russian. In Formal Approaches to Slavic Linguistics:
The Second MIT Meeting 2011, ed. A. Podobryaev, 243–258. Ann Arbor, MI: Michigan Slavic Publications.
Agreement attraction: a novel view from Russian
HSE Moscow & St. Petersburg State University
St. Petersburg State University
[email protected]
[email protected]
We discuss two experiments on number agreement attraction (AA) and on a potentially
similar phenomenon in Russian and their implications for psycholinguistic approaches to
agreement attraction and for the views on the grammar architecture.
1. Background. Attraction errors received a lot of attention in the last two decades (e.g. Bock
& Miller 1991; Eberhard et al. 2005; Franck et al. 2002, 2006; Vigliocco et al. 1995). An
example of an attraction error in subject-predicate agreement is given in (1). The verb are
agrees not with the head of the subject DP key, but with a dependent DP cabinets (attractor).
Corpus studies and production and comprehension experiments looked at such errors in various languages, and, among many things, noted that: (ia). Only Plural DPs cause significant
attraction: errors like (1) are produced more often and cause smaller delay in comprehension
than errors like (2). (ib). The formal resemblance factor. Errors appear more often in
production and cause smaller delay in comprehension if the form of the attractor coincides
with the Nom.Pl form, like in German (3a) as opposed to (3b) (Hartsuiker et al. 2003).
The key to the cabinets are rusty.
The keys to the cabinet is rusty.
(3) a. die Stellungnahme gegen die Demonstrationen ‘theNOM.SG position against
theACC.PL=NOM.PL demonstrations’
b. die Stellungnahme zu den Demonstrationen ‘theNOM.SG position on theDAT.PL≠NOM.PL
Two types of accounts have been proposed: (iia). Subject DPs get their number specification
from a dependent DP due to erroneous feature percolation or similar mechanisms, i.e. the syntactic structure is represented incorrectly (e.g. Franck et al. 2002; Eberhard et al. 2005). So we
produce an error or miss it in comprehension. (iib). The error arises not when we build syntactic structure, but when we access information in it: when we encounter a wrong verb form
in comprehension or generate it in production, we recheck the subject DP and an attractor can
interfere with this process (e.g. Solomon & Pearlmutter 2004; Wagers et al. 2009).
2. Our study. Exp. 1. In Russian, Nom.Pl forms of some nouns coincide not only with Acc.Pl,
but also with Gen.Sg forms: e.g. vecherinki from vecherinka ‘party’. We conducted an experiment to compare the number of errors with subject DPs like (4a-c). In every trial, participants
saw a predicate and then a subject and were asked to pronounce a complete sentence. Half of
the predicates did not agree with the subjects in number (a modified version of the method
originally proposed by (Vigliocco et al. 1995)). In target stimuli, subjects consisted of a head
noun, a preposition selecting Acc or Gen and a dependent noun in all possible number
combinations (inanimate nouns for which the relevant forms coincide were used).
(4) a. bilet na koncerty ‘ticketNOM.SG for concertACC.PL=NOM.PL’
b. komnata dlja vecherinki ‘roomNOM.SG for partyGEN.SG=NOM.PL’
c. komnata dlja vecherinok ‘roomNOM.SG for partyGEN.PL≠NOM.PL’
Results. Out of all possible number combinations in subject DPs of our target stimuli (four in
Acc sets and four in Gen sets), number agreement errors occurred only in three conditions
containing DPs like (4a-c) and were distributed as follows: (4a)>(4b)>(4c). I.e. there were
significantly more errors with Gen.Sg=Nom.Pl attractors than with Gen.Pl≠Nom.Pl ones.
Exp. 2. Exp.1 is compared with Exp.2 looking at another example of morphological
ambiguity. In Russian, some adjective forms are ambiguous between different cases, in
particular, Gen.Pl and Prep.Pl coincide (Prep stands for Prepositional case). Rusakova (2009)
who studied naturally occurring errors in Russian noted several examples like (5). We studied
these errors in a comprehension experiment using self-paced reading methodology (unlike
with subject-predicate agreement errors, there is no obvious way to induce them in
production). We had 36 sets of target sentences: 12 sets with prepositions selecting Gen, Prep
and Dat. Gen and Prep sets contained sentences in the three conditions shown below (Dat sets
were used as controls and contained only the errors analogous to (6c), i.e. had two WICs):
v tex razmerov ‘in thosePREP.PL(=GEN.PL) sizeGEN.PL
(6) a. Neudachi v proshlyh sezonah zastavili komandu potrudit’sja. ‘Correct’ condition, CC
‘failureNOM.PL in previousPREP.PL seasonPREP.PL makePST.PL teamACC.SG workINF’
b. Neudachi v proshlyh sezonov… ‘Wrong compatible’ condition, WCC
‘failureNOM.PL in previousPREP.PL(=GEN.PL) seasonPREP.PL’
c. Neudachi v proshlyh sezonam… ‘Wrong incompatible’ condition, WIC
‘failureNOM.PL in previousPREP.PL(≠DAT.PL) seasonDAT.PL’
Results. The violation was detected significantly later and its effect was less pronounced in
WCC conditions than in WIC conditions.
3. Conclusions. Previously, the formal resemblance factor in (ib) was assumed to boost the
Plural markedness effect in (ia). Our results from Exp.1 indicate that (ib) is more important
than (ia), AA effects are driven more by the form overlap than by featural overlap. Results
from Exp.2 provide an interesting parallel (such constructions have not been experimentally
studied before). We argue that in this case we deal with a phenomenon analogous to AA. The
implications are manifold:
The arguments used in the discussion of two approaches to AA (iia-b) have been
inconclusive so far. Our results are incompatible with (iia) and thus provide a strong argument
in favor of (iib). In cases like (4b), there are no Pl features that could percolate, yet AA errors
are attested and are even more frequent than in cases like (4c). In (5) or (6b), there is also no
feature that could percolate from the adjective to the noun. Approaches in (iib) need to be
modified to account for our data. Wagers et al. (2009) assume that when we have a wrong
verb form and rechecking is prompted, an attractor can provoke a mis-take if it contains a Pl
feature. Our data show that the surface form of the attractor is important.
Our results are hardly compatible with grammar architectures where full-fledged morphological forms appear only at late stages (Late Insertion). All previous studies of number
AA discussed only cases like (1), (3a-b), (4a) or (4c). Both types of approaches above assumed that errors are triggered by the featural specifications of the attractor. Our results show
that the actual form of the attractor is also (and even more) important. This form should be
available and visible at early stages of syntactic derivation to be able to influence agreement.
BOCK, J. K., & MILLER, C. A. (1991). Broken agreement. Cognitive Psychology, 23, 45–93. • EBERHARD, K. M.,
CUTTING, J. C., & BOCK, J. K. (2005). Making syntax of sense: Number agreement in sentence production.
Psychological Review, 112, 531-559. • FRANCK, J., LASSI, G., FRAUENFELDER, U., & RIZZI, L. (2006).
Agreement and movement: A syntactic analysis of attraction. Cognition, 101, 173-216. • FRANCK, J., VIGLIOCCO,
G., & NICOL, J. (2002). Attraction in sentence production: The role of syntactic structure. Language and
Cognitive Processes, 17, 371-404. • HARTSUIKER, R., SCHRIEFERS, H., BOCK, K., & KIKSTRA, G. (2003).
Morphophonological influences on the construction of subject–verb agreement. Memory and Cognition, 31,
1316–1326. • RUSAKOVA, M. (2009). Rechevaja realizacija grammaticheskix elementov russkogo jazyka
[Speech realization of some grammatical features of Russian]. Habilitation dissertation, St.Petersburg State
University. • SOLOMON, E. S., & PEARLMUTTER, N. J. (2004). Semantic integration and syntactic planning in
language production. Cognitive Psychology, 49, 1-46. • 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. • WAGERS, M.W., E.F. LAU & C. PHILLIPS (2009). Agreement
attraction in comprehension: Representations and processes. Journal of Memory and Language 61: 206-223.
Turkish synthetic compounds and the categorial status of the Root
Yeditepe University
[email protected]
Within the recent constructionalist approaches to language, particularly in the grammar
architecture assumed in Distributed Morphology (Halle and Marantz 1993), syntax not only
shoulders phrase/sentence formation, but also word formation, including both simplex and
complex words. In these approaches, Roots, which are acategorial, nongrammatical
morphemes with semantic content, are manipulated by functional categories to derive words,
which receive phonological expression only after all syntactic and postsyntactic operations are
completed. For instance, the Root √DESTR, when selected by a little n, creates the context for
the insertion of the Vocabulary Item destruction whereas in the context of a little v, it is
realized as destroy.
There are two main views about the categorial status of Roots. According to the first one,
Roots are considered syntactic categories that project phrases and specifiers and take
complements. In this view, the complement position of the √P is where the internal argument
can be hosted (Marantz 1997; Harley 2009, 2011). According to the second view, Roots are
non-projecting categories, such that they do not project √Ps or specifiers, take complements,
or impose any position for arguments (Borer 2009, 2011; de Belder 2011; de Belder and
Craenenbroeck 2011). In this view, a verbal projection, such as a vP is needed to host the
internal argument. In this study, I consider a range of seemingly synthetic compounds in
Turkish, and show that nominals that are derived directly from Roots can never form true
synthetic compounds insofar as our definition of these is such that the first element in the
compound is an argument of the second. This shows that Roots cannot take internal
arguments as complements, supporting the view that they are non-projecting syntactic
A compound that appears to be synthetic in nature at first glance can be parsed at least in two
ways. For instance truck-driver in English can be parsed as a true synthetic compound as in
(1), in which the first element initially incorporates into the verbal stem before the whole
construction gets nominalized by -er, or as a primary compound as in (2) analogous to, let’s
say, truck-garage where the semantic relation between the first and the second element is
determined by world knowledge rather than strict syntactic means.
[truck [drive]-er]
[truck [driver]]
In Turkish, the difference between synthetic and primary compounds is less subtle.
This is because the primary compounds would always necessitate the presence of a peculiar
compound marker -(s)İ(n), typically analyzed as agreement morphology. For instance, the
Turkish equivalent of truck-driver and truck-garage would both have this marker, suggesting
that they are both primary compounds.
‘truck driver’
‘truck garage’
True synthetic compounds in Turkish never take the compound marker -(s)İ(n). Furthermore,
these allow by-phrases, agent-oriented modifiers, adverbial and aspectual modification, which
all contrast with primary compounds, showing the presence of an event structure in the former
and the lack of it in the latter. These contrasts are illustrated in examples (5-6).
Primary compound:
kadın çiz-im-i
woman √draw-NOM-SİN
Intended: ‘drawing of a woman by the student’
kadın çiz-im-i
woman √draw-NOM-SİN
Intended: ‘drawing of a woman voluntarily’
(*bir saat-te) (*nihayet)
kadın çiz-im-i
one hour-LOC eventually
woman √draw-NOM-SİN
Intended: ‘eventually drawing of a woman in one hour’
True synthetic compound:
öğrenci-ler tarafından
information √share-VER-NOM
‘sharing of information by the students’
information √share-VER-NOM
‘sharing of information voluntarily’
two week-EP
hardship-COM information √share-VER-NOM
‘sharing of information with difficulty for two weeks’
In this paper, I conclude that (i) Roots are non-projecting categories, (ii) a verbal stem is
required to host a theme argument, (iii) and, therefore, only nomimalizations with verbal
stems can act as the head of true synthetic compounds. I further show that the presence of an
overt verbalizing or causativizing morpheme in certain nominalizations does not always entail
an event structure; nominals of this type always denote entitites. I refer to compounds headed
by these as ‘pseduo-synthetic compounds’ which I claim are just a sub-set of primary
BORER, H. 2009. Roots and categories. Paper presented at the 19th Colloquium on Generative Grammar.
University of the Basque Country. April 1-3, 2009.
BORER, H. 2011. In the event of a nominal. In Everaert, M., Maelj, M. and T. Siloni (eds). The theta system:
argument structure at the interface. Oxford University Press.
DE BELDER, M. 2011. Roots and affixes: eliminating lexical categories from syntax. Brussels: HU Brussel /
Utrecht University.
DE BELDER, M. and VAN CRAENENBROECK, J. 2011. How to merge a root. M.s. HU Brussel & Utrecht
HALLE, M. and MARANTZ, A. 1993. Distributed morphology and the pieces of inflection. In A view from
Building 20: Essays in linguistics in honor of Sylvain Bromberger, ed. K.
HALE, K. and S. J. KEYSER, 111-176. Cambridge, MA: MIT Press.
HARLEY, H. 2009. Compounding in Distributed Morphology. In The Oxford handbook of compounding, R.
Lieber and P. Stekauer (eds). Oxford: OUP. p. 129-143.
HARLEY, H. 2011. On the identity of Roots. M.s. University of Arizona.
MARANTZ, A. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own
lexicon. UPenn Working Papers in Linguistics 4: 201-225.
Verbal morphemes need not entail verbal structure
Yeditepe University
University of Arizona
[email protected]
[email protected]
Derivation from category-neutral Roots has been observed to differ in important respects from
derivation from syntactic categories (Marantz 1997, Arad 2003). In terms of semantics, the
merger of the first functional head with an acategorial Root is a domain of special meanings.
A verbalizing head that merges directly with a Root may have an idiosyncratic meaning;
further derivation from this syntactic object cannot result in a new, renegotiated meaning, but
must instead be an extension of the semantics produced at the first merger with the Root.
Marantz (2013) addresses apparently counter-examples to this generalization. Among these
are cases in Japanese where a nominalizer interacts with a Root to produce idiosyncratic
semantics despite being separated from the Root by a causative or inchoative verbalizing
√nag nag-e-(ru) ‘to flow’ nas-as-(u) ‘to make flow’
(from Volpe (2005), cited in Marantz (2013))
nag-ashi ‘a sink’
Marantz proposes that the verbalizing morpheme in the nominalization is semantically null
insofar as it does not entail an event. This is what allows the nominalizing head to negotiate a
special meaning with the Root: the intervening morpheme is not semantically visible, such
that the nominalizer is semantically adjacent to the Root. Marantz’s proposal suggests that the
intervening morpheme plays no role in deriving the idiosyncratic semantics. However, a
number of Turkish derivations call this into question. There are several transitive verbs in
Turkish that optionally have single or double morphological exponence of the causative with
no consistent difference in meaning.
Single Exponence
çık-ar- ‘bring up, out’
kop-ar- ‘detach (tr.)’
kap-a- ‘close (tr.)’
uç-ur- ‘make fly’
Double exponence
çık-ar-t- ‘bring up, out’
kop-ar-t- ‘detach’
kap-a-t- ‘close (tr.)’
uç-ur-t- ‘make fly’
Nominalizations based on these forms have special meanings, much like the Japanese
examples. What is striking is that these semantics vary sharply and idiosyncratically
according to whether the causative verbalizer has single or double exponence.
Single Exponence
a. çık-ar-ma ‘naval landing’
b. kop-ar-ma ‘clean and jerk’
c. kap-a-ma ‘lamb stew’
d. uç-ur-ma (no special meaning)
Double Exponence
çık-ar-t-ma ‘sticker, decal’
kop-ar-t-ma ‘(a kind of) pastry’
kap-a-t-ma ‘kept woman’
uç-ur-t-ma ‘kite (the toy)’
In the face of these facts, it cannot be maintained that the morphemes intervening between the
Root and the nominalizer are invisible to the semantics. We propose instead that the
nominalizations are not derived from the corresponding verbal structures. Building on
Marantz’s insight that these nominalizations do not entail an event, we claim that the nouns in
(3) do not have an argument structure, unlike the corresponding verbal forms in (2). The
meanings of these nouns, like those of their Japanese counterparts, indicate that they are Rnominals (entitydenoting nominals) in the sense of Borer (2003, 2011), and thus do not allow
agent-oriented modifiers, adverbial or aspectual modification, and cannot participate in the
formation of true synthetic compounds. We therefore conclude that the different
interpretations of nominalizations such as the ones in (3) do not occur at LF because their
nominalizing head is semantically Root-adjacent but rather occur at the Encyclopedia
component of the grammar architecture generally assumed in the Distributed Morphology
framework (Harley and Noyer 1999), which contains the list of idioms in a language that need
to be memorized by the language learner.
ARAD, M., 2003. Locality constraints on the interpretation of roots: The case of Hebrew denominal verbs.
Natural Language and Linguistic Theory 22: 737-778.
BORER, H. 2003. Exo-skeletal vs. Endo-skeletal explanations. In Moore, J. and M. Polinsky (eds). The nature of
explanation in linguistic theory, Chapter 2. CSLI.
BORER, H. 2011. In the event of a nominal. In Everaert, M., Maelj, M. and T. Siloni (eds). The theta system:
argument structure at the interface. Oxford University Press.
HARLEY, H. and R. NOYER. 1999. State-of-the-article: Distributed Morphology. GLOT 4.4: 3-9.
MARANTZ, A. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own
lexicon. UPenn Working Papers in Linguistics 4: 201-225.
MARANTZ, A. 2013. Locality domains for contextual allomorphy across the interfaces. In Matuhshansky, O. and
A. Marantz (eds). Distributed Morphology today. MIT Press.
Two types of i-defectivity in Spanish verbal paradigms
CLLE-ERSS UMR 5263, CNRS & Université Toulouse Jean Jaurès
[email protected]
A curiosity of the Spanish -ir conjugational class is the existence of verbs like abolir
“abolish” that are said to have a defective paradigm. According to grammatical tradition, the
verbs in this class can only be conjugated when the ending begins with the letter ‹i› (Bello
§587ff), hence the name “i-defective”. The other forms — pres. ind. 1/2/3/6, pres. subj., and
sing. imperative — are said to be missing. So the pertinent parts of the paradigm of abolir
have the following shape, with the defective cells highlighted:
past part.
According to some sources, however, abolir is no longer an i-defective verb and now has a
complete conjugation, with fully regular forms as indicated above. Normative authorities in
fact disagree about the precise membership of the class of i-defective verbs, and limited
empirical studies (notably Albright 2003) suggest that i-defectivity is a gradient and variable
phenomenon. The verbs identified are always low frequency if not completely archaic, but if
pressed, speakers will produce and accept supposedly defective forms, albeit with a low
degree of confidence. The psychological reality and robustness of i-defectivity are therefore
less than fully established, and the distribution shown above may be best understood as an
idealization, possibly corresponding more to the situation in earlier stages of modern Spanish
than to that of current standard varieties.
The proper characterization of the i-defective pattern also remains open to debate. It is clear
that the semantic features of the missing forms provide no satisfactory explanation. Albright
(2003) proposes an “anti-stress” constraint, meaning that root-stressed forms are avoided.
This accounts for almost all of the paradigm except for persons 4/5 of the present subjunctive
(underlined in the paradigm above), which have an unstressed root but are nevertheless
considered to be defective. Maiden & O’Neill (2010) see in the i-defective distribution a
combination of the morphomic L- and N-patterns that are recurrent in Romance verbal
morphology, primarily in connection with stem allomorphy. There are indeed verbs whose
stems exhibit precisely this distribution in the present tenses (e.g. ped- vs. pid- for pedir), but
note that the parallelism with abolir does not extend to the complete paradigm:
past part.
In the “stem space” approach of Boyé & Cabredo-Hofherr (2006, 2010), Spanish verbs have
11 interrelated stems, each associated with a region of the paradigm. The defective cells in the
conjugation of abolir correspond exactly to the regions for stems “S1”, “S2”, and “S5”, so the
defective paradigm can be modeled by stipulating that those three stems are suppletively
specified as Ø. But none of the previous analyses captures the traditional idea that idefectivity is correlated with the segmental content of the verb ending: the endings of the
existing forms all begin with [i] or [jV]. It is hard to see how the absence of such an ending
could be a trigger for defectivity, but no other generalization appears to give the desired
result, and we conclude that the analysis of the abolir-type conjugation must make direct
reference to the surface phonological form of the ending (or to an abstract thematic vowel /i/).
Bello (§591f) further describes a variant of i-defectivity that was already obsolete by the mid19th century. Verbs like escarnir “mock” could only be conjugated with endings starting
with [i], so that in addition to the abolir-type defectivity, forms in [jV] were also excluded:
*escarnió, *escarnieron, *escarniendo. These missing forms were supplied by the doublet
verb escarnecer; these are indicated in green in the following paradigm.
past part.
imp.subj. escarneciera
It is not widely recognized that modern Spanish once had heteroclitic conjugations involving
the inchoative suffix (cf. Italian -isc-, French -iss-), and what is most striking is that the
distribution of stems was exactly the same as in the case of pedir shown above. The [e]/[i]
alternation for verbs like pedir is relatively well studied, with analyses typically invoking
phonological notions (dissimilatory lowering of [i] before í, metaphonic raising of [e] before
[j]). We propose that the pedir-type distribution of stems became a morphologized
paradigmatic pattern that gave rise to the obsolete escarnir/escarnecer-type mixed
conjugation, for which no phonological motivation can be identified.
Selected references
ALBRIGHT, A. 2003. “A Quantitative Study of Spanish Paradigm Gaps”. In G. Garding & M. Tsujimura (eds.),
WCCFL 22 Proceedings. Somerville, MA: Cascadilla Press, pp. 1–14. • BELLO, A. 1847. Gramática de la
Lengua Castellana destinada al uso de los Americanos. Santiago: El Progreso. • BOYÉ, G. & P. CABREDO
HOFHERR. 2006. “The structure of allomorphy in Spanish verbal inflection”. Cuadernos de Lingüística Ortega y
Gasset 13:9–24. • BOYÉ, G. & P. CABREDO HOFHERR. 2010. “Defectiveness as Stem Suppletion in French and
Spanish Verbs”. In M. Baerman et al. (eds) Defective Paradigms. London: British Academy, pp. 35–52. •
MAIDEN, M. & P. O’NEILL. 2010. “On morphomic defectiveness: evidence from the Romance languages of the
Iberian peninsula”. In M. Baerman et al. (eds.) Defective Paradigms. London: British Academy, pp. 103–124.
The transcategorial irrealis suffix in some Mayan languages
The Institute of Linguistics, Russian Academy of Sciences
[email protected]
Though the term “irrealis” (or “irreal status”) is widely used in linguistic studies and
grammatical descriptions, there are some doubts about its typological validity (cf. Bybee et al.
1994, Bybee 1998, and the recent study de Haan 2012). In essence, these doubts arise because
of the vagueness and uncertainty of possible definitions of irreality. As a marker of the set of
unreal events, irrealis is usually considered a part of verbal grammar, and consequently
irreality is regarded as a verbal semantic domain. As I argue below, even such a
circumscription of the applicable domain of irreality is incorrect for some languages. The
study is mainly based on the analysis of original texts.
The modern Mayan languages of the Cholan-Tzeltalan and Q’anjobalan branches have a
suffix of irrealis that is notable for its transcategorial nature. It can combine not only with
verbs, but also with a wide range of other parts of speech. Etymologically, this suffix
originates in the Protomayan modal suffix denoting dependent status for intransitive verbs
*-oq (Robertson 1992).
There is a range of possible meanings of this suffix and these meanings are highly
dependent on the part of speech of the stem. Most frequently it is added to finite verb forms.
In this case it has the widest realm of uses that consists of counterfactual conditions,
expressions of goal or intention and various desiderative/optative constructions (1). Generally,
the list of verbal meanings of the Mayan irrealis does not significantly differ from the
typological standard obtained in various comparative studies (Elliott 2000, for instance).
When the irrealis suffix is used with another part of speech it reduces the range of possible
meanings. Furthermore the semantics of this suffix depend on the syntactic role of the word
form, specifically on its predicative status. Non-verbal predicates are formed without any
auxiliary component in Mayan languages, so in some cases the irrealis suffix added to a nonverbal stem can be regarded as syntactic marker of predicativity. With non-verbal predicates,
the realm of uses of the irrealis suffix is limited to negative (2) or, less often, interrogative
When the nominal, adverbial or numeral stem that does not act as a predicate carries the
irrealis suffix, it usually denotes approximateness or indefiniteness (3). The combination of
the numeral ‘one’ and the irrealis suffix (4) is to a certain extent similar to the indefinite
article in European languages, an element that is absent in Mayan languages of the CholanTzeltalan and Q’anjobalan branches.
The irrealis suffix can sometimes perform the word-formative function as well. Indefinite
pronouns are derived from interrogative ones by means of this suffix in some Mayan
There are some differences in the use of irrealis suffix in a particular language. For
example, the usage of irrealis in order to form a question was noted with certainty only in
Tzeltal, while the derivation of an indefinite pronoun by means of the irrealis suffix is
possible only in Tzotzil and Ch’orti’. In general, however, the two branches of the Mayan
family under consideration exhibit significant similarities.
The transcategorial nature of the Mayan irrealis suffix, as a special morphological feature
of a particular morpheme, reveals at least two significant semantic features of the category of
irrealis. Firstly, the Mayan irrealis provides clear evidence that irreality does not correspond
with modality, being a wider semantic domain. It exceeds the limits of verbal grammatical
semantics, spreading to nominal and numeral semantic domains, such as, for instance,
approximateness and indefiniteness. Secondly, as regard to the possible definitions of
irreality, the Mayan languages from the Cholan-Tzeltalan and Q’anjobalan branches prove the
importance of referentiality as the determining factor when considering an event or an object
to be real or unreal. The non-referential status of an object can trigger the marker of irrealis, at
least in some languages. This idea was first suggested in Givón (1994), but is as yet still not
generally accepted as irreality is considered an exclusively verbal category.
(1) Chol (Cholan-Tzeltalan)
che’_jini la’-to
‘Well, let him still grow’ (Warkentin & Whittaker 1965: 16)
(2) Akatek (Q’anjobalan)
man y-et-oj
NEG 3.POSS-ownership-IRR CLF(metal)
‘The money is not for him’ (Andrés & Dakin 1989: 294)
(3) Tzotzil (Cholan-Tzeltalan)
mi mas
3.ERG.PFV-waste maybe
‘He wasted maybe a week or more’ (Laughlin 1977: 55)
(4) Tzeltal (Cholan-Tzeltalan)
s-jok’oy te
ya s-nuk’
3.POSS-cigarette 3.POSS-cause
IPFV 3.ERG-smoke
‘He asked if he had taken a cigarette to smoke’ (Pérez López et al. 1994: 282)
ANDRÉS, J. J. & K. DAKIN. 1989. El wakax kan y otros textos acatecos. Tlalocan XI: 285-300.
BYBEE, J. 1998. “Irrealis” as a grammatical category. Anthropological linguistics 40/2: 257-271.
BYBEE, J., R. PERKINS & W. PAGLIUCA. 1994. The evolution of grammar: tense, aspect, and modality in the
languages of the world. Univ. of Chicago Press.
DE HAAN, F. 2012. Irrealis: fact or fiction? Language sciences 34: 107-130.
ELLIOTT, J. 2000. Realis and irrealis: forms and concepts of the grammaticalisation of reality. Linguistic
typology 4/1: 55-90.
GIVÓN, T. 1994. Irrealis and the subjunctive. Studies in language 18/2: 265-337.
LAUGHLIN, R. 1977. Of cabbages and kings: tales from Zinacantán. Washington DC, Smithsonian Institution
PÉREZ LÓPEZ, E., M. HIDALGO PÉREZ & A. GÓMEZ GÓMEZ (eds.). 1994. Cuentos y relatos indígenas, vol. V.
México DF, UNAM.
ROBERTSON, J. 1992. The history of tense/aspect/mood/voice in the Mayan verbal complex. Austin, UT Press.
WHITTAKER, A. & V. WARKENTIN. 1965. Chol texts on the supernatural. Oklahoma, SIL.
Defining direct/inverse systems: a canonical approach
1. DDL (CNRS, U. Lyon 2) 2. LLF (CNRS, U. Paris 7) 3. CRLAO (CNRS, INALCO)
[email protected] [email protected] [email protected]
The concept of CANONICAL INFLECTION (Corbett, 2007) has been used to characterise a broad
variety of inflectional systems (Thornton, 2010; Baerman et al., 2010; Stump, 2012; Brown
and Hippisley, 2012; Palancar, 2012). However, none of the existing work on canonical inflection seems to consider the specificity of DIRECT / INVERSE SYSTEMS (Zúñiga, 2006) and their
relationship to the criteria defining canonical inflection.
This paper constitutes a first attempt towards the proposition of a set of criteria for defining the
concept of direct/inverse systems. Our definition is set within the CANONICAL TYPOLOGICAL
FRAMEWORK (Corbett, 2003) and focuses more specifically on the canonicity of direct/inverse
marking. We also show how regular direct/inverse systems compare with the more general standard of CANONICAL INFLECTION as defined in (Corbett, 2007).
Direct/inverse systems are traditionally thought of as systems displaying hierarchical agreement. Transitive verbs within direct/inverse systems typically index two arguments and are
marked so that, e.g., a 1 SG agent and a 3 SG patient (henceforth 1 SG >3 SG) are marked in the
same way as a 3 SG agent and a 1 SG patient (3 SG >1 SG), except for a distinctive INVERSE
MARKER . Direct vs. inverse forms are then distinguished through their coherence with an independent agentivity/empathy hierarchy, which typically stipulates that arguments corresponding
to speech act participants (SAP) are higher in the hierarchy than third person animate arguments,
which are again higher than inanimate arguments. A typical hierarchy is one by (DeLancey,
1981, 644): SAP > 3rd pers. pronoun > human > animate > natural forces > inanimate.
In our example, 1 SG >3 SG complies with the hierarchy and would be considered a direct form,
marked by a dedicated DIRECT MARKER (DM), while 3 SG >1 SG does not comply with the hierarchy and would be typically marked as inverse by the contrasting INVERSE MARKER (IM).
Moreover, in traditional typological approaches, the hierarchy among SAP is considered to be
unstable across languages (Jacques and Antonov, 2013): 1>2 and 2>1 can be marked with special forms (Algonquian); 2>1 may receive the IM (Situ Rgyalrong); both 2>1 and 1>2 may
receive the IM (Mapuche or Kiranti); the presence of DM or IM independently follows from
the context (Japhug Rgyalrong). This variation is often seen as the image of the languages’
specific hierarchy and typological properties.
Within this broad range of variation, it makes sense to resort to a canonical approach. It allows
for defining an independent, purely formal (and typically non-existing) standard from which
deviation may be measured. The standard we define for canonical direct/inverse (DI-canon)
systems is based on three converging criteria: (i) compliance with a language independent,
maximally regular hierarchy (any verbal argument should be higher or lower w.r.t. the other),
as sketched hereafter: 1>2>3 HIGHEST>. . . >3 LOWEST, (ii) perfect symmetry across the cells
of the inflectional paradigm involving the same verbal arguments (see Table 1), (iii) autonomy
with respect to linguistic features external to the hierarchy.
2 2>1
3 H 3 H>1
3 L 3 L>1
1>3 H
2>3 H
3 H>2
3 L>2 3 L>3 H
1>3 L
2>3 L
3 H>3 L
Table 1. Schematic symmetrical structure of a canonical direct-inverse system (dark cells mark inverse, white
cells mark direct; 3 H represents a 3rd person argument higher in the hierarchy, and 3 L a 3rd person lower
We illustrate possible types of deviation from the direct/inverse canon based on data from Algonquian, Kiranti, Mapuche, Rgyalrong, and Sahaptian. For example, the Kiranti data in the
Khaling sub-paradigm in Table 2 shows deviation from the DI-canon in two of above mentioned
dimensions: (i) inverse marking does not strictly follow the directness hierarchy: third person
forms are all marked as direct forms; (ii) the IM Pi- is not symmetrically distributed across the
paradigm: (a) the empathy hierarchy does not strictly structure 3rd person arguments; (b) the
IM also appears in supposedly direct forms such as 2>3, 1DE>2 and 1PE>2, and (c) 1>3 and
3>1, resp. 1>2 and 2>1, are not symmetrical. The Japhug data, where inverse marking is reliant
on genericity, will illustrate deviation from the DI-canon with respect to criterion (iii), namely
the autonomy of the hierarchical marking w.r.t. other linguistic features.
1 DI
1 DE
1 PI
1 PE
1 DI
1 DE
1 PI
1 PE
Table 2. Non-canonical direct-inverse marking in Khaling (Kiranti), data from (Jacques et al., 2012)
Finally, we characterise the canonical direct/inverse system w.r.t. the notion of canonical inflection (CI). We show that the DI-canon only partly complies with the criteria for CI. CI can be seen
as an instance of maximal regularity and minumum ambiguity in the paradigmatic realisation of
individual morphosyntactic property sets (henceforth MPS). CI also features COMPLETENESS
(each MPS has a corresponding form realisation) and UNIQUENESS (each MPS is realised by
one form only). In the maximally regular case, the paradigm of a direct/inverse system (i.e.,
the DI-canon) seems to fully satisfy the set of criteria drawn for defining CI. However, when
one considers the individual feature realisation, hierarchical direct/inverse marking does deviate from the canonical definition of inflection. Corbett (2007) indicates that CI expresses each
value of the overall set of morphosyntactic features through an unambiguous single affix. In
the case of direct/inverse based hierarchical agreement marking, affixes are typically ambiguous in the sense that they underspecify the status of the argument they are supposed to encode.
Only the DM and IM allow for disambiguation. Moreover, the structure of the paradigm is not
completely unmotivated (as expected for CI), but conditioned by a canonical empathy hierarchy. The DI-canon thus appears as a maximally regular and minimaly ambiguous system, also
complying with the completeness and uniqueness criteria, that deviates from the general canon
for inflection in the individual encoding of the argument verbal features and the (at least partial)
motivation of the paradigm’s symmetric shape through a specific feature hierarchy.
Acknowledgments This work was partly supported by the LabEx ASLAN (ANR-10-LABX0081) operated by the French National Research Agency (ANR).
M. BAERMAN , G. G. C ORBETT AND D. B ROWN, Eds. 2010. Defective Paradigms: Missing Forms and What They Tell Us.
Oxford University Press.
B ROWN , D. AND A. H IPPISLEY. 2012. Network Morphology: A Defaults-based Theory of Word Structure. Cambridge
University Press.
C ORBETT, G. G.. 2003. Agreement: the range of the phenomenon and the principles of the surrey database of agreement.
Transactions of the philological society, 101, 155–202.
C ORBETT, G. G.. 2007. Canonical typology, suppletion and possible words. Language, 83, 8–42.
D E L ANCEY, S.. 1981. An interpretation of split ergativity. Language, 57(3), 626–57.
JACQUES , G. AND A. A NTONOV. 2013. Direct-inverse systems. Language and Linguistics Compass. In press.
JACQUES , G., A. L AHAUSSOIS , B. M ICHAILOVSKY, AND D. B. R AI. 2012. An overview of Khaling verbal morphology.
Language and linguistics, 13.6, 1095–1170.
PALANCAR , E.. 2012. The conjugations classes of tilapa otomi: An approach from canonical typology. Linguistics, 50(4),
S TUMP, G. T.. 2012. The formal and functional architecture of inflectional morphology. In A. R ALLI , G. B OOIJ , S. S CALISE
AND A. K ARASIMOS, Eds., Morphology and the Architecture of Grammar: On-line Proceedings of the 8th Mediterranean
Morphology Meeting (MMM8), Cagliari, Italy, 14-17 Sept. 2011, p. 255–271.
T HORNTON , A. M.. 2010. Towards a typology of overabundance. Communication au colloque des Décembrettes 7.
Z ÚÑIGA , F.. 2006. Deixis and Alignment - Inverse systems in indigenous languages of the Americas. Amsterdam: Benjamins.
The Opacity-Compactness Tradeoff:
Morphomic Features for an Economical Account of Khaling Verbal Inflection
1. DDL (CNRS, U. Lyon Lumière) 2. LLF (CNRS, U. Paris Diderot)
3. CRLAO (CNRS, Inalco) 4. ALPAGE (Inria, U. Paris Diderot)
[email protected] [email protected] [email protected]
Morphomic features (Aronoff, 1994) have been argued to allow for more elegant and more economical
descriptions of inflectional systems (Bonami and Boyé, 2010). However, morphomic accounts are opaque
from the morphosyntactic point of view and thus require an additional mapping between the morphomic
features and the actual morphosyntactic property sets (MPS) expressed by the inflected forms (Stump, 2006)
— something which is not needed if the MPSs are realised directly. Yet, no measures have been proposed
so far to quantitatively assess the tradeoff between the lower complexity achieved by relying on opaque
morphomic features and the additional need for an explicit feature mapping.
In this paper, we show that the descriptive economy achieved by capturing the inherent morphomic structure
of inflectional systems can effectively be assessed through dedicated compactness measures.
We present two competing analyses of Khaling verbal inflection based on original data from Jacques et al.
(2012). These two analyses will be called descriptions A and B hereafter. Khaling is a highly inflectional
language of the Kiranti subbranch of the Sino-Tibetan language family, spoken by about 15,000 people
in Eastern Nepal. Verbs realise up to 572 verb forms, among which 300 realis forms. Due to numerous
syncretisms, these 300 realis forms reduce to only 100 distinct forms.
Among our two descriptions, DESCRIPTION A uses realisation rules based on standard MPS (such as
A 1 S . P 2 S ). DESCRIPTION B comprises rules realising morphomic features (such as B , C , D , etc. in Table 2) and and their corresponding feature mapping functions (such as B : 1 DU . INCL). We show that the
paradigm structure of Khaling verbs is best described using morphomic features of the type proposed for
Nepali by Bonami and Boyé (2010) together with another morphomic feature related to the direct/inverse
marking system based on the language’s empathy hierarchy, as commonly used for describing Algonquian
languages (Silverstein, 1976; Zúñiga, 2006).
We show that if one ignores the direct/inverse marking (i.e., the presence of the inverse prefix Pi-), the
resulting additional syncretisms decrease the number of distinct forms from above mentioned 100 to only
68. These additional syncretisms are illustrated by the coloring in Table 1 for the non-past positive subparadigm. Additional morphomic feature mappings, such as posing a unique feature F for all forms expressing a P =2 SG, then allows for further syncretic reduction. Table 2 shows how for the non-past positive
subparadigm alone, the number of 75 distinct MPSs can be collapsed into 22 morphomic feature bundles
such as single feature F or feature bundle X >1.2 SG. Among these 22 morphomic feature bundles, 12 are
complex features, but 10 are simple features, thus greatly simplifying the the expression of form realisation
rules within the grammar (l8̂:p-nu becomes simply the realisation of feature K instead of that of features
3 PL >3, 3 DU >3 DU , 3 DU >3 PL , 2 SG >3 PL plus the inverse marked 2 SG >3 PL).
↓A | P→
1 SG
2 SG
2 DU
2 PL
3 SG
3 DU
3 PL
1 SG
2. SG
2 DU
2 PL
3 SG
3 DU
3 PL
Table 1. Positive non-past paradigm for the Khaling verb LOP ‘catch’, represented with standard features
Our two descriptions have been implemented on 167 verbs, producing 50,100 verb forms. We have measured the descriptions in terms of descriptive economy following Sagot and Walther (2011). This measure
relies on the information theoretical concept of Minimum Description Length (MDL) (Rissanen, 1984).
The measure allows in particular for showing the amount of information within different parts of the description, such as regular morphonological rules at morph boundaries (“phono+morphono”), the types of
2 SG
2 DU
2 PL
3 DU
3 PL
X >1
1> X
Table 2. Positive non-past paradigm for the Khaling verb LOP ‘catch’, reorganised after having introduced morphomic feature mappings
E.g., B : 1 DU . INCL, F : P =2 SG, I : 2 SG >3 SG or 3 SG >3 SG, . . .
phonological “operations” applied by the realisation rules, the information comprised within the “feature”
specification and the feature mapping rules, and the general “structure” of the description, including the
complexity of the realisation rules or the specified inflection classes.
In our experiment, the measure shows that the length of standard description A (with realisation rules realising transparent fully specified MPSs) is 6.2 Kbits, while the length of morphomic description B (realising
above mentioned morphomic features in the rule sets coupled with separate inverse marking and morphomic
mapping functions) is 5.4 Kbits.
These results show that using morphomic features makes for a description that is by 12% more compact
than the non-morphomic description. In particular, the additional cost of feature mapping functions (here
formalised with a mechanism similar to rules of Paradigm Linkage in PFM (Stump, 2006)) does not lead
to a higher overall description length (due to the simplification of the realisation rules themselves). As
illustrated by Figure 1, the additional information needed to stipulate the mapping between the morphomic
features and the actual realised morphosyntactic property sets appears to be a more than reasonable tradeoff
compared to the reduced length of the general description’s structure.
Combined with supplementary morphomic features, direct/inverse marking as traditionally used as a distinctive feature within descriptive linguistics thus proves to be quantitatively relevant for describing some
morphological systems, as it can lead to formulating maximally compact descriptions.
Acknowledgments This work was performed within the framework of the LabEx EFL (ANR-10-LABX0083), Strand 6 “Language Resources”, operation LR4.11. It was partly supported by the LabEx ASLAN
(ANR-10-LABX-0081). Both LabEx’s are operated by the French National Research Agency (ANR).
(including realisation rules)
A: standard
B: morphomic
(including feature mapping
if relevant)
1 Kbit
2 Kbits
Fig. 1. Comparative description lenghts for two accounts of Khaling verbal inflection, breakdown on grammar components
A RONOFF , M.. 1994. Morphology by Itself: Stems and Inflectional Classes. Linguistic Inquiry Monographs. MIT Press.
B ONAMI , O LIVIER AND G ILLES B OYÉ. 2010. Opaque paradigms, transparent forms in nepali conjugation. Workshop On Theoretical
Morphology 5.
morphology. Language and linguistics, 13.6, 1095–1170.
R ISSANEN , J.. 1984. Universal coding, information, prediction, and estimation. IEEE Transactions on Information Theory, 30(4), 629–636.
S AGOT, B ENOÎT AND G ÉRALDINE WALTHER. 2011. Non-canonical inflection : data, formalisation and complexity measures. In C. M AHLOW
AND M. P IOTROWSKI , Eds., Systems and Frameworks in Computational Morphology, volume 100, p. 23–45, Zurich, Switzerland: Springer.
S ILVERSTEIN , M ICHAEL. 1976. Hierarchy of Features and Ergativity. In R. M. D IXON, Ed., Grammatical Categories in Australian Languages.
S TUMP, G REGORY T.. 2006. Heteroclisis and paradigm linkage. Language, 82, 279–322.
Z ÚÑIGA , F ERNANDO. 2006. Deixis and Alignment - Inverse systems in indigenous languages of the Americas. Amsterdam: Benjamins.
The nominal morphology of Lovari from an analogical perspective
Doctoral School in Linguistics, Theoretical Linguistics Programme, ELTE
[email protected]
The available information regarding the nominal inflectional paradigms of the Lovari dialect
of Romani spoken in Hungary is contradictory, especially concerning the oblique stem. The
data provided by descriptive grammars and textbooks do not always and fully coincide with
the information provided by native informants. The relevant literature generally approaches
the phenomena from a rule-based aspect, synchronic or diachronic, mentioning analogical
effects only as a “last resort”. (Analogy at work in Lovari is very conspicuous; the analogical
processes and effects in the Lovari verbal paradigms are discussed by Baló 2012 for instance.)
This paper attempts to provide evidence that analogy-based approaches make language
change, as well as unstable and variegated forms and inflectional classes easier to understand
and grasp than they would be in a traditional synchronic framework or through a diachronic
analysis when the processes and the data are otherwise inexplicable.
Romani is a dialectally most diverse Indo-European language that is often exposed to contactrelated influences (cf. Matras 1995). One important result of an intensive contact period of
Early Romani with Greek is a cross-linguistically unique feature (Elšík 2000), the strict split
between the morphology of inherited and borrowed vocabulary (also termed as the thematicathematic dichotomy in Romani linguistics; for further information cf. Matras 2002), which is
seen in a new light if we examine the possible processes behind the apparent erosion of this
age-old system.
The difference between the inherited and borrowed layers of the lexicon manifests itself in the
oblique stem. What we can see on a synchronic level is that the thematic oblique ending for
the masculine is -es-/-en- (singular/plural), as opposed to athematic -os-/-on-. Elšík 2000
considers this from a diachronic perspective and claims that the reconstructed Proto-Romani
ending of all oblique stems is -en-/-es-. In Early Romani, the lexical items borrowed newly
from Greek took on different oblique endings, such as -os-, -us-, -is-, -as-, through intraparadigmatic levelling. He also adds that the oblique endings were re-analysed into an oblique
marker -s- and a “classification” vowel, or, we might as well say following a traditional term,
a thematic vowel, which he considers to belong to the stem in some dialects. These additional
oblique endings, if they existed in Lovari at some point, have completely disappeared. Not
even descriptive grammars list different athematic classes based on a thematic vowel (cf. e.g.
Cech—Heinschink 1999, Hutterer—Mészáros 1967 or Choli-Daróczi—Feyér 1989). No
matter what the nominative ending of a noun is, it will fall into one of the two classes
mentioned above (the oblique ending being either -es-/-en- or -os-/-on-). Therefore, it is
probably more useful for us to consider the surface forms and their relation, and interpret
words in a more holistic manner instead of breaking them down into even smaller morphemes
(cf. Ackerman et al. 2009). The -os-/-on- pattern, possibly due to its frequency, simply spread
onto the whole athematic layer.
Let us consider two examples: Early Romani *sapuni ‘soap’ (present-day Lovari sapuj) and
Early Romani *?doktori ‘doctor’. The Early Romani oblique form sapunis- has been replaced
by sapujos-, and similarly “doktoros-, i.e. not the original *doktoris-” (Elšík 2000: 23). The
second example, however, is problematic, because there is no reference to the word in earlier
sources, and the oblique appears as doktoros- in the above-mentioned, recent Lovari sources.
Other dialects may behave differently, but doktori could easily be a fairly new loanword,
which has never had an oblique in -is- in Lovari.
Another well-known example is the case of the Greek-derived word fōro ‘town’. We learn
from the sources and informants that the forms of both the singular and the plural oblique
stems are ambiguous: they may be fōres-/fōros- and fōren-/fōron-, respectively. As Elšík 2000
states, diachronically, fōros- replaced fōres-, so that the oblique form could resemble the
nominative singular. This process, however, goes against the basic layout of the “thematic”
inflection, where the oblique singular stem ends in -es-, no matter what the nominative ending
is (for example nominative singular bālo ‘pig’ and oblique singular stem bālés-). It might as
well be the case that the loss of the word-final consonant resulted in a form similar to many
thematic nouns, and the oblique form is slowly taking on the thematic pattern, too – or, if the
historical aspect is considered, re-acquiring it. Using the Saussurean proportion, we can
represent it as follows.
(1) bāló : bālés- = fōró : x
x = fōrésAs for the oblique plural stem, Matras 2002 claims that fōren- became fōron-, possibly based
on an analogy to the nominative singular. On the other hand, Hutterer—Mészáros 1967
mentions that the original form of the oblique plural stem of sokro ‘father-in-law’, from
Romanian socru, is sokron-, but it appears more and more frequently in the form sokren-, and
even the oblique singular stem can be sokres- instead of sokros-, as attested by informants. A
similar process to the singular might have taken place in the plural as well.
If we disregard the thematic dichotomy for a moment, we find that basically there are two
declension classes in Lovari: one masculine and one feminine. The thematic dichotomy, as
Elšík 2000 already notes, is decaying. But is it actually decaying, or do we just see a balance
in variation maintained by constant language change?
In the case of Romani, it is difficult to determine what the original forms of a certain word
were exactly. The fact that the forms fōren-, fōres-, sokren- and sokres- are in use – and they
are spreading, weakening the role of the forms in -os-/-on- – suggests that the thematic classes
seem to exert an analogical force on the athematic borrowings, at least as far as the masculine
is concerned.
This can be related to the fact that many borrowings become obscure; for a bilingual speaker
of Hungarian and Lovari, the words tudōšo and juhāsi might be transparent borrowings; older
borrowings, like fōro and socro might have become more integrated into the system.
Generalisations may be made based on surface patterns that are “stronger” in some aspect;
this might be the case for the historically deeper-rooted thematic pattern which is in constant
opposition with the athematic pattern, which is not as old but has become well-established due
to the high degree of contact Romani has been exposed to.
ACKERMAN, FARRELL, JAMES P. BLEVINS & ROBERT MALOUF. 2009. “Parts and Wholes: Implicative Patterns in
Inflectional Paradigms”. Analogy in Grammar: Form and Acquisition ed. by James P. Blevins & Juliette Blevins,
54-82. Oxford: Oxford University Press. • BALÓ, ANDRÁS MÁRTON. 2012. “Arguments from Lovari loan-verb
adaptation for an analogy-based analysis of verbal systems.” Current Issues in Morphological Theory:
(Ir)Regularity, Analogy and Frequency (Current Issues in Linguistic Theory 322.) ed. by Ferenc Kiefer, Péter
Siptár P. & Mária Ladányi, 3-22. Amsterdam: John Benjamins. • CECH, PETRA & MOZES F. HEINSCHINK. 1999.
Basisgrammatik. Arbeitsbereicht 1a des Projekts Kodifizierung der Romanes-Variante der österreichischen
Lovara (ed. by Dieter W. Halwachs). Vienna: Verein Romano Centro. • CHOLI-DARÓCZI, JÓZSEF & LEVENTE FEJÉR.
1988. Cigány nyelvkönyv [Romani Textbook]. Budapest: Magyarországi Cigányok Kulturális Szövetsége. •
ELŠÍK, VIKTOR. 2000. “Romani nominal paradigms: their structure, diversity, and development”. Grammatical
relations in Romani: The noun phrase. (Current Issues in Linguistic Theory, 211.) ed. by Viktor Elšík & Yaron
Matras, 9-30. Amsterdam: John Benjamins. • HUTTERER, MIKLÓS & GYÖRGY MÉSZÁROS. 1967. A lovári cigány
dialektus leíró nyelvtana [Descriptive Grammar of the Lovari Dialect of Romani]. Budapest: Magyar
Nyelvtudományi Társaság. • MATRAS, YARON ed. 1995. Romani in Contact. The History, Structure and Sociology
of a Language. Amsterdam: John Benjamins • MATRAS, YARON. 2002. Romani: A linguistic introduction.
Cambridge: Cambridge University Press. • SAUSSURE, FERDINAND DE. 1966. Course in General Linguistics. New
York: McGraw-Hill Book Company. • VEKERDI, JÓZSEF. 2000. A Comparative Dictionary of Gypsy Dialects in
Hungary. Budapest: Terebess.
Italian VN compounds are not exocentric
*Universidade Nova de Lisboa; ^Università di Firenze
[email protected] - [email protected] - [email protected]
What is the nature of the verb in Italian ‘exocentric’ Verb-Noun (VN) compounds, cf. (1)?
wash-dishes (‘dishwasher’)
break-ice (‘icebreaker’)
There are many different approaches in the literature to this issue and such point is still a
matter of contemporary debate (Floricic, 2008; von Heusinger & Schwarze, 2013). The
relevant V item has been considered: an imperative form (Rohlfs, 1968; Progovac & Locke,
2009); an indicative third person singular (Di Sciullo, 1992); a deverbal nominal form (Zuffi,
1981; Bisetto, 2004); a root plus a thematic vowel (Scalise, 1992; Vogel & Napoli, 1995; cf.
also Ralli, 2008, who argues that the thematic vowel acts as a linker). All these hypotheses
have weak points (see Ferrari Bridges, 2005; Floricic, 2008). Let us assume that the first
constituent of Italian VN compounds consists of a root form, i.e. a predicative lexical base,
followed by a ‘thematic vowel’. Following an insight of Manzini & Savoia (2005: 487-489),
we propose that the thematic vowels of verbs and nouns have the same categorial signature
N(ominal class). Thus rompi (break) in (1b) or lava (wash) in (1a) can be analysed as in (2)
where the lexical base, indicated as √, expressing predicative content, combines with an N
inflection (-i, -a). In the same way, ghiaccio consists of the predicative base ghiacci- ‘ice’ and
the N inflection –o (we take the -i of piatti in (1a) to be plural morphology). Stems of the √+N
forms are not individuated as verbs or noun. A √+N structure is read as a Noun as a default –
only if embedded under TMA (Tense/Mood/Aspect structures) is it read as a verb. Under
these assumptions, structures like (2) need not be exocentric. Rather they are regular leftheaded compounds where the category neutral stem lava, rompi provides both the semantic
head and the syntactic head. In particular, in the absence of other specifications, the lava,
rompi stems correctly project N by default.
One obvious question concerns the exact nature of N. We assume that predicative bases √ –
not only ‘wash’ or ‘break’, but also ‘ice’ or ‘dish’ – are associated with at least an argumental
slot. The internal argument slot of the latter is generally referred to in the literature as the RRole (Higginbotham, 1985; Williams, 1994). We take it that the N class vowel is the lowest
level saturation of the R-role, eventually to be closed by the determiner D. We assume much
the same for ‘wash’ or ‘break’, eventually to be closed by referential arguments. Another
question concerns the non-existence of ‘nouns’ such as *rompi, *lava in isolation; this has
proved in the past a powerful objection to assuming that they head the N compound in (2). In
other words, why would the stem on the left in (2) be an N only in compounds? The natural
assumption is that the internal theta-role of the lexical base must be closed by more than the N
inflection, specifically by an N level constituent, such as piatti or ghiaccio in (2) – whence the
compounding. A third question concerns the fact that rompi or lava are perfectly good as
verbal forms in isolation. We naturally assume that this is so only to the extent that they are
embedded TMA structures. One of the proposals alluded to at the beginning takes the left-
hand member of the compound in (2) to be a 2 sg imperative. On the contrary, we argue that
such forms can become verbal only to the extent that they are embedded in particular under
the modal position C characteristic of imperatives (for Romance, see Rivero, 1988). Manzini
& Savoia (2005, cf. also Rohlfs, 1969; Giannelli, 1976; Stefanini, 1970) further note that in
the Italian variety of Firenze (Florence) forms morphologically corresponding to the 2 sg
imperative can appear as complements of some verbal classes which usually require
infinitival complements in Romance, as in (3a-b). (3a’) provides comparison with the
prosodically different morphological infinitive, (3b’) with the morphologically identical
imperative. Again modal value accrues to the spiega ‘explain’ and rompi ‘lit: break’ in (3)
from the high modal position of the verb (associated with enclisis, cf. Kayne, 1991).
he …
go.IMP.2SG to
explain√-him COMP (‘go to explain to him that…’)
go.IMP.2SG to
explain.INF-him COMP (‘go to explain to him that…’)
un 'rompi-ʎʎi 'l anima
be.enough.PRS.3SG NEG break√-him the soul (‘it is better not to bother him’)
θanto 'l anima!
break.IMP.2SG-him so.much the soul (‘don’t bother him so much’)
Finally, we shall provide some cross-linguistic discussion of the ways in which different
languages realize agentive/instrumental synthetic compounds (henceforth: A/ISC). Languages
seem to realize the A/ISC configuration, using inflectional morphology or derivational
morphology. Thus an architecture containing inflectional and derivational processes in a
single grammatical component (Williams, 1981) seems to be on the right track. Germanic
languages, like English, tend to use ‘derivational’ morphemes (e.g. the morpheme –er) to
express forms of the dishwasher or icebreaker type (forms like scarecrow would have a zero
morpheme instead of the overt -er suffix, according to e.g. Marchand, 1969; Scalise &
Bisetto, 2005; see also Bauer, 2008). The idea pursued here is that English uses a so-called
derivational tool to do what in Romance is done with inflection, that is to join a nominal class
slot, N, to a (verbal) root, √.
Interestingly, Old English in (5) seems to adopt the inflectional strategy for A/ISCs, since
according to the analysis given in Kastovsky (1985: 246ff, cf. also Gast, 2009) the endings -a,
-e have as their primarily function to determine the inflectional class (i.e. case/number) of the
lexical item to which they attach. In fact, the Old English examples closely resemble Latin
NV items with analogous agentive/instrumental function such as the ones in (6).
law-break-infl (‘law-breaker’)
field-gen-farm-infl (‘farmer’)
sea-travel-infl (‘sea-traveller, sailor’)
part-gen-take-infl (‘part-taker’)
Selected References: BAUER, L. 2008. Exocentric compounds. Morphology 18, 51-74. – FLORICIC, F. 2008. The
Italian verb–noun anthroponymic compounds. Morphology 18, 167-193. – HIGGINBOTHAM, J. 1985. On
Semantics. Linguistic Inquiry 16, 547-621. – MANZINI, M.R. & SAVOIA, L.M. 2005. I dialetti italiani e romanci.
Morfosintassi generativa, vol. III, Alessandria, dell’Orso. – SCALISE, S. 1992. Compounding in Italian. Rivista
di Linguistica 4, 175-199. – WILLIAMS, E. 1994. Thematic Structure in Syntax. Cambridge, MA., MIT Press.
What makes me sleepy? Morphological polyfunctionality of causatives and
reflexives in Mari and Udmurt
Eötvös Loránd University
[email protected]
Impersonal constructions of Finno-Ugric languages have principally been studied from the
viewpoint of verbal morphosyntax (cf. Stipa 1962, Schiefer 1981). Functional investigations
have emerged recently in linguistic typology (Malchukov & Siewierska 2011) and this kind of
approach has been applied for the major Finno-Ugric languages (Posio & Vilkuna 2013). In
my presentation, I will focus on two lesser-studied FU languages, these are Mari and Udmurt.
The goal of this study is to describe some particular impersonals within the terms of
morphological polyfunctionality.
In both languages there are very rich verbal derivational and case marking systems (cf.
Bartens 2000). The present study will describe two verbal suffixes demonstrating
morphological polyfunctionality; causatives and reflexives.
Causation is a valency (transitivity) increasing strategy (Song 2001). Morphological causation
is expressed by the derivational affix -kt and -alt in Mari and -t in Udmurt. In canonical cases,
the causer is marked with the nominative followed by a verb bearing the causative suffix,
while the causee is marked with the accusative in Udmurt. In Mari, the causee is sometimes
marked with the dative:
(1) Udmurt
gožti̮ -t-iśko. (Winkler 2001: 56)
(s)he-ACC letter[ACC]
‘I have the boy write the letter.’
(2) Mari
šolt-əkt-en-Ø. (Kalinina et al. 2006: 446)
Helen-DAT/Helen-ACC soup-ACC boil-CAUS-PST-3SG.II
’Peter made Helen cook (lit. boil) soup.’
In constructions (3) and (4), the causative verb has only one obligatory argument standing in
the accusative:
(3) Ud.
(4) Ma.
beri̮ k-t-e. (Y. S.)
Ač́ a-m
vakǝ̑-kt-a. (T. E.)
’The father is nauseous.’
According to functional considerations, these constructions demonstrate a slightly different
usage of causative verbs that I reckon as impersonal. Following Siewierska (2008: 116), I
consider given constructions to be impersonal if they lack a subject bearing canonical
properties. Subject is determined here in functional terms; a canonical subject is an argument
that is referential, topical, agentive, animate and definite (cf. Malchukov & Ogawa 2011: 23).
As the translations of the latter two examples show, the accusative-marked argument is the
logical subject of the causative impersonal construction, that is, it bears the thematic role of
the Patient or the Experiencer.
A same functional differentiation is presented in the usage of reflexive suffixes in both
languages. The dative-marked argument of the construction displays functional properties
similar to the former type but can be considered less salient:
(5) Ud.
śoti-śk-i-z. (Y. S.)
‘I gave (something) to the child unintentionally.’/‘I was forced to give
(something) to the child.’
The data examined in this study were elicited from native speakers using two questionnaires
which focused on pragmatic neutrality and acceptability factors. Examples were provided by
three informants for each language. To outline the characteristics of the use of these special
kinds of impersonals, I will present data of a corpus based survey. The corpus consists of the
parallel translations of a Russian novel (PM 2010), which contains 12500 words and texts
collected from reference grammars (Bereczki 1990, Csúcs 1990).
BARTENS, RAIJA 2000: Permilaisten kielten rakenne ja kehitys. MSFOu 238.
BERECZKI, GÁBOR 1990: Chrestomathia Ceremissica. Budapest: Tankönyvkiadó.
CSÚCS, SÁNDOR 1990: Chrestomathia Votiacica. Budapest: Tankönyvkiadó.
increase markers interacting with verb semantics. Evidence from Finno-Ugric
languages. In: Kulikov, L., A. Malchukov, P. de Swart (eds.): Case, Valency and
Transitivity, 441–464. Amsterdam: John Benjamins.
MALCHUKOV, ANDREJ & AKIO OGAWA 2011: Towards a typology of impersonal
constructions. A semantic map approach. In: Malchukov, A. & A. Siewierska (eds.):
Impersonal constructions. A cross-linguistic perspective, 19–56. Amsterdam: John
MALCHUKOV, ANDREJ & ANNA SIEWIERSKA 2011: Introduction. In: Malchukov, A. & A.
Siewierska (eds.): Impersonal constructions. A cross-linguistic perspective, 1–15.
Amsterdam: John Benjamins.
PM 2010: Pavlik Morzov -paralleelitekstikorpus. Turku: Research Unit for Volgaic
Languages, University of Turku. (Manuscript)
POSIO, PEKKA & MARIA VILKUNA 2013: Referential dimensions of human impersonals in
dialectal European Portuguese and Finnish. Linguistics 51 (1): 177–229.
SCHIEFER, ERHARD 1981: Das anonyme Subjekt in den finnisch-ugrischen Sprachen. In:
Bereczki Gábor‒Molnár József (eds.): Lakó-emlékkönyv, 140‒157. Budapest.
SIEWIERSKA, ANNA 2008: Introduction: Impersonalization from a subject-centred vs. agentcentred perspective. Transactions of the Philological Society, 106 (2): 115–137.
SONG, JAE JUNG 2001: Linguistic Typology. Morphology and Syntax. Harlow: Longman.
STIPA, GÜNTER 1962: Impersonale Ausdrucksformen. In: Commentationes Fenno-Ugricae in
honorem Paavo Ravila. MSFOu 125: 579–592.
WINKLER, EBERHARD 2001: Udmurt. Lincom Europa.
The Grammatical Category of Case in Early Development of Slovak-speaking Children
Faculty of Arts University of Presov in Presov
[email protected]
The aim of the research is to describe and to explain the development of grammatical ca tegory
of case in early development of Slovak-speaking children (up to 36 months). The main
research question is as follows: What is the development of understanding and production of
case in early development of Slovak-speaking children? In order to find an answer to this
question, we must decide the following sub-questions:
(i) What is the relation between case semantics and lexical semantics?
(ii) Which cases represent preferentially acquired core of case subsystem in Slovak-speaking
(iii) What is the relation between case semantics and pragmatics of child utterance?
We are using two methods: qualitative and quantitative. Qualitative method presents
longitudinal studies of 5 Slovak-speaking children communicating with their parents and care
givers in natural conditions and in standard situations (bathing, feeding and playing). Video
records were made by children’s parents (one hour per month). We are using transcription
system CHAT of CHILDES (CHIld Language Data Exchange System;
http://www.laboratorium.detskarec.sk/ukazka.php). Quantitative method presents Slovak
adaptation of MacArthur-Bates Communicative Development Inventories – Test of
Communicative Behaviour II (TEKOS II http://www.laboratorium.detskarec.sk/tekos2.php).
The research sample consists of 1065 children (539 girls and 526 boys) from 17 to 36 months.
Information about child’s speech development was provided by parents.
(i) The first case forms in early development of Slovak-speaking children appear in 4
semantic categories. The first category is objectness within which names of things and people
in nominative and accusative are formed. The second category called locationality includes
formation of the grammatical case in Slovak-speaking children. The substantive cases are
formed as three opposites: up/down, close/far, inside/outside. We notice these opposites in
non-verbal gestures first, followed by development of verbal means (lexis and grammar). All
case forms are formed within this category:
the opposite inside/outside
beginning of contact with interior of an object: the preposition do/into + the genitive
do domčeka – into the house
duration of contact with interior of an object: the preposition v/in + the locative case
v lese – in the forest
beginning of contact with surface of an object: na/on + the locative case
na deku – on the blanket
duration of contact with surface of an object: na/on + the accusative case
na deke – on the blanket
the opposites close/far (spatial and equally social contact as a means of showing the
need to be with somebody)
beginning of contact with a person: k/to + the dative case
k mame – to mum
duration of contact with a person: s/with + the instrumental case
s mamou – with mum
the opposite down/up
down: pod/under + the instrumental case
pod stolom – under the table
up: nad/above + the instrumental case
nad fľaškou – above the bottle
The third semantic category is benefitiality. A child uses dative or accusative case with
preposition for to name a beneficiary (mame, pre mamu – for mum). The sociativity is the
fourth semantic category which reflects a child’s need to be with somebody, be close to
somebody: with + the instrumental case (s mamou – with mum). This category coincides with
locationality category.
(ii) The case formation in early development of Slovak-speaking children can be divided into
two periods: successive a simultaneous. The grammatical cases (nominative and accusative)
appear in a child’s speech successively. In the development of grammatical case, first the
opposition of nominative – accusative is formed. The quantitative research shows that from
21st month accusative is used by more than a half of Slovak-speaking children. Semantic
cases (dative, locative and instrumental) appear together (simultaneously) in a different order,
age and with different frequency in each child (in 5 longitudinal studies from 19 to 25
months). However, semantic concretization of cases and pragmatic functions of utterances
with case forms develop in all children identically. An example of instrumental: While in 17 th
month an instrumental is used by 5 % of children, in 36 th month it is used by almost 90 % of
children. In early development of Slovak-speaking children the first to be formed is sociative
instrumental (with father) and instrumental of device (by car). Instrumental of place with
prepositions appear less frequently (in order under – in front of – above – behind). The
development of the case is gradual not in stages. The preferential core of the case system
(non-prepositional accusative, dative of benefit, sociative instrumental, accusative case of
plural, direction genitive, direction dative, nominative of plural, locative of place,
instrumental of means, direction accusative) is acquired by Slovak-speaking children in this
order: in 17th month of age from 1 to 10 % of children uses the cases. In 24th month
approximately a half of children do so and in 36 th month almost all (90%) of children in our
sample (1065 children).
(iii) In children acquiring Slovak the preferential core of the case system can be seen in
utterances with three pragmatic functions: con-situational information (a child comments an
activity, that he or she is currently doing), information about an intention (a child wants to do
something with a tool), challenge or a wish (a child wants to do an activity with a close
person). Development of grammatical category of case in early development of Slovakspeaking children demonstrates how cognition (transition to structures of relatedness,
cognitive coordination of an aim and means to achieve it), social experience, experience with
item manipulation and with movement in space reflect into preferentially acquired cases.
EI SENBEISS, S. – NARASIMHAN, B. – VOEIKOVA, M. 2009. The acquisition of case. In: Malchukov, A. and
Spencer, A. (eds.) The Oxford Handbook of Case. Oxford OUP, 369– 383. • GVOZDEV, A. N. 1961. Voprosy
izučenija detskoj reči, 273– 358. Moskva: Izdateľstvo APN RSFSR • PAČESOVÁ, J. 1968. The Development of
vocabulary in the child, 162–251. Brno: Univerzita J. E. Purkyně. • VOEIKOVA, M. – DRESSLER, W. 2002:
Introduction. In: Maria D. Voeikova and Wolfgang Dressler (eds.) Pre- and Protomorphology: Early Phases of
Morphological Development in Nouns and Verbs. Linco m Studies in Theoretical Linguistics 29. Munich:
Lincom Europa, 3–5.
Modular affixation
Form-related operations in an Item-and-Process model of word formation
University of Szeged
[email protected]
In this poster, I shall argue for a modular conception of affixation processes and other formrelated operations that allows for flexible and general solutions of problems in wordformation description. This conception is part of a formal theory of word formation in the
Item-and-Process tradition, called “Pattern-and-Restriction Theory” (PR; Nolda 2012), which
makes crucial use of functional, multidimensional word-formation patterns. In this respect,
PR resembles theories such as Lexeme Morpheme Base Morphology (Beard 1995) or Lexical
Relatedness Morphology (Bochner 1993), although it surpasses them considerably in terms of
formal explicitness and flexibility.
While PR has been fully formalised in axiomatic manner, the poster will present the basic
theoretical tenets and the empirical advantages of the conception as informally as possible, by
applying it to selected examples from German. The analyses of the examples discussed in the
poster will, in addition, be made available as a computer implementation of the theory for a
hands-on experience.
According to PR’s basic idea, the word-formation component of a linguistic system provides a set of word-formation patterns. A pattern is viewed as a multidimensional combination of functional word-formation means: a formal means, certain categorial means, and a
semantic means. Those means are modelled as mathematical operations on forms, categories,
or concepts, respectively. One and the same formal means can be combined with different
categorial and semantic means of the same arity, resulting in formally related, but categorially
and/or semantically different patterns. Given that formal means also include affixation operations, apparent polysemy or homonymy of affixes is not treated as a lexical matter, but as a
grammatical one (for arguments against a lexical conception of affixes, cf. Wurzel 1989;
Beard 1995: chap. 4).
Patterns not only differ in terms of the means they involve, but also in terms of the restrictions they impose on possible bases. This is taken into account by assigning each pattern
of the system a corresponding base restriction.
Formal means capture all sorts of form-related morphological – and, possibly, morphophonological – manipulation, including additive, subtractive, and modificational operations.
Instead of taking basic operations such as concatenation, subtraction, or substitution as primitives of the theory, formal means are typically conceived in PR as specialisations or combinations thereof. By doing so, complex formal means can be defined in a modular way from simpler operations.
Simple affixation operations are defined as one-place operations concatenating a variable
base form and a constant affix form in a given order. For example, the formal means of ungsuffixation in German can be defined as the concatenation of a base form with the affix form
ung (in that order). From simple affixation operations, complex affixation operations can be
derived in various ways. I shall indicate three of them here.
In a first type of complex affixation operations, a simple affixation operation is combined
with a suprasegmental operation. An example at hand is the ˈei-suffixation used in a pattern
for the derivation of location nouns from profession nouns in German: ˌMetzger ˈei ‘butcher’s
shop’ < ˈMetzger ‘butcher’, Konˌditor ˈei ‘confectionery’ < Konˈditor ‘confectioner’, ˌAbt ˈei
‘abbey’ < ˈAbt ‘abbot’, etc. (“ˈ” and “ˌ” denote primary and secondary lexical accents, respectively). This suffixation operation combines simple ˈei-suffixation by functional composition
with a previously applied deaccentuation operation, the latter substituting primary accents in
base forms by secondary ones. Provided that base and affix forms have inherent suprasegmental phonological properties, complex affixation operations of this type make distinctions between ‘class I’ and ‘class II’ affixes (or ‘level I’ and ‘level II’ affixes) dispensable – at least
insofar, as such distinctions are to account for suprasegmental differences in the product
forms (for discussion, cf. Booij 1987, Bochner 1993: chap. 6).
A second type of complex affixation is conditional affixation, which is particularly useful
for modelling allomorphic variation with complementary distribution of formally related affix
forms. Such affixation operations are defined by distinguishing several cases, involving different affix forms, with one of the cases possibly functioning as a default. As an example
consider a variant of the ˈei-suffixation mentioned above, figuring in a pattern for the derivation of iterative process nouns from verbs in German, e.g. ˌTanz er ˈei ‘pointless dancing’ <
ˈtanz (en) ‘to dance’, ˌKletter ˈei ‘pointless climbing’ < ˈkletter (n) ‘to climb’, ˌWackel ˈei
‘pointless wobbling’ < ˈwackel (n) ‘to wobble’ (orthographic case is, of course, phonologically insignificant). Here, ˈei-suffixation, with preceding base deaccentuation, is only applied to
base forms ending in er or el; in the default case, however, deaccentuated base forms are suffixed first by (‘morphomic’) er and then by ˈei. (This affixation operation appears also to be
involved in a pattern deriving location nouns from verbs; for an overview cf. Motsch 2004:
334 f., 351 f.)
A third type of complex affixation operations are substitutive affixations, which substitute
an affix form for an element of the base forms. Such an affixation operation may be assumed
for a pattern for the derivation of German profession nouns from competence domain nouns,
such as ˌPoˈlit ik er ‘politician’ < ˌPolit ˈik ‘politics’ or ˈChem ik er ‘chemist’ < ˌChem ˈie
‘chemistry’ (a pattern which presumably is restricted to certain neoclassical base nouns) . The
formation of those nouns can be analysed in a uniform manner by reference to a formal means
which first deletes the final morphological segment of the base form and accents the final syllable of the remainder, before suffixing both ik and er. As a result, ˈie in ˌChem ˈie is replaced
by ik in ˈChem ik er, while the ˈik in ˌPolit ˈik is only changed suprasegmentally in
ˌPoˈlit ik er. Note that this analysis neither resorts to adjustment rules in the sense of Aronoff
(1976) nor necessarily has to assume bound stem variants like Chem (as Aronoff 1994 does
for nomin ee < nomin ate).
ARONOFF, M. 1976. Word Formation in Generative Grammar. Linguistic Inquiry Monographs 1. Cambridge,
Mass.: MIT Press. • ARONOFF, M. 1994. Morphology by Itself: Stems and Inflectional Classes. Linguistic Inquiry Monographs 22. Cambridge, Mass.: MIT Press. • BEARD, R. 1995. Lexeme Morpheme Base Morphology:
A General Theory of Inflection and Word Formation. Albany: State University of New York Press. • BOCHNER,
H. 1993. Simplicity in Generative Morphology. Berlin: Mouton de Gruyter. • BOOIJ, G. 1987. Lexical Phonology
and the organization of the morphological component. In: Gussmann, E. (ed.), Rules and the Lexicon: Studies in
Word-formation, Lublin: Katolicki Uniwersytet Lubelskiego, 43–65. • MOTSCH, W. 2004. Deutsche Wortbildung
in Grundzügen. 2nd ed. Schriften des Instituts für deutsche Sprache 8. Berlin: de Gruyter. • NOLDA , A. 2012.
Konversion im Deutschen – Muster und Beschränkungen: Mit einem Grundriss einer allgemeinen Theorie der
Wortbildung. Habilitation thesis, Humboldt-Universität zu Berlin. • WURZEL, W. U. 1989. Von der Inadäquatheit einer Affixmorphologie: Weshalb morphologische Kategorienmarker nicht als eigene Einheiten im Lexikon
repräsentiert sein können. Linguistische Studien A 194: 277–298.
Morphological priming in Turkish nominal compound
Middle East Technical University, Ankara, TURKEY
[email protected],
[email protected]
Compounding, constructing new words out of previously known lexemes by means of simple
concatenation mostly, can be counted as one of the major word production mechanisms in the
majority of languages including Turkish.[1,10,11 and 4] Their importance in the history of
human languages warrants a detailed study with respect to the language faculty and related
cognitive aspects.[12] In the last decade, compound production as well as comprehension have
become highly debated and investigated areas of research. Morphological priming is one
frequently employed paradigm for the investigation of compounding. Whether morphologically
complex words undergo a decomposition-composition process, respectively, during
comprehension and production or whether they are all listed in full form in the lexicon is one
key question hitherto addressed using picture naming task in several studies related to English,
German and Dutch nominal compound words [3, 6, 14, and 15].
The present study is concerned with compound production in Turkish. Various types of Turkish
compounds were investigated [2, 5, 7, 8, 9, and 13]:
(1) bare JCs (‘akbalık’, ‘dace’),
(2) indefinite (‘dil balığı’, ‘flounder’)
(3) definite (‘gölün balığı’, ‘fish of the lake’) izafet constructions
in a morphological priming paradigm by means of a picture naming task. In the general
implementation of this task, subjects name black-white line drawings of simple objects in a
limited and pre-specified time-interval while at the same time, they have to ignore distractor
words which are presented visually(or auditorily). The locus of interest in this paradigm is the
evaluation of possible linguistic effects of the distractor word presentation on picture naming
performance. In this study, distractor words were Turkish nominal compounds and picture
names(e.g., ‘balık’, ‘fish’) were morphologically related (depicted either first or second part of
the compound) or completely unrelated to these distractor words. Results of the experiment
revealed equal amounts of morphological priming effect in all compound types investigated,
that is, morphologically related compounds led to shorter naming latencies compared to
unrelated distractors, a result which is in line with the decompositional (also possibly
opportunistic framework proposed by Libben[12]) view of compound processing. Furthermore,
significant animacy effect found on naming latencies irrespective of the compound type,
underlines another possible essential factor in compound processing. Finally, distractor-wise
analysis revealed marginally significant reaction time advantages for the head part of the
compound as compared to the modifier part, a finding which suggests a possible special role for
the head constituent during lexical access.
[1] BAUER, Laurie. Introducing linguistic morphology. 2003. [2]DEDE, Müşerref Ağan. A syntactic and semantic
analysis of Turkish nominal compounds. 1978. PhD Thesis. University of Michigan. [3]DOHMES, Petra;
ZWITSERLOOD, Pienie; BÖLTE, Jens. The impact of semantic transparency of morphologically complex words
on picture naming. Brain and Language, 2004, 90.1: 203-212. [4]GÖKSEL, Aslı. Compounds in turkish. Lingue e
linguaggio, 2009, 8.2: 213-236. [5]GÖKSEL, Asli; KERSLAKE, Celia. Turkish: A comprehensive grammar.
Psychology Press, 2005. [6]KOESTER, Dirk; SCHILLER, Niels O. Morphological priming in overt language
production: Electrophysiological evidence from Dutch. Neuroimage, 2008, 42.4: 1622-1630. [7]KÖNIG, Wolf
Dietrich. Nominalkomposita im Türkischen. Zeitschrift für Sprachwissenschaft, 1987, 6.2: 165-185.
[8]KORNFILT, Jaklin. Turkish. descriptive grammars. London and New York, 1997. [9]LEWIS, Geoffrey L.
Turkish grammar. Clarendon Press, 1985, 2001. [10][LIEBER, Rochelle; STEKAUER, Pavol (ed.). The Oxford
handbook of compounding. Oxford University Press, 2009. [11]LIBBEN, Gary; JAREMA, Gonia (ed.). The
representation and processing of compound words. Oxford University Press, 2006. [12] LIBBEN, Gary. The nature
of compounds: A psychocentric perspective. Cognitive neuropsychology, 2014, ahead-of-print: 1-18.
[13]UNDERHILL, Robert. Turkish grammar. MIT Press, 1976. [14]ZWITSERLOOD, Pienie; BÖLTE, Jens;
DOHMES, Petra. Morphological effects on speech production: Evidence from picture naming. Language and
Cognitive Processes, 2000, 15.4-5: 563-591. [15]ZWITSERLOOD, Pienie; BÖLTE, Jens; DOHMES, Petra. Where
and how morphologically complex words interplay with naming pictures. Brain and Language, 2002, 81.1: 358367.
Irony processing and contextual effects in schizophrenia – A neurolinguistic perspective
a Department of Linguistics and Institute of Psychology, University of Pécs, Hungary
b Department of Psychiatry and Psychotherapy, Medical Faculty, University of Pécs, Hungary
c Neuro CT Diagnostic Center, Hungary
d Department of Neurosurgery, Medical Faculty, University of Pécs, Hungary
e Department of Neurology, Medical Faculty, University of Pécs, Hungary
f MTA-PTE Clinical Neuroscience MR Research Group, Hungary
*Supported by the National Brain Research Program (NAP) KTIA_NAP_13_1_2013_001 Grant
#Supperted by TÁMOP 4.2.4. A/2-11-1-2012-0001 „National Excellence Program and SROP-4.2.2.C-11/1/KONV-20120005 Well-Being in the Information Society project
[email protected]
Schizophrenic patients are known to be diagnosed with atypical Theory of Mind (ToM)
mechanisms even during remission, which is believed to be responsible, or at least, to
contribute to their deficit in pragmatic competence, within that, to the processing of noncompositional constructions. Several studies reveal a connection between irony
comprehension and ToM capacities (Frith & Frith, 2003; Happé, 1993), given that
schizophrenic patients demonstrate difficulty in irony processing, which has been associated
with their deficient mentalization skills. We examined the neural correlates of irony
understanding in schizophrenic patients, as an indicator of ToM capacity, and evaluated how
surface cues (i.e., linguistic help) could affect irony comprehension. Schizophrenic patients in
remission and healthy controls were submitted to event-related functional MRI scanning
while performing three tasks: 1. irony, 2. irony with linguistic help and 3. control tasks.
Mental background of ToM and schizophrenia
Schizophrenia is a pervasive mental disorder often resulting in incapacitation and in social
marginalization. The basis of this neuropsychiatric illness is based on a failure of social
participation, since patients generally have problems with the perception of reality. Their
mental illness generates unreal sensations that they perceive as part of reality. This mental
disturbance is named as mentalization deficit, also known as Theory of Mind (ToM) deficit in
literature, stating that schizophrenic patients have difficulty representing their own and others’
mental states (will, desire, intention, feelings, beliefs and knowledge) realistically (Dennett,
1989; Frith & Frith, 1999; Woodruff & Premack, 1978), making Theory of Mind a core
component of social cognition. For this reason the close examination of mentalization ability
and its interaction with language in communicative settings based on social cognitive skills
can eventually shed light on the yet unknown cognitive background mechanism responsible
for this complex neuropsychiatric disorder resulting in social marginalization and often
isolation, and may result in a training paradigm that can contribute to the reintegration of such
patients into society.
Linguistic background
Irony is a type of polysemy, a non-compositional construction, where intended meaning does
not correspond to the meaning of the constituents. That is, ironic meaning cannot be
interpreted on the basis of sheer linguistic input (semantically), only and strictly integrating
contextual cues, i.e., pragmatically. Thus, to be able to decode the intended (that is, ironic)
meaning of the utterance, a holistic meaning construction is necessary that breaks away from
the semantic analysis. For the successful deciphering of ironic meaning the listener relies on
cognitive strategies, which we define as mentalization, based on a change of perspectives, for
which cues are drawn from the situational and linguistic context. This process of meaning
construction involves the coordination of an implicit communicative intent, where the implicit
meaning is the opposite of what is explicitly expressed (Giora 2001). Irony comprehension
thus represents a form of non-compositional language use that requires not only semantic and
syntactic decoding, but also the deciphering of the speaker’s non-linguistic implicature
(Sperber & Wilson, 1995; 2002, Alberti 2011). Therefore, the integration of intended meaning
and social context is essential for irony comprehension in order to be able to represent the
speaker’s mental state, and decipher their intention, and eventually the intended meaning of
the utterance heard. It is important for the listener to recognize that the actual intention
expressed by the speaker is in fact, the opposite of the utterance heard. Thus, irony
understanding requires not only the correct interpretation of communicative intentions, but
also the ability to construct a coherent narrative based on contradictory information. Up to
now, only a few brain imaging studies have examined the neural basis of irony
comprehension. The present study would like to contribute to this line of research and reveal
potential correlations between language and cognition which can eventually yield a
communicative strategy-based treatment to facilitate integration and reintegration of those
affected by the disorder.
The study
Our aim was to investigate irony comprehension and the underlying brain activity in patients
with schizophrenia. We examined various phases of irony processing separately (such as
context phase, statement phase, as well as question on comprehension and answer phase). We
also evaluated whether providing more explicit contextual information modifies the patients’
performance in irony comprehension and the related brain activities. We hypothesized that
patients with schizophrenia would perform worse and exhibit an abnormal brain activation
pattern during irony comprehension (Varga et al. 2013, 2014). We assumed that linguistic cues
depicting the speaker’s emotional state improve patients’ performance and modify the
network activation. To model complex, real life discourse settings as they take place in natural
languages, we used short social scenarios presented as auditory stimuli in an event related
design, and examined the differences in irony processing at various stages of the paradigm in
patients and controls.
The results indicate that patients handled irony significantly worse than healthy controls. The
patients showed stronger brain activity in the parietal and frontal areas in the early phase of
irony task, however the healthy controls exhibited higher activation in frontal, temporal and
parietal regions in the latter phase of the irony task. Interestingly the linguistic help not only
improved the patients’ ToM performance, but it also evoked similar activation pattern to
healthy controls.
Selected references
ALBERTI, G. 2011. eALIS: Interpreters in the World, Worlds in the Interpreter. Budapest. Akadémiai Kiadó •
DENNETT D. 1989. The intentional stance. A Bradford Book • FRITH, C. D., & FRITH, U. 1999. Interacting
minds – A biological basis. Science. 286(5445), 1692–1695 • FRITH, U., & FRITH, C. 2003. Development and
neurophysiology of mentalizing Philosophical Transactions: Biological Sciences, 358, 459–473 • GIORA, R.
2001. Irony and its discontent. Amsterdam. John Benjamins. 163-182 • HAPPÉ, F. G. 1993. Communicative
competence and theory of mind in autism: A test of relevance theory. Cognition, 48(2), 101–119 • SPERBER,
D., & WILSON, D. (1995). Relevance: Communication and cognition (2nd ed.) Oxford: Blackwell •
SPERBER, D., & WILSON, D. (2002). Pragmatics, modularity and mind-reading. Mind and Language,
17(1&2), 3–23 • VARGA et al. 2013. Irony comprehension and context processing in schizophrenia during
remission – A functional MRI study. Brain and Language 126. 231-242 • VARGA et al. 2014. Compensatory
effect of general cognitive skills on non-literal language processing in schizophrenia. Journal of
Neurolinguistics • WOODRUFF, G., & PREMACK, D. (1978). Does the chimpanzee have a theory of mind.
Behavioral and Brain Sciences 1(4), 515–526.
Unusual and double plural markers in a representative sample of German pre-school
children and adults
Department of Phoniatrics and Pediatric Audiology, Goethe-University, Frankfurt/Main,
Institute of Medical Psychology and Medical Sociology, Center of Psychosocial Medicine,
Faculty of Medicine, University of Göttingen, Germany
Department of Phoniatrics and Pediatric Audiology, Ruhr-University, Bochum, Germany
[email protected], [email protected], [email protected], [email protected]
Numerous studies on plural acquisition in German examined the distribution of plural
allomorphs of the target language in the speech of children, mostly of pre-school age, and in
rare cases in adult samples. However, there is hardly any information on infrequent
phenomena of double plural markers and pseudosuffixes used as plural allomorphs. In this
study, both children and adults were examined in this respect. 810 German children (age
range 3;8-6;0 years, median=4;2) and 582 adults (18-96, median=24;0) were asked to produce
plural forms of 8 nonce words taken from the validated language test SETK 3-5 (Grimm
2001): eine Ribane > Ribanen, ein Tulo > Tulos, eine Plarte > Plarten, ein Biwo > Biwos, ein
Dolling > Dollinge, eine Kland > Klände, ein Ropf > Röpfe, eine Tapsel > Tapseln. Indefinite
articles allowed for consideration of dichotomized (feminine vs. non-feminine) grammatical
gender. Children were examined by means of color pictures, adults had to fill out
Apart from 9,713 usual (single) plural markers, 175 double (and sometimes even triple) plural
markers were documented: 70 in the answers of children and 105 in the answers of adults: -se
(38), -ner (7), -sen (29), -nen (53), -ser (1), -ers (3), -ne (14), -ren (3), -ns (13), -re (5), -res
(2), -ses (3), -nse (2), -rne (2). All plural markers with a double-digit frequency contained
either -(e)n or -s, that is two most frequent plural allomorphs in our sample: -s (2,964
occurrences=30.0% of all plural markers), -e plus umlaut (864=8.7%), -(e)n (4,432=44.8%),
umlaut (98=1.0%), -e (1,289=13.0%), -er (246=2.4%).
A categorical regression was conducted with the classification of double plural markers (-ns,
-re, -res etc.) as the dependent variable and the following independent variables: classification
of test subjects as children or adults, number of syllables in the singular forms of SETK 3-5
items, expected (correct) plural allomorphs, dichotomized gender, application or nonapplication of the schwa deletion rule, and word final sounds (e.g., /n/, schwa, /f/): adjusted
R2=.65, the only important variable being expected plural allomorphs: β=1.37***. However, it
should be mentioned that the extension of the same model to other SETK 3-5 items (real
words Fisch, Vogel, Apfel, Schiff, Gabel, Buch, Stuhl, Bild, Hand, Glas) explained much more
variance (94%) with significant results for expected plural allomorphs (β=.73***), number of
syllables (β=-1.32***), and word final sounds (β=.40***).
There was a highly significant difference in the distribution of double plural markers in the
answers of children and adults: χ2(13)=86.40***. The most obvious difference was a clear
preference of adults for the plural marker -nen (47% of all double plural markers vs. 6% in the
answers of children) and a clear preference of children for the plural marker -se (41% vs. 9%).
Other tendencies were a high frequency of -ne in the answers of adults compared to children
(12% vs. 1%) and a high frequency of -ns in the answers of children compared to adults (16%
vs. 2%). Children often (in≥10% of all double plural markers) used markers -se (Ropfse), -sen
(Ropfsen, Plartesen), and -ns (Tulons, Plartens, Tapselns), adults often used -nen (Biwonen,
Tulonen), -sen (Tulosen, Ribanesen, Biwosen), and -ne (Tulone, Biwone). In the next step an
attempt was made to explain the preference for certain double plural markers in these most
frequent double plural forms given in brackets.
In order to control which word final sounds are associated with which plural markers in the
target language, a word frequency list was built based on a German language corpus (Leipzig
University 2010). The list contains the 1,000 most frequent nouns of each of the three
genders: masculine, feminine, and neuter (tokens). The schwa (Plarte, Ribane) occurs most
frequently either with -(e)n or -s, /f/ (Ropf) either with -e (with or without umlaut) or -s, full
vowels (Tulo, Biwo) either with -s or -n, /l/ (Tapsel) either with -e or -(e)n (however, in cases
where the schwa deletion rule is applicable, -e is substituted by -s). Hence, the combinations
of the two most expected plural markers (according to phonotactic rules) can account for the
forms Ropfse, Plartesen, Tulons, Plartens, Tapselns, Tulosen, Ribanesen, Biwosen. According
to the same frequency list, the sounds /s/ and /n/ in the word final position can be combined
with either -e or -(e)n, hence the high frequency of forms Ropfse(n), Biwone(n), Tulone(n).
The tendency of both children and adults to overgeneralize -s and -(e)n might be linked to the
almost universal applicability of -s and a high frequency of -(e)n. For instance, in the same
frequency list, -s is compatible with 21 word final sounds, -e with 16, -(e)n with 17, -er with
11, umlaut with 3, and -e plus umlaut with 12. Nouns requiring -(e)n account for 46% of the
frequency list, followed by -e (29%), -s (9%), -e plus umlaut (9%), -er (6%), and umlaut
(1%). Noteworthy is a low frequency of -s in comparison with our sample.
14 children and 2 adults used one plural marker which exists neither in the Modern High
German nor in dialects: Klandel, Röfpel. Furthermore, 52 children and one adult tended to
delete -el and thus to reduce nouns to what they probably believed to be plural forms: Tapsel
> Tapse, Tapsen, Taps, where the word final sounds look like plural markers. This uncertainty
about the link between the pseudosuffix -el and pluralization might be traced back to the high
frequency of nouns ending in -el and receiving umlaut – the most seldom and the least iconic
German plural allomorph – as plural marker, which might have led some children to
misinterpret -el as the plural marker. Furthermore, because innumerable German nouns
sometimes receive -el in its diminutive meaning and lose it again depending on context, some
children might have internalized schemata making this element removable (cf. 6 occurrences
of another diminutive suffix -i(e)/y only with the item Tapsel > Tapsi).
Hence, the preference of both children and adults for double plural markers containing -s and
-(e)n can be attributed to an almost universal compatibility of -s with word final sounds and a
very high frequency of -(e)n. Another obvious regularity was found in combinations of two
most expected plural markers allowed by the rules of phonotactics. Although both -s and -n
can function as linking elements in compounds like Arbeitsrecht and Kartenhaus, it can
hardly be assumed that children or adults misused them as linking elements between stems
and second plural markers in the words like Tulonen. Rather, once again, the rules of
phonotactics prompted a higher compatibility of /n/ in the word final position with the plural
marker -(e)n, which was demonstrated on the basis of a frequency list. Interestingly, both
children and adults tended to add double plural markers to the four items following simplest
German pluralization rules (-s after full vowels, -n after schwa), probably as an expression of
creativity in linguistically advanced groups. There is some uncertainty as far as the link
between the pseudosuffix -el and pluralization is concerned. The most important factor
predicting the distribution of double plural markers in nonce words is the correct plural
allomorph which occurs more often in such combinations than other allomorphs. If real words
are also taken into account, number of syllables and word final sounds are also of importance.
GRIMM, H. 2001. SETK 3-5. Sprachentwicklungstest für drei- bis fünfjährige Kinder. Göttingen, Germany:
LEIPZIG UNIVERSITY, DEPARTMENT FOR NATURAL LANGUAGE PROCESSING. 2010. Wortschatz, http://corpora.unileipzig.de/download.html.
Modelling compound properties - new approaches
and new explanations
Sabine Arndt-Lappe (Düsseldorf)
Melanie J. Bell (Cambridge, UK)
Martin Schäfer (Jena)
Barbara Schlücker (Berlin / Konstanz)
From bread baskets to muffin tops:
What can we learn about the mental lexicon from lexical creativity?
Eötvös Loránd University & Indiana University
[email protected]
One of the most remarkable features of linguistic creativity in English is the predominance of
metaphorical and metonymical compounds in everyday language. This is immediately
apparent from the vast number of examples that can be found in dictionaries. One such
example is belly button, denoting the ‘navel’, coined in 1934 (Oxford English Dictionary;
henceforth OED). Belly, however, fell from grace in the Victorian era and became a taboo
word; the wonderfully apt bread basket and pudding house were coined instead for the same
concept (Burridge 2004: 42). Needless to say, both expressions are metaphor- and metonymybased: 1) an image schematic metaphor establishes the image-based resemblance between the
shape of a bread basket or pudding house (the latter of which is a round container for
puddings) and the belly; and 2) a CAUSE FOR EFFECT metonymy provides further motivation by
creating a mapping between the thing that is eaten (bread and pudding, respectively) and the
result (a bulging belly).
The ongoing appeal of metaphor- and metonymy-based compounds is also demonstrated
by the more recently formed muffin top, which can be considered as one of the success stories
of English word-formation. It was originally coined in 2003 by two Australian TV presenters
to denote the spare flesh that overhangs loose-fitting jeans. In 2006, it was named as “Word of
the Year” by the Australian Macquarie Dictionary and was also elected among the “most
creative” terms the same year by the American Dialect Society. It eventually entered the
online edition of the OED in March 2011.
Nevertheless, despite the preponderance of metaphorical and metonymical compounds in
English, not much has been said about them in morphological literature on the grounds that
they are not based on productive word-formation processes due to their “exocentric” nature.
Apparently, in exocentric (or headless) compounds the head “falls outside” of the expression
(hence the term “exocentric”), and the compound expression is not a hyponym of the head
element. Such coinages are distinguished from endocentric ones, where the compound
represents a subcategorization of the entity expressed by the head element (e.g., apple tree is a
type of tree).
The endocentric–exocentric distinction has had a profound – and rather negative – effect
on the scope of morphological research into compounding in the sense that linguistic literature
has a strong tendency to focus on exocentric combinations only peripherally (if they are
mentioned at all). This fact is also underlined by Scalise and Guevara (2006: 185): “In fact,
while there is an extensive literature on endocentric compounding, the references to the
theoretical and/or typological treatment of exocentric compounds are very rare” (emphasis
added). Although descriptivist approaches do make reference to exocentric combinations,
these are typically more superficial than the detailed classifications of endocentric compounds
(see, for example, Jespersen 1954; Adams 1973). An exception to this trend is Marchand
(1960), who devotes a whole chapter to the classification of exocentric compounds. As
regards to the transformational generativist account, it left the issue of metaphorical and
metonymical compounds untouched, probably for the simple reason that the theoretical
framework was unable to accommodate such combinations.
Cognitive linguistics, however, has demonstrated that these “exocentric” compounds are
indeed analysable with the application of conceptual metaphor and metonymy and blending
among others. As emphasized by Langacker (1987), Talmy (1988) and Croft and Cruse
(2004), both metaphor and metonymy can be considered as a type of construal operation, and
as such, a certain way of interpreting/conceptualizing the world around us. What this implies,
therefore, is that the use of and reliance on conceptual metaphors and metonymies in word
formation must also be an absolutely natural process. Benczes (2006) has extensively shown
that such expressions can be analyzed remarkably well within a cognitive linguistic
framework. The use of metaphors and metonymies in novel compound formation opens up a
limitless supply of innovation and creativity in novel word-formation, as such expressions
make use of the creative associations that exist between concepts; associations based on
similarity, analogy or contiguity.
However, how can all this be applied to a better understanding of the mental lexicon? In
line within the cognitive linguistic framework, and by analyzing numerous examples, the talk
will attempt to bring together the semantic network approach of Langacker (1987) with more
recent psycholinguistic results on compound representation (El-Bialy, Gagné & Spalding
2013; Gagné 2009: 265; Ji, Gagné & Spalding, 2011; Libben 2006). As it will be
demonstrated throughout the presentation, creative language use involves ad hoc, contextual
and analogical information that is “rooted in long-term memory or the immediate physical or
linguistic context” (Gerrig & Gibbs 1988: 14), and any account of the mental lexicon must
necessarily be able to explain and accommodate such examples of everyday creativity.
ADAMS, V. 1973. An Introduction to Modern English Word Formation. London: Longman.
BENCZES, R. 2006. Creative Compounding in English: The Semantics of Metaphorical and Metonymical Noun–
Noun Combinations. Amsterdam & Philadelphia: John Benjamins.
BURRIDGE, K. 2004. Blooming English: Observations on the roots, cultivation and hybrids of the English
language. Cambridge: CUP.
CROFT, W. & D.A. CRUSE. 2004. Cognitive Linguistics. Cambridge: Cambridge University Press.
EL-BIALY, R., Ch.L. GAGNÉ & Th.L. SPALDING. 2013. Processing of English compounds is sensitive to the
constituents’ semantic transparency. Mental Lexicon 8/1: 75–95.
GAGNÉ, Ch.L. 2009. Psycholinguistic perspectives. In: R. Lieber & P. Štekauer (eds.), The Oxford Handbook of
Compounding. Oxford: OUP, 253–71.
GERRIG, R.J & R.W. GIBBS. 1988. Beyond the lexicon: Creativity in language production. Metaphor and
Symbolic Activity 3/1: 1–19.
JESPERSEN, O. 1954. A Modern English Grammar on Historical Principles. Part VI: Morphology. London:
Bradford and Dickens.
JI, H., Ch.L. GAGNÉ & Th.L. SPALDING. 2011. Benefits and costs of lexical decomposition and semantic
integration during the processing of transparent and opaque English compounds. Journal of Memory and
Language 65: 406–30.
LANGACKER, R.W. 1987. Foundations of Cognitive Grammar, Vol I. Stanford, CA: Stanford UP.
LIBBEN, G. 2006. Why Study Compound Processing? An overview of the issues. In: G. Libben & G. Jarema
(eds.), The Representation and Processing of Compound Words. Oxford: OUP, 1–22.
MARCHAND, H. 1960. The Categories and Types of Present-Day English Word-Formation: A SynchronicDiachronic Approach. Wiesbaden: Otto Harrassowitz.
Oxford English Dictionary, The. 1989. 2nd ed., online version. Oxford: Oxford University Press.
SCALISE, S. & E. GUEVARA. 2006. Exocentric compounding in a typological framework. Lingue e Linguaggio 2:
TALMY, L. 1988. The relation of grammar to cognition. In: B. Rudzka-Ostyn (ed.), Topics in Cognitive
Linguistics. Amsterdam & Philadelphia: Benjamins, 165–205.
Compounds in German – Influences on prosodic constituency
University of Freiburg – German department
[email protected]
In prosodic morphology, compounds in German are traditionally mapped onto two prosodic
words that together recursively form another prosodic word (henceforth: pword).
Accordingly, the compound Zahnarzt ‘dentist’ (lit. ‘tooth doctor’) is analyzed as
((zahn)ω(arzt)ω)ω. Evidence for this analysis stems from the fact that the syllable related law of
onset maximization is blocked by the intervening pword boundary: Instead of [tsa:.naPtst],
which we would expect according to onset maximization, we yield [tsa:n.ʔaPtst]. Syllable
boundaries thus coincide with pword-boundaries. Additionally, glottal stop insertion
strengthens the initial boundary of the second pword. Although this view on prosodic
constituency of morphological compounds is rather uncontroversial (cf. Hall 1999,
Raffelsiefen 2000, Wiese 2000), two points of criticism may be raised. The first one refers to
the fact that the analyses are mostly based on introspectional data only. There are hardly any
accounts of how compounds are actually realized in spoken German, and whether the
phonological and phonetic evidence for pword status – like blocking of resyllabification and
glottal stop insertion – is really existent, and if so, whether there are gradient differences in
boundary strength. The second point of criticism relates to the first one by focussing on
factors that may systematically influence the phonetic realization of compounds. While it is
hardly ever denied that factors like speech rate or speech style may have an influence on
acoustic reductions at the pword boundary, other factors like frequency or morphological
informativeness are only beginning to being looked upon (but see e.g. Pluymaekers et al.
2005, 2010 for Dutch and Ernestus & Hanique 2012 for a critical overview).
Against this background, the purpose of the present talk is twofold: Firstly, it will shed some
light on the acoustic-phonetic realization of the pword boundary in German compounds, and
secondly, it will discuss the effect of word frequency and other independent variables on
pword boundary strength. The results to be presented stem from two types of data:
(1) In a carefully designed acoustic production study, twelve north-western German speakers
produced target items in a laboratory setting. All targets were embedded into carrier
sentences. The boundary phenomena of interest are glottal stop insertion / glottalization and
degemination across the pword boundary. The independent variables are word frequency,
accentuation, vowel quantity and segmental context. For instance, glottal stop and/or
glottalization are investigated in word pairs like high-frequency Zahnarzt (‘dentist’) vs. lowfrequency Bahnarzt (‘doctor of the railway’) (n = 756); degemination is investigated in word
pairs like high-frequency Stadttour (‘city tour’) vs. low-frequency Watttour (‘tour of the
tidelands’) (n = 896). The results indicate that prosodic boundary strength in compounds is
indeed systematically influenced by frequency, but structural factors like vowel quantity and
segmental context cannot be dismissed either.
(2) An ongoing corpus study of spontaneous speech investigates durational reduction at the
pword boundary in compounds with /l/-initial second constituents (e.g. Urlaubslohn ‘holiday
pay’) which are then compared to derivations with –los (e.g. arbeitslos ‘unemployed’) (n =
414). From a prosodic point of view, both types of complex words are similar in that they
consist of two pwords: ((arbeits)ω(los)ω)ω. Despite this similarity, however, it may be
hypothesized that durational reduction at the pword boundary varies systematically with other
factors. The independent variables of primary interest are morphological structure
(compounding vs. derivation), semantic transparency and absolute word frequency.
Preliminary results suggest that semantic transparency does not play a role for boundary
realization in complex words, whereas morphological structure and word frequency lead do
systematically different realizations.
HALL, A.T. 1999. The Phonological Word: A Review. In: A. T. Hall and U. Kleinhenz (eds.),
Studies on the Phonological Word. Amsterdam: John Benjamins, 1-22.
HANIQUE, I., ERNESTUS, M. 2012. The role of morphology in acoustic reduction. Lingue e
linguaggio 2, 147-164.
PLUYMAEKERS, M., ERNESTUS, M. AND BAAYEN, R.H. 2005. Lexical Frequency and Acoustic
Reduction in Spoken Dutch. Acoustical Society of America Journal 118, 2561-2569.
PLUYMAEKERS, M., ERNESTUS, M., BAAYEN, H.R. AND BOOIJ, G. 2010. Morphological effects
on fine phonetic detail: The case of Dutch -igheid. In: C. Fougeron, B. Kühnert, M.
D'Imperio and N. Vallée (eds.): Laboratory Phonology. Berlin / New York:de Gruyter,
RAFFELSIEFEN, R. 2000. Evidence for word-internal phonological words in German. In: R.
Thieroff, M. Tamrat, N. Fuhrhop and O. Teuber (eds.): Deutsche Grammatik in Theorie
und Praxis. Tübingen:Max Niemeyer Verlag, 43-56.
WIESE, R. 2000. The phonology of German. Oxford: Oxford University Press.
Relational competition and meaning construction in compound interpretation and
Department of Psychology, University of Alberta
[email protected] [email protected]
The available evidence on the processing of morphologically complex words is consistent
with a lexical system that is highly productive and actives both whole-word and constituent
representations (Ji, Gagné, & Spalding, 2011; Kuperman, Bertram, & Baayen 2010; Libben
2006). Such a system requires that compound processing involve the integration of
constituent representations. Thus, it is critical to understand what this integration entails.
Several studies indicate that integration involves meaning construction (e.g., Fiorentino &
Poeppel 2007) and the use of semantic relations to link the constituents (Gagné & Spalding
2009; Ji, et al. 2011). Of course, compounds vary in terms of the extent to which the
constituents’ meanings contribute to the meaning of the compound, and there is ongoing
debate about whether the semantic representations of the constituents of opaque compounds
are available during compound processing. Even so, there is some recent evidence that even
opaque compounds are affected by relational competition (Gagné & Spalding, 2014; Spalding
& Gagné, 2014), suggesting that meaning construction is attempted even for opaque
In this talk, we present studies that examine whether semantically priming the first constituent
of a compound influences the processing of that compound, and whether any such priming
effect depends on the semantic transparency of the compound’s constituents. We investigate
this question using a lexical decision task to evaluate interpretation, where the primary
evidence for meaning construction exists. We also use a typing task to evaluate written
production, where one might believe that meaning construction would be unnecessary. We
find that semantic priming affects processing in both the interpretation and production tasks,
but that such effects are complicated. Semantic transparency of the constituents affects
whether semantic priming results in changes to processing, but it is not just the semantic
transparency of the primed constituent that exerts an influence—for example, the semantic
transparency of the head affects whether semantically priming the modifier results in a change
in lexical decision latencies or typing times. In fact, the semantic transparency of the primed
constituent seems not to matter in the typing task, and matters only in conjunction with the
semantic transparency of the head in the lexical decision task.
Overall, the results are consistent with a meaning construction process that gives rise to
competing meanings during both interpretation and production of compound words. If
meaning construction is occurring, then competition among the interpretations might offset
benefits due to increased availability of lexical and morphemic information (see, for example,
El-Bialy et al. 2013; Ji et al. 2011). That is, because constructed meanings compete,
manipulations that increase the availability of a constituent’s semantic, lexical, or
morphological representations may either facilitate or inhibit processing, depending on the
other properties of the item being processed and on the properties of the processing task. A
better understanding of the integration process might help resolve some of the conflicting
results in the literature, especially those concerning the availability of morphemic and
semantic representations of the constituents.
EL-BIALY, R., GAGNÉ, C. L. & SPALDING, T. L. (2013). Processing of English compounds is sensitive to the
constituents’ semantic transparency. The Mental Lexicon, 8, 75-95.
FIORENTINO, R. & POEPPEL, D. (2007). Compound words and structure in the lexicon. Language and Cognitive
Processes, 22, 953-1000.
GAGNÉ, C. L. & SPALDING, T. L. (2009). Constituent integration during the processing of compound words:
Does it involve the use of relational structures? Journal of Memory and Language, 60, 20-35.
GAGNÉ, C. L., & SPALDING, T. L. (2014). Relation diversity and ease of processing for opaque and transparent
English compounds. In F. Rainer, Dressler, W., Gardini, F, Luschützky, HC (Eds.). Morphology and
Meaning: Current Issues in Linguistic Theory. Amsterdam: John Benjamins.
JI, H., GAGNÉ, C. L., & SPALDING, T. L. (2011). Benefits and costs of lexical decomposition and semantic
integration during the processing of transparent and opaque English compounds. Journal of Memory and
Language, 65, 406-430.
KUPERMAN, V., BERTRAM, R., & BAAYEN, R. H. (2010). Processing trade-offs in the reading of Dutch derived
words. Journal of Memory and Language, 62, 83-97.
LIBBEN, G. (2006). Why study compound processing: An overview of the issues. In G. Libben & G. Jarema
(Eds.), The representation and processing of compound words (pp. 1-22). Oxford, New York.
SPALDING, T. L., & GAGNÉ, C. L. (2014). Relational diversity affects ease of processing even for opaque English
compounds. The Mental Lexicon, 9, 48-66.
From genitive suffix to linking element in German
Johannes Gutenberg University Mainz
[email protected]
Nominal compounds in German can have one of two structures: In the first one, similar to
English, the two nouns are simply combined (Tisch+decke ‘tablecloth’). The second makes
use of intervening linking elements (Geburt+s+ort ‘birth-LE-place’, Akte-n-ordner ‘file-LEfolder’), a phenomenon that can also be observed in several other Germanic languages (e.g.
Dutch, Luxemburgish).
The origin of most German linking elements is generally considered to be syntactic in nature
(cf. Pavlov 1983, Nitta 1987, Demske 2001). Due to the frequent preposing of genitive attributes in Early New High German (1350–1600), reanalysis of a genitive attribute and its head
lead to a new, compound-like structure:
(1) des
Erbe-n > die
Leib- s-
erbe-n ‘legitimate heirs’
By analogy, these fossilized genitive suffixes spread to new formations or existing compounds in which a former syntactic origin is no longer indicated due to specific morphological
or semantic properties. This resulted in a new pattern of compounding that quickly spread.
German compounds with linking elements are a widely studied aspect of word formation.
However, thorough quantitative corpus studies that focus on the genesis and expansion of this
pattern are still lacking. The present study aims to fill this gap by addressing the following
1. At which point can we assume a new pattern of compound formation as opposed to singular, lexeme specific instances of reanalysis?
2. Can the ambiguous structures leading to reanalysis be quantified?
3. Was the expansion and current distribution of German linking elements guided by morphological, phonological or semantic factors and how did these interact?
The study is based on data from a corpus of Early New High German and New High German
with texts ranging from 1500 up 1710. It is a strong(e)n
ly modified version of the Bergmann & Nerius
(1996) corpus. The corpus is divided in 8 periods, 5
regions and two text types, making a detailed analysis of a changing system possible.
This diachronic data is essential for the interpretation of today’s system. About 39% of today’s news
ly formed compounds make use of linking elements 27%
(see Figure 1). A number of theories have been pro59%
posed to explain the persistance of this seemingly
useless phonological material (Aronoff & Fuhrhop
2002, Wegener 2003, Krott et al. 2004, Berg 2006,
Nübling & Szczepaniak 2008). The present study
seeks to shed light on the role the proposed morpho- Figure 1: Distribution of linking elements in
recent German neologisms (2009–
logical, phonological or semantic functions have
2010, n=1,754; own study).
played in the emergence of linking elements.
Aronoff, Mark & Fuhrhop Nanna (2002): Restricting suffix combinations in German and English: Closing suffixes and the monosuffix constraint. In: Natural Language & Linguistic Theory 20, 451–490. Berg, Thomas
(2006): The internal structure of four-noun compounds in English and German. In: Corpus Linguistics and Linguistic Theory 2, 197–231. Bergmann, Rolf/Nerius, Dieter (eds., 1996): Die Entwicklung der Großschreibung
im Deutschen von 1500 bis 1700. Heidelberg. Demske, Ulrike (2001): Merkmale und Relationen. Diachrone
Studien zur Nominalphrase des Deutschen. Berlin, New York. Krott, Andrea & Hagoort, Peter & Baayen,
Harald (2004): Sublexical Units and Supralexical Combinatorics in the Processing of Interfixed Dutch Compounds. In: Language and Cognitive Processes 19.3, 453–471. Nitta, Haruo (1987) : Zur Erforschung der
'uneigentlichen' Zusammensetzungen im Frühneuhochdeutschen. In: ZfdPh 106. 400–416. Nübling, Damaris
& Szczepaniak, Renata (2008): On the way from morphology to phonology: German linking elements and the
role of the phonological word. In: Morphology 18, 1–25. Pavlov, Vladimir M. (1983): Von der Wortgruppe
zur substantivischen Zusammensetzung. Zur Ausbildung der Norm der deutschen Literatursprache (1470–1730),
Vol. 4. Berlin. Wegener, Heide (2003): Entstehung und Funktion der Fugenelemente im Deutschen, oder:
warum wir keine *Autosbahn haben. In: Linguistische Berichte 196, 425–457.
Branching direction and phonetic reduction in triconstituent compounds
Heinrich-Heine-Universität Düsseldorf
[email protected], [email protected]
Studies investigating the order of derivational affixes in English (e.g. Plag & Baayen 2009)
have provided evidence that affix boundaries differ in boundary strength: in a word of the
form [[base-X]-Y], the outer boundary between [base-X] and affix Y is stronger than the inner
boundary between the base and affix X. Hay (2003) and Plag & Baayen (2009) argue that
these strength differences affect the degree not only of decomposability, but also of
phonological integration: affixes at weaker boundaries will show a higher degree of
phonological integration than affixes at stronger boundaries.
Similar boundary effects are reported in Sproat (1993) and Sproat & Fujimura (1993),
who investigate the phonetic implementation of the same segment string at different types of
boundary (no boundary, affix, compound, phrasal, utterance). Their findings are compatible
with the conclusion that weaker boundaries show more phonological integration: for instance,
the acoustic duration of the same segments at an affix boundary is shorter than at a compound
boundary. Likewise, Hay (2007) finds a difference in duration between the prefix un- and
non-morphemic word-initial un-, as well as differences in vowel reduction.
Taken together, the results from the phonetic and the affix order studies predict that
the phonetic implementation of a word with more than two morphological constituents will
reflect its morphological structure, i.e. its internal bracketing. We hypothesize that the
embedded form shows more phonetic reduction than forms at higher derivational levels
(‘Embedded Reduction Hypothesis’).
To test this prediction, we analyzed triconstituent compounds. Traditionally, the
branching direction of triconstituent compounds (e.g. [child care] center vs. university
[textbook]) has been extensively discussed in the context of stress assignment. Recent studies
(for example, Kösling 2013 and Kösling et al. 2014) have shown, that, contrary to what has
been assumed in much of the earlier literature (in the wake of Liberman and Prince 1977), the
branching direction of noun-noun-noun compounds cannot be read off from the stress pattern.
In the present study, we use the data from Kösling et al.’s experiment to explore the
prediction of the Embedded Reduction Hypothesis that the embedded constituents show more
reduction than the third, non-embedded, constituent. The statistical analysis confirms that
prediction. The constituents of the embedded compound were significantly shorter than the
single constituent at the higher compositional level. This effect holds for left-branching and
right-branching compounds. The results present clear evidence that the branching direction,
i.e. morphological structure, is phonetically encoded.
This poses a challenge for most formulations of Lexical Phonology (e.g. Kiparsky
1982), which argue that the internal structure of morphologically complex forms is not
accessible anymore at the post-lexical stage, with which phonetic variation like sub-phonemic
durational differences or phonetic reduction are usually associated. If there is reliable
evidence that the acoustic signal contains phonetic detail signaling the internal structure of
morphologically complex words, and if this detail affects the processing of the acoustic signal
by listeners, the strict division between lexical and post-lexical components needs to be
HAY, JENNIFER B. 2003. Causes and consequences of word structure. London: Routledge.
HAY, JENNIFER B. 2007. The phonetics of 'un'. In Judith Munat (ed.), Lexical creativity, texts and contexts.
Amsterdam: John Benjamins. 39-59.
KÖSLING, KRISTINA. 2013. Prominence assignment in English triconstituent compounds. PhD thesis. Universität
Siegen. Available at http://dokumentix.ub.uni-siegen.de/opus/volltexte/2013/674/.
KÖSLING, KRISTINA, GERO KUNTER, R. HARALD BAAYEN & INGO PLAG. 2013. Modelling pitch contours in
English tripartite compounds. Language and Speech. DOI: 10.1177/0023830913478914
KIPARSKY, PAUL. 1982. From cyclic phonology to lexical phonology. In Harry van der Hulst & Norval Smith
(eds.), The structure of phonological representations. Dortrecht: Foris Publications. 131-175.
LIBERMAN, MARK Y. & ALAN PRINCE. 1977. On stress and linguistic rhythm. Linguistic Inquiry 8(2): 249-336.
PLAG, INGO & R. HARALD BAAYEN. 2009. Suffix Ordering and Morphological Processing. Language 85(1):
SPROAT, RICHARD. 1993. Looking into words. In Sharon Hargus & Ellen M. Kaisse (eds.), Studies in lexical
phonology. San Diego, CA: San Diego Academic Press. 173-195.
SPROAT, RICHARD & OSAMU FUJIMURA. 1993. Allophonic variation in English /l/ and its implications for
phonetic implementation. Journal of Phonetics 21: 291-311.
Individual differences in compound processing
Brock University
[email protected]
In the psycholinguistic literature, a great deal of progress has been made in the understanding
of how compound constituents contribute to the overall meaning of compound words and the
understanding of the conditions under which compound constituents are activated in lexical
processing (e.g., Spalding, Gagné, Mullaly & Ji. 2010; Marelli & Luzzatti, 2012).
Traditionally, however, matters of individual variation have received relatively less attention.
In this paper, a psycholinguistic model of the mental representation of compounds is
presented that highlights the extent to which individuals and populations can differ in terms of
how compounds are represented and processed. The model has two key components:
(1) Morphological Transcendence: This is the claim that the representation of compound
words in the mind changes through the lifespan as a result of the experience that an individual
language user has with processing specific compounds and morphological families.
(2) Morphological Superstates: This is the claim that compound constituents exist
psychologically in a morphological superstate up to the point at which they are measurable
through acts of language production or comprehension.
The postulation of morphological transcendence and morphological superstates predicts that
the constituent activation effects that can be measured through psycholinguistic experiments
will depend on the compound processing experience of the individual language user and the
specific performance demands of the experimental paradigm. These predictions were tested
through a series of experiments with native speakers of English.
Experiment 1 employed the P3 experimental paradigm (Libben, Weber, & Miwa, 2012) in
which participants are visually presented with compound stimuli in a Primed Progressive
Demasking paradigm. The task of the participants is to say the stimulus word aloud as quickly
as possible and then to type it. It was found that individual participant characteristics
including education, linguistic background and metalinguistic knowledge affected
performance on both tasks. Moreover, the typing production data revealed a substantial effect
of task performance characteristics as predicted by the morphological superstate hypothesis.
In Experiment 2, the P3 paradigm was employed in dyadic word recognition and production
paradigm in which two participants are tested at the same time. One member of the dyad sees
progressively demasked stimuli and says them aloud. The other participant types the
compound stimuli (as one would in a classic dictation task). Results show that performance on
the task is not only influenced by the characteristics of individual participants and the
disparity between the two participants in the dyad. A critical difference between these two
variables is that the former is plainly evident to participants in the experiment at the time of
testing, whereas the latter is not. We introduce a new data visualization technique, Participant
Profiles, that facilitates the understanding of the relation between individual and dyadic
participant characteristics as well as their relation to patterns of compound processing.
LIBBEN, G., WEBER, S., & MIWA, K. (2012). P3: A technique for the study of perception, production, and
participant properties. The Mental Lexicon, 7:2, 237-248.
MARELLI, M., & LUZZATTI, C.G. (2012). Frequency Effects in the Processing of Italian Nominal Compounds:
Modulation of Headedness and Semantic Transparency. Journal of Memory and Language, 66, 644-664.
SPALDING, T., GAGNÉ, C. MULLALY A.,& JI, H. 2010. Relation-based interpretations of noun–noun phrases: A
new theoretical approach. In Susan Olsen (ed.), New impulses in word-formation (Linguistische Berichte
Sonderheft 17), 283–315. Hamburg: Buske.
A model of present-day English compound spelling
University of Erlangen-Nuremberg
[email protected]
A question many language users are frequently faced with when spelling English compounds
is whether to use open spelling (drinking fountain), hyphenation (far-off) or solid spelling
(airport). In the literature, two views dominate:
1. The spelling of English compounds is chaotic (e.g. Fowler 1926: 243; MerriamWebster 2001: 99).
2. The spelling of English compounds develops from an open via a hyphenated stage
towards solid spelling (e.g. Quirk et al. 1985: 1537).
These two views are not necessarily mutually exclusive, since a different speed of
development of individual compounds may result in what looks like superficial chaos. A
large-scale empirical study was conducted in order to show that neither of these general views
is entirely correct and that there are determinants of present-day English compound spelling.
In my paper, I am going to present the central points of that study, which goes beyond
previous research on the spelling of English compounds (e.g. Sepp 2006; Rakić 2009;
Kuperman and Bertram 2013), since is not limited to noun+noun compounds and uses a wider
compound concept. Compounds are defined as complex lexemes
a) which consist of at least two constituents occurring as free synchronically recognisable
and semantically relevant lexemes each and
b) which contain no affixation on the highest structural level.
The study uses various types of linguistic evidence to test over 60 hypotheses concerning the
spelling of English compounds. Among other things, it analyses
• lemma lists from six dictionaries, one of which served for the compilation of the
Master Compound List comprising more than 10,000 compounds.
• corpus data from analogously structured corpora for different periods and varieties of
English (BLOB-1931, LOB, FLOB, BE2006 for British English; Brown and FROWN
for American English).
• information on the first spelling recorded in the Oxford English Dictionary.
In contrast to previous research, which has mainly focused on variation in compound spelling,
the study that I am going to present in my paper follows an approach that is based on the
observation that the spelling of a number of English compounds is largely standardised and
does not seem to vary. My study establishes empirically which out of a large number of
features (such as word length, stress pattern etc.) are responsible for this special behaviour.
The variables with a statistically significant effect are then used as predictors for the spelling
of other compounds. Not only do the results that I am going to present in my paper call into
question the general views of English compound spelling outlined above: in addition, they
were used to develop various spelling algorithms with different degrees of complexity. The
predictive accuracy of the decision trees was tested on compounds belonging to different
1. compounds with exclusively open, hyphenated or solid spelling in 5-6 dictionaries.
2. compounds with spelling variation across dictionaries – but which exhibit a preference
for one particular spelling variant.
3. compounds from CompText, an 8,864-word corpus of present-day British English
compiled specifically for the study.
I am going to show in my paper that even a minimal algorithm performs better than educated
English native speakers in a decision test on the most prototypical English compounds. This
seems to suggest that while it is indeed possible to recognise principles underlying the
present-day spelling of English compounds, these are not immediately obvious.
The theoretical analysis of the data discusses the application of different theoretical models
to the phenomenon of English compound spelling. My paper presents an overview of the most
important aspects and reaches the conclusion that both a modified prototype model (cf. e.g.
Rosch 1973) and a constructionist account (cf. e.g. Goldberg 2006; Stefanowitsch and Gries
2003) are suitable for the modelling of the data. In addition, analogy (cf. e.g. Bybee 2003;
Krott, Baayen and Schreuder 2001; Plag, Kunter and Lappe 2007), which is frequently
contrasted with rule-based principles, can be reconciled with my feature-based approach by
taking into account that the features used have actually emerged from a large set of exemplars
(cf. also Bell and Plag 2012: 517).
BELL, MELANIE, and INGO PLAG (2012): “Informativeness is a determinant of compound stress in English.”
Journal of Linguistics 48/03, 485-520.
BYBEE, JOAN (2003): Phonology and Language Use. 2nd ed. Cambridge: CUP.
FOWLER, HENRY W. (1926): A Dictionary of Modern English Usage. Oxford: Clarendon.
GOLDBERG, ADELE (2006): Constructions at Work: The Nature of Generalization in Language. Oxford: OUP.
KROTT, ANDREA, HARALD BAAYEN and ROBERT SCHREUDER (2001): “Analogy in morphology: Modeling the
choice of linking morphemes in Dutch.” In: Linguistics 39/1, 51-93.
KUPERMAN, VICTOR and RAYMOND BERTRAM (2013): “Moving spaces: Spelling alternation in English nounnoun compounds.” In: Language and Cognitive Processes 28/07, 939-966.
MERRIAM-WEBSTER (2001): Merriam-Webster’s Guide to Punctuation and Style. 2nd ed. Springfield,
Massachusetts: Merriam-Webster.
Oxford English Dictionary (2009). 2nd ed. on CD-Rom (v. Oxford: OUP.
PLAG, INGO, GERO KUNTER and SABINE LAPPE (2007): “Testing hypotheses about compound stress assignment
in English: A corpus-based investigation.” In: Corpus Linguistics and Linguistic Theory 3/2, 199-233.
QUIRK, RANDOLPH et al. (1985): A Comprehensive Grammar of the English Language. London: Longman.
RAKIĆ, STANIMIR (2009): “Some observations on the structure, type frequencies and spelling of English
compounds.” In: SKASE Journal of Theoretical Linguistics 6. http://www.skase.sk/Volumes/
ROSCH, ELEANOR (1973): “On the internal structure of perceptual and semantic categories.” In: Moore, Timothy
E. (ed.): Cognitive Development and the Acquisition of Language. New York: Academic Press, 111-144.
SEPP, MARY (2006). Phonological Constraints and Free Variation in Compounding: A Corpus Study of English
and Estonian Noun Compounds. American Dissertation.
STEFANOWITSCH, ANATOL and STEFAN TH. GRIES (2003): “Collostructions: Investigating the interactions of
words and constructions.” In: International Journal of Corpus Linguistics 8, 209-243.
The polysemy of Italian VN compounds in a diachronic perspective
Charles University in Prague
[email protected]
The VN compounding (of the type portalettere) in Italian (and, in general, in the Romance
languages) is a well-studied and well-described phenomenon (cf., e.g., Tollemache 1945;
Scalise 1994: 134-136; Bisetto 2004: 45-47; Scalise - Bisetto 2008: 133-135; Dardano 2009:
190-198; Ricca 2010, among others) – to the extent that one may question the necessity to
deal with it in another study. The structural patterns (e.g., the nature of verb form) and also
the semantic outputs (agentive, instrumental, eventive, etc.) are effectively very well known.
However, recent research has shown that at least two questions are worth being further
developed. First, it is the quantitative aspect of the productivity and of the overall distribution
of the VN compounds, which is a question revisited especially thanks to the development of
large language corpora (cf. Ricca 2010). Second, it is the diachronic development of the
pattern in question. Even though it is generally known that VN compounds represent a
morphological innovation which is common to virtually all Romance languages (cf. Bauer
2010), a precise quantitative – diachronically based description – is still missing (but cf.
Moyna 2011 for a diachronic treatment of Spanish compounding). Thus, for example, the
question of the limited number of verb bases which enter the pattern (porta-, salva-, mangia-,
...), as well as the question of the historical coverage of the meanings (e.g., is the agentive
meaning the primary one?), calls for a thorough elaboration.
In fact, beside the empirical results which are obviously very important, a theoretical
question arises as well, namely the diachronic variability of word-formations rules (whatever
definition is adopted – Aronoff 1976 vs. Corbin 1987 or some other new formulations...) and
the associated semantic instructions of these WFRs (e.g., the status of the unitary output base
hypothesis, cf. Ricca 2010).
Both questions – empirical and theoretical – are precisely what I wish to address in
this paper, and I wish to analyse them along the following lines. As far as the empirical basis
is concerned, I will present the diachronic data drawn from the corpus LIZ 4.0 within the time
span that goes from the 16th to the 19th century (the 20th century being the period on which all
the available synchronic studies are based and will thus be excluded from the survey), for
each century separately; I shall also present the complementary data from the major
diachronic dictionaries (from the 1612 Vocabolario degli Accademici della Crusca up to the
1859 Tommaseo’s dictionary, along with some specialised dictionaries, cf. for a general
lexicographic overview Marazzini 2009).
This empirical basis will reveal that the quantitative characteristics are well different
for the 16th century as opposed to the present-day situation in that even within the class of the
most frequent „leading bases“ (Ricca 2010: 247; Gather 2001), the number of different types
(and their token frequencies) is rather low (for porta- the 16th century corpus offers just
portaflagello, portalettere, portanovelle, portaseggiette). This indicates that even if the VN
compounding has always been available (in Corbin’s sense of disponibilité), the pattern has
become fully profitable (again in the sense of rentabilité) only recently. At the same time,
also the range of possible meanings is rather limited being there – in the time span in question
– especially the agentive VN compounds.
This empirical situation thus leads directly to the theoretical question whether the
semantic instruction associated with the WFR in question has always been identical regardless
of any diachronic variability. In fact, as shown by Ricca (2005; 2010; cf. also Magni 2010),
the VN compounds may not only denote agents, instruments, events, locations (though the
quantitative representation of each type is not equal), but they also function – productively –
as adjectival modifiers (e.g., cannone sparaneve, giro spaccagambe). Ricca raises the
theoretical question whether this adjectival output is part of the WFR; and he also alludes to
the marginality of the above mentioned meanings in the history of Italian (claiming that event
VN compounds are rare also in diachrony).
I maintain that both questions are closely connected, i.e. the diachronic empirical
research could provide us with some concrete answers also to the theoretical problems
regarding the development of the WFRs in general. I intend to put forward the hypothesis
according to which there is a core semantic instruction tied up with the readily available
pattern which is, at least as far as the VN compounds are concerned, diachronically constant
(cf. for Spanish similar observations made by Moyna 2011: 210-211), and the increasing
profitability could effectively exploit it or also leave it unexploited; at the same time, the
modifier function is simply a contextual enrichment which is not part of the WFR (along the
lines of Ricca 2010: 254).
The first claim is supported by the diachronic evidence (e.g., salvadanaio is an old
instrument noun showing thus that also in diachrony the VN compounds were not exclusively
agents; and much new evidence comes also from specialised historical dictionaries); the
second claim is a little harder to support diachronically given the data scarcity in the period
under investigation. Nevertheless, we shall present a couple of examples which go in this
ARONOFF, M. 1976. Word Formation in Generative Grammar, Cambridge (MA): The MIT Press.
BISETTO, A. 2004. Composizione con elementi italiani. In: Grossmann, Maria e Rainer, Franz (a cura di), La
formazione delle parole in italiano. Tübingen: Niemeyer.
BAUER, B. 2010. Word formation. In: Maiden, Martin and Smith, John Charles, Ledgeway, Adam (eds.), The
Cambridge History of the Romance Languages. Volume I. Structures. Cambridge: Cambridge
University Press, 2010, pp. 532–563.
CORBIN, D. 1987. Morphologie dérivationnelle et structuration du lexique, 2 voll., Tübingen: Niemeyer.
DARDANO, M. 2009. Costruire parole. La morfologia derivativa dell'italiano. Bologna: il Mulino.
GATHER, A. 2001. Romanische Verb-Nomen Komposita. Wortbildung zwischen Lexikon, Morphologie und
Syntax. Tübingen: Narr.
IL TOMMASEO. 2004. Dizionario della lingua italiana di Niccolò Tommaseo e Bernardo Bellini (1857-1879).
Testo integrale degli otto tomi originali e interrogazione per lemmi e a tutto testo in CD-ROM.
Bologna: Zanichelli.
LIZ 4.0. 2001. Letteratura italiana Zanichelli. CD-ROM dei testi della letteratura italiana, a cura di Pasquale
Stoppelli ed Eugenio Picchi. Bologna: Zanichelli.
MAGNI, E. 2010. From the periphery to the core of Romance [VN] compounds. Lingue e Linguaggio, 9, 1, 2010,
pp. 3–39.
MARAZZINI, C. 2009. L’ordine delle parole. Storia di vocabolari italiani. Bologna: il Mulino.
MOYNA, M. I. 2011. Compound Words in Spanish. Theory and History. Amsterdam: John Benjamins.
RICCA, D. 2005. Al limite tra sintassi e morfologia: i composti aggettivali V-N nell'italiano contemporaneo. In:
Grossmann, M. – Thornton, A. M. (a cura di). La formazione delle parole. Atti del XXVII Congresso
Internazionale di Studi della Società di Linguistica Italiana. L’Aquila, 25-27 settembre 2003. Roma:
Bulzoni, pp. 465–486.
RICCA, D. 2010. Corpus data and theoretical implications. With special reference to Italian V-N compounds. In:
Scalise, Sergio and Vogel, Irene (eds.), Cross-Disciplinary Issues in Compounding. Amsterdam: John
Benjamins, 2010, pp. 237–254.
SCALISE, S. 1994. Morfologia. Bologna: il Mulino.
SCALISE, S.  A. BISETTO 2008. La struttura delle parole. Bologna: il Mulino.
TOLLEMACHE, F. 1945. Le parole composte nella lingua italiana. Roma: Rores.
Information-theoretic approaches to morphology
James P. Blevins (Cambridge)
Uncertainty and discriminability in information-theoretic WP
J B
M R
University of Cambridge Eberhard Karls University, Tübingen
[email protected]
Classical WP models depart from prevailing symbol-processing paradigms in two fundamental respects. e first is that classical WP models aim to minimize the 
that arises in associating units with grammatical properties, not to minimize the size of units,
or the ‘redundancy’ of inventories. Associated with this aim is a conception of languages as
‘complex systems’ in which dimensions of variation do not function as individually meaningful ‘signs’ or even as ‘realizations’ but instead serve to  larger units that are
meaningful within the system. is talk suggests how formalization in terms of information
theory and discriminative learning theories clarifies these properties of classical WP models.
e relevance of information theory to morphology generally is enhanced by the fact that
the communicative function of morphological variation conforms well to the idealization of
communication as “selecting from a set of possible messages” (Shannon : ). WP models are particularly amenable to formalization in information-theoretic terms given their
emphasis on minimizing indeterminacy in the mapping between features and forms. e
goal of uncertainty reduction dictates the choice of units and structures; neither words nor
paradigms have a privileged status a priori in WP models. e abstraction of word-sized
units and their assignment to larger paradigmatic structures is justified by the reduction in
uncertainty that this achieves in a grammatical analysis. e word is the basic unit of morphological analysis because it is, on the one hand, a maximally discriminable “perceptual
gestalt” Hockett (: ), and, on the other, “a more stable and solid focus of grammatical
relations than the component morpheme by itself ” (Robins : ). e association between words and paradigm cells ‘anchors’ a form within a system of grammatical contrasts,
facilitating implicational deductions that further constrain uncertainty within the system.
Pushing analyses below the word level increases indeterminacy, not only by isolating units
that are individually more ambiguous than the forms they are extracted from, but also by
disrupting patterns that locate a form within an inflectional paradigm or derivational family.
e roles that words and paradigms play in reducing uncertainty are summarized in ().
a. e uncertainty that arises in relating word forms and grammatical properties
is less than (or equal to, in the case of simple forms) the sum of the uncertainty
that arises in associating their component morphs with properties.
b. e uncertainty associated with a paradigm is less than (or equal to, in the case
of single-cell paradigms) the sum of the uncertainty of their component cells.
(a) reformulates the claim that “grammatical statements … are more profitably abstracted
from words as wholes than from individual morphemes” (Robins : ), whereas (b) ex-
presses a version of the Low Conditional Entropy Conjecture of Ackerman & Malouf ().
Both claims can be formalized in terms of standard entropy measures.
A discriminative perspective further clarifies the structure of WP models, and offers a principled account of morphological patterns that are unexpected or problematic on decompositional models, whether morphemic or realizational. Although classical WP models are
oen faulted for failing to recognize morphemic structure, the classical position reflects a
discriminative rather than an associative interpretation of form variation.
As with the rest of western Antiquity, Priscian’s grammatical model is word and
paradigm and he expressly denied any linguistic significance to divisions in what
would now be called morphemic analysis below the word. (Robins : )
e locus of form contrasts is clearly identified in analyses that specify what Matthews
(: ) terms “different formal changes by which oblique Cases were derived from the
‘upright’ Case”. Yet this variation does not signify or even realize grammatical meaning but
discriminates between larger meaningful units. Hence discriminative learning approaches
(Ramscar et al. ) provide the most appropriate models for the formalization of the exponence patterns of classical WP accounts. A discriminative perspective also offers natural
explanations for the fact that combinations of elements are oen distinctive, that elements
in isolation may differ from the ‘same’ element when it occurs within a larger form, as well
as for the observation that no division or multiple divisions of a form may be motivated.
Ackerman, F. & Malouf, R. (). Morphological organization: e Low Conditional Entropy Conjecture. Language , –.
Hockett, C. F. (). Refurbishing our Foundations: Elementary Linguistics from an Advanced Point
of View. Amsterdam: John Benjamins.
Matthews, P. H. (). Morphology. Cambridge: Cambridge University Press.
Ramscar, M., Yarlett, D., Dye, M., Denny, K. & orpe, K. (). e effects of feature-label-order
and their implications for symbolic learning. Cognititive Science , –.
Robins, R. H. (). In defence of WP. Transactions of the Philological Society , –. Reprinted
in Transactions of the Philological Society , , –.
Robins, R. H. (). A Short History of Linguistics. London: Longman.
Shannon, C. (). A mathematical theory of communication. e Bell System Technical Journal ,
–, –.
Implicative structure and inflectional classification
Université Paris-Sorbonne, Institut Universitaire de France, Laboratoire de Linguistique Formelle
[email protected]
Descriptions of inflection class systems take many forms, depending on the level of documentation of the language, the theoretical preferences of the author, and the goals of the classification.
Pedagogical grammars are often content with giving a broad classification in major classes. At the
other end of the spectrum, various attempts at making sense of the structure of inflection systems
presuppose a classification into fine-grained micro-classes that exhaustively partition the set of
lexemes (see e.g. Stump and Finkel’s (2013) notion of a plat). The two types of classifications can
be linked by assuming a hierarchically-organized system of classes, where macro-classes are seen
as groupings of micro-classes (Dressler and Thornton, 1996). Although there are various ways of
designing such hierarchies, we limit ourselves here to tree-shaped hierarchies with monotonous
inheritance of inflectional properties.
Inflectional classifications are often used to reason about the typology of inflection systems.
For such reasoning to be meaningful, it is crucial that the classifications defined for different languages be commensurable. Linguists use combinations of heuristics to decide on a classification,
which are hardly ever defined precisely. Thus existing classifications for different languages may
not be commensurable. Of particular concern is the fact that inflectional classifications are often
dependent on disputed decisions on segmentation: whether two lexemes should be taken as belonging to the same class is deduced from the fact that they use the same affixal exponents in the
same paradigm cells. However in the absence of a consensus as to the exact boundary between
stem allomorphy and affixal exponence, this results in the possibility of nontrivially different classifications.
One obvious way of addressing this issue is to implement systematized heuristics and apply
these to large inflected lexica (Brown and Evans, 2012; Lee and Goldsmith, 2013). While the
heuristics might not capture equally well the structure of different systems, at least criticizing the
heuristics will be made easier, and the classifications will be commensurate. In this presentation
we will exemplify a formally simple-minded way of doing this, where infection classes are deduced
directly from implicative relations between surface forms. For concreteness I will show examples
from French conjugation. All data is taken from the Flexique lexicon (Bonami et al., 2014).
The general strategy is very similar to that of Brown and Evans (2012). We start by establishing a distance matrix between all lexemes in the lexicon, and then use a clustering algorithm to
deduce a tree-shaped classification. To evaluate distance between lexemes, we start by inferring
a classification of patterns of relatedness between pairs of paradigm cells. Notice that this is the
same initial step used by Ackerman et al. (2009) and subsequent literature. The inflection of each
lexeme can then be represented by a tuple of patterns relating each pair of cells, and the distance
between two lexemes can be defined as the Hamming distance between tuples. This is exemplified
in Tables 1 and 2, where a sample of French verbs are reduced to three of their indicative present
cells. In this subparadigm,
inflect in exactly the same way: the distance between
them is 0. Both
differ from
in two positions, and are thus at distance 2
from it, but they differ from each other in all three positions.
differs from
in all three
dimensions and is thus more distant from
than either
From the distance matrix one may deduce a hierarchy using some hierarchical clustering algorithm. For concreteness we use here agglomerative average linkage clustering (Sokal and Michener, 1958): items that are minimally distant form a cluster; distance between clusters is evaluated
by taking the average of the pairwise distances between the items they contain. The resulting
clustering for the current dataset is shown in Figure 1.
Tuple of patterns
,3 ⇀
↽3 ,2 ⇀
⟨X ⇀
↽ Xe
⟨X ⇀
↽ Xe
⟨ XɛC ⇀
↽ XəCe , X
⟨X ⇀
↽ Xse
⟨ Xwa ⇀
↽ Xs
↽ Xv
, Xe ⇀
, Xe ⇀
, XəCe ⇀
↽ XɛC
, Xe ⇀
, XyCe ⇀
↽ XwaC
Table 1: A toy inflection system based on French conjugation
Table 2: Hamming distances
deduced from Table 1
Figure 1: Dendrogram deduced by hierarchical
clustering from the distance matrix in Table 2
The present method strikes a balance between unsupervised learning and the practice of descriptive linguists. This makes it useful as a tool for empirical exploration. Importantly, we implement a clear notion of inflectional microclass: two lexemes that intuitively have the exact same
inflection have a distance of 0. By contrast, Brown and Evans’s method always finds a non-zero
distance between two nonidentical paradigms. The price to pay for this is that some linguistic
knowledge is built in the classification of patterns of alternation. This is done however without
relying on a constant segmentation scheme, avoiding the methodological quandaries that plague
traditional discussions of inflection classes.
In the talk we will show the results of applying the present method to the full paradigm of
the 4952 verbs in the Flexique dataset, and compare the resulting classification to the handdesigned classification of Kilani-Schoch and Dressler (2005). We will also discuss the heuristics
implicit in the choice of Hamming distances and particular clustering algorithms, and argue that
these decisions, although they must be discussed and justified, are direct analogues of the decisions
taken by a descriptive linguist building a classification. Thus if anything, inference of inflection
classes from patterns of alternation should help assessing the reliability of classification schemes.
Finally, we will discuss how paradigm entropy in the sense of Ackerman et al. (2009) can be used
to assess the cohesion of clusters.
, F., B
, J. P.,
, R. 2009. Parts and wholes: implicative patterns in inflectional
paradigms, pp. 54–82. In J. P. Blevins and J. Blevins (eds.), Analogy in Grammar. Oxford University Press, Oxford.
, O., C
, G.,
, C. 2014. Construction d’un lexique flexionnel phonétisé libre du français.
In Actes du quatrième Congrès mondial de linguistique française, Berlin. • B
, D.
, R. 2012. Morphological complexity and unsupervised learning: validating Russian inflectional classes using high frequency data,
pp. 135–162. In F. Kiefer, M. Ladányi, and P. Siptár (eds.), Current Issues in Morphological Theory: (Ir)regularity,
analogy and frequency. John Benjamins, Amsterdam. • D
, W. U.
, A. M. 1996. Italian nominal inflection. Wiener Linguistische Gazette 55-57:1–26. • K
, M.
, W. 2005. Morphologie
naturelle et flexion du verbe français. Gunter Narr Verlag, Tübingen. • L , J.
, J. A. 2013. Automatic morphological alignment and clustering. Presented at the 2nt AIMM. • S
, R. R.
, C. D.
1958. A statistical method for evaluating systematic relationships. The University of Kansas Scientific Bulletin 38.
, G. T.
, R. 2013. Morphological Typology: From Word to Paradigm. Cambridge University
Press, Cambridge.
Reading skill and early inflectional processing
The University at Albany, SUNY &
Haskins Laboratories
The University at Albany, SUNY
Eberhard Karls University Tubingen,
University of Novi Sad, Serbia
A now classic debate in the domain of morphological processing is whether native
speakers process regular (talked-talk) and irregular (spoke-speak) verb forms in the same
manner. Those who advocate a single processing account characterize effects of
regularity in terms of graded form similarity along with consistent meaning similarity and
argue that processing of all words benefits from similarity between prime and target
(Basnight-Brown, Chen, Shu, Kostić, & Feldman, 2007; McClelland & Elman, 1986;
Rueckl, Mikolinski, Raveh, Miner, & Mars, 1997; Rueckl & Raveh, 1999; Rumelhart &
McClelland, 1986; Seidenberg & Elman, 1999). As a result, regularity effects emerge
because many irregular past and present forms have less similarity between their
orthographic and/or phonological form (e.g., letter overlap is higher in TALKED-TALK
than in SPOKE-SPEAK) or because regular stems tend to differ from irregular stems in
terms of their similarity to other words along semantic dimensions (Baayen & del Prado
Martín, 2005). Alternatively, advocates of a dual route account claim that recognition of
regular past-tense verbs (talk–talked) is a rule-based process where an -ed past tense
marker is affixed to the stem of the verb. For the many (about 180) irregular verbs in the
English language characterized by a past tense that does not include an -ed ending, there
is no rule. Further, when past tense forms of irregular verbs cannot be formed by rule,
those forms must be stored in rote memory (Pinker & Ullman, 2002; Prado & Ullman,
2009). Because many assert that frequency effects reflect storage (but see Baayen, 2009;
Baayen, Wurm, Aycock, 2007), "dual routists" also assert that frequency effects should
be more robust for irregular than for regular forms.
In the visual lexical decision task, we compare forward masked morphological
facilitation after identity, inflected ED and ING and unrelated primes for regular and
irregular verbs. Within regularity, we manipulate surface frequency.
Results replicate the study of Milin et al. (2014): an effect of spelling proficiency
that fails to interact with either target type or prime type, and a tensor product of two
item-based principal components of form (PC1) and frequency (PC2). In addition, effects
of morphological versus unrelated prime type were significant and it did not interact with
the target inflection (simple, ED and ING). Crucially, the contrast between regular vs.
irregular past tense primes was weak for simple targets (talked-talk marginally faster than
spoke-speak) and absent for inflected forms of those same targets (talked-talking and
spoke-speaking fail to differ).
Results challenge the decomposition account of facilitation that is at the core of
the dual route account of inflectional processing.
Baayen, R. H. (2009). Demythologizing the word frequency effect: A discriminative learning
perspective The Mental Lexicon. 5(3):436-461.
Baayen, R. H., Hendrix, P., & Ramscar, M. (2013). Sidestepping the combinatorial explosion:
Towards a processing model based on discriminative learning. Language and Speech, page in
Baayen, R. H., & del Prado Martín, F. M (2005). Semantic density and past-tense formation in
three Germanic languages Language, 666-698
Baayen, R. H., Milin, P., Filipović Durdević, D., Hendrix, P., and Marelli, M. (2011). An
amorphous model for morphological processing in visual comprehension based on naive
discriminative learning. Psychological Review, 118(3):438{481.
Baayen, R. H., Wurm, L. H., Aycock, J. (2007). Lexical dynamics for low-frequency complex
words: A regression study across tasks and modalities The Mental Lexicon 11/2007;
Basnight-Brown, D., Chen, H., Shu, H., Kostić, A. and Feldman, L. B. (2007). Monolingual and
Bilingual Recognition of Regular and Irregular English Verbs: Does Sensitivity to Word
Form Vary with Language Experience? Journal of Memory and Language.57, 65-80.
Milin, P., Ramscar, M., Cho, K., Baayen, R.H. & Feldman, L. B. (2014). Processing partially
and exhaustively decomposable words: An amorphous approach based on discriminative
Pinker, S., & Ullman, M. T. (2002). The past and future of the past tense. Trends in Cognitive
Sciences, 6, 456–463.
Prado, E. & Ullman, M., (2009). Can Imageability Help Us Draw the Line Between Storage and
Composition? Journal of Experimental Psychology: Learning, Memory, and Cognition 2009,
Vol. 35, No. 4, 849–866
Rueckl, J. G., Mikolinski, M., Raveh, M., Miner, C. S., & Mars, F. (1997). Morphological
priming, connectionist networks, and masked fragment completion. Journal of Memory and
Language, 36, 382–405.
Rueckl, J. G., & Raveh, M. (1999). The influence of morphological regularities on the dynamics
of a connectionist network. Brain and Language, 68, 110–117.
Rumelhart, D., & McClelland, J. (1986). On learning the past tense of English verbs: implicit
rules or parallel distributed processing? In J. McClelland, D. Rumelhart and the PDP
Research Group (Eds.), Parallel distributed processing: Explorations in the microstructure of
cognition. Cambridge, MA: MIT Press.
Seidenberg, M. S., & Elman, L. M. (1999). Networks are not hidden rules. Trends in Cognitive
Sciences, 3, 353–365.
Morphemes in context and in isolation: A recursive
view on measures of frequency and complexity.
Ghent University, 2University of Novi Sad, 2Eberhard
Karls University of Tübingen, 3Cambridge University
[email protected]
Measures of frequency and complexity are used pervasively in the study of word processing.
For instance, in research on the recognition of complex words, surface frequency, lemma
frequency, family size, and derivational entropy, have all been shown to have an influence on
morphological processing (e.g., Amenta & Crepaldi, 2012).
In this talk, we will show that nearly all measures of frequency and complexity are variants of
three fundamental measures: element frequency, element dispersion, and element entropy,
which can be derived recursively by assuming that the contexts in which elements are seen
become themselves the elements of interest in the next recursion.
For instance, let us consider unbound morphemes to be elements of interest, and let us
consider context to be the material surrounding the element up to the word boundaries. The
token count of the contexts in which the element (morpheme) occurs is the measure of
element frequency (in this case morpheme frequency). The type count of the contexts is the
measure of element (morpheme) dispersion and the entropy of the probability distribution of
contexts is the element entropy. Interestingly, the final two measures are also known as family
size and derivational entropy (e.g., Baayen, Feldman, Schreuder, 2006).
If we now consider the elements in context to be the elements of interest, we are operating
over words in the context of surrounding words. The measure of element frequency now
corresponds the classic measure of word frequency. Next, the measure of element dispersion,
which could be called word family size, simply indicates in how many word trigrams a word
of interest occurs centrally. Finally, the probability distribution of word trigrams is the basis
for the computation of the measure of element entropy at this level.
It should now be clear that in the following recursion, measures of word trigram frequency,
word trigram dispersion, and word trigram entropy are obtained, and so on.
While we have so far considered context as elements directly surrounding the elements of
interest, the definition of context is variable. For instance, starting with morphemes as
elements of interest, but looking only at material directly to the left or to the right give us
measures closely related to prefixation and suffixation at the first step of recursion. At a
higher level, considering the edges of a document in which a word occurs as boundaries, gives
us contextual diversit (Adelman, Brown, & Quesada, 2006) as the measure of element
This systematic approach yields several interesting results. In our talk, we will focus on three
findings in particular.
First, to our knowledge, word dispersion (the family size measure at the word level) has not
been investigated. However, this measure seems to be superior to word frequency in
predicting lexical decision times on words presented in isolation. This finding by itself is quite
surprising, as word frequency is the most important and most robust predictor of isolated
word processing. This shows that our approach fills gaps: it allows us to identify important
measures which are simply absent from the literature.
Second, at a lower level of granularity, considering phonemes as elements of interest
generates the measure of phoneme entropy. In a phoneme-trigram context, this measure
provides a simple and elegant insight in the concept of sonority, with elements with higher
sonority having higher entropies.
Finally, the same finding occurs at the level of morphemes, where measures of morpheme
entropy give an indication of the freedom with which morphemes can combine.
ADELMAN, J. S., BROWN, G. D. A., & QUESADA, J. F. (2006). Contextual Diversity, Not Word Frequency,
Determines Word-Naming and Lexical Decision Times. Psychological Science, 17(9), 814–823.
AMENTA, S., & CREPALDI, D. (2012). Morphological Processing as We Know It: An Analytical Review of
Morphological Effects in Visual Word Identification. Frontiers in Psychology, 3.
BAAYEN, R. H., FELDMAN, L. B., & SCHREUDER, R. (2006). Morphological influences on the recognition of
monosyllabic monomorphemic words. Journal of Memory and Language, 55(2), 290–313.
The effect of grammatical context on processing Serbian inflected nouns and verbs
Laboratory for Experimental Psychology
Faculty of Philosophy, University of Belgrade
[email protected]
Inflected languages are characterized by grammatical congruency among different
word classes. Thus, for example, Serbian nouns are congruent with adjectives and possessive
pronouns in case, grammatical number and gender, verbs are congruent with personal
pronouns in person, grammatical number and sometimes gender etc. Any violation of this
congruency will produce an ungrammatical sentence. The congruency is in most cases
realized via inflectional suffixes attached to the root morpheme of an open class word.
In four experiments using the lexical decision task it was demonstrated that Serbian
inflected noun forms are primed by adjectival and pronominal context. When preceded by
congruent adjectival and pronominal primes, noun forms were processed faster than when
preceded by incongruent context. (Gurjanov, Lukatela, Moskovljević, Savić & Turvey,
1985a; Gurjanov, Lukatela, Savić & Turvey, 1985b).
The information-theoretic approach to processing inflected morphology is based on
the assumption that the cognitive system is sensitive to the uncertainty of inflected word
forms, specified in terms of the amount of information (bits). The general formalism from
which the amount of information is derived consists of two distinct terms: the probability of
an inflected form and the grammatical component. The grammatical component in case of
nouns is the number of syntactic functions/meanings (R) of an inflected noun form, while in
case of verbs it is the number of congruent personal pronouns (Q) /Equ. 1 and 2/.
⎛ F
⎜ Rm
I m = ⎢ − log 2 ⎜ k
Fm j
⎜ j =1 Rm
⎛ F
⎜ Q
I m = ⎢ − log 2 ⎜
k F
m j
⎝ j =1 m j
In a series of experiments it was demonstrated that the amount of information derived
from the above equations accounts for almost all processing variability of Serbian inflected
noun and verb forms presented in isolation (Kostić, 1995; Kostić & Havelka, 2002; Kostić,
Marković & Baucal 2003). The question is how to account for priming effects observed with
Serbian inflected nouns. The approach adopted in this study is the following: facilitation and
inhibition effects are co-implicative and take place simultaneously within a defined
morphological paradigm. For example, Serbian regular feminine nouns appear in six distinct
inflected forms which constitute the paradigm of feminine nouns. Congruent primes will
decrease the amount of information carried by the target stimulus and proportionally increase
the amount of information carried by other paradigm constituents.
This approach was evaluated in three experiments using the lexical decision task. In
the first experiment two noun forms (suffixes –i and –u) were preceded by neutral context
(***) and congruent and incongruent adjectival primes. In the second experiment the same
noun forms were preceded by neutral context and congruent and incongruent possessive
pronouns. In the third experiment the first and second person verb in the present tense were
preceded by neutral context and congruent and incongruent personal pronouns.
In case of adjectival and pronominal priming, modification of the amount of
information carried by a noun form should be applied at the level of probability of a noun
form because adjective forms, personal pronoun forms and noun forms share the same
syntactic functions/meanings. Likewise, the amount of information carried by verb person
should also be modified at the level of its probability (Equ. 1 and 2).
There are two ways to carry out these modifications. The most obvious way is to
apply the formalism for conditional probabilities. The equation for conditional probabilities is
defined as the ratio of the probability of the joint of events A and B, and the probability of B.
An alternative way is to assume that the modified probability of the target in congruent
contexts is the sum of the probabilities of the prime and the target. Consequently, the
probability of the remaining forms within a paradigm is proportionally decreased.
In regression analyses it was demonstrated that the amount of information derived
from the equation for conditional probabilities does not account for significant proportion of
processing latency variability in the three experiments. The informational values derived
from the alternative approach (i.e. summing up prime and target probabilities) accounted for
almost all processing variability in the experiments (see Figures 1 and 2).
Figure 1. Relation between the amount of information and processing latency in Experiments 1 and 2. Left:
noun targets with suffixes –i and –u in Exp. 1; Middle: suffix –i in Exp. 2, right: suffix –u in Exp. 2
Figure 2. Relation between the amount of information and processing latency in Experiment 3. Left scattergram:
first person present tense; Right scattergram: second person present tense
There is, however, a serious conceptual problem with this approach. Summing prime
and target probabilities implies that the two events are independent (mutualy exclusive). This,
in turn, requires revision of the general information-theoretic framework that describes the
processing consequences of grammatical congruency among different word classes.
GURJANOV, M., LUKATELA, G., MOSKOVLJEVIĆ, J., SAVIĆ, M., &: TURVEY, M. T. 1985a. Grammatical priming
of inflected nouns by inflected adjectives. Cognition, 11, 55-71.
GURJANOV, M., LUKATELA, G., LUKATELA, K., SAVIĆ, M., &: TURVEY, M. T. 1985b. Grammatical pr iming of
inflected nouns by the gender of possessive adjectives. Journal of Experimental Psychology: Learning,
memory, and Cognition, Vol 11, No 4,. 692-701.
KOSTIĆ, A. 1995. Informational load constraints on processing inflected morphology. In L. B. Feldman (Ed.),
Morphological Aspects of Language Processing. New Jersey. Lawrence Erlbaum, Inc., Publishers.
KOSTIĆ, A. & HAVELKA, J. 2002. Processing of a verb tense. Psihologija, 35, 3-4, str. 299-316.
KOSTIĆ, A., MARKOVIĆ, T. & BAUCAL, A. 2003. Inflectional morphology and word meaning: orthogonal or coimplicative cognitive domains? u H. Baayen & R. Schreuder (Eds.): Morphological Structure in Language
Processing. Mouton de Gruyter. Berlin. str.1- 45.
Why is Maria read slower than Μαρία?
UiL OTS, Utrecht University
[email protected]
Introduction Consider the following:
(i) John praises, very often, Maria and not his other friends - low inflectional entropy (LH)
(ii) John helps, very often, Maria and not his other friends
- high inflectional entropy (HH)
Processing of the Proper Name (PN) Maria is influenced by the ease of processing of the verb
with which it forms a predicate, namely, Maria in (i) is processed harder than Maria in (ii).
What differentiates praises from helps is their accessibility in long term memory (LTM) as
measured by their value of the inflectional entropy of their verbal paradigm (H), an
information-theoretic measure that quantifies the representational similarity of an inflectional
paradigm in the lexicon and is a function of each verb type’s frequency and syntactic
functions it serves [1]. High inflectional entropy (HH) describes a more “uniform”
distribution of the verb types in the inflectional paradigm that increases the type’s
accessibility. Low inflectional entropy (LH) describes a more “distinct” distribution of the
verb types in the inflectional paradigm that lowers the type’s accessibility. Consequently,
response times of inflected verb types that belong to an HH paradigm (in comprehension) are
accelerated in comparison to inflected verb types that belong to a LH paradigm when the
words are in isolation ([2], [3] a.o.). The effect pertains also within a sentence and since
inflectional entropy is associated with the amount of resources consumed during first lexical
activation it influences, accordingly, the processing time of the object of the sentence as well
[4]. So, in the case of an object like Maria verbs with HH require fewer resources during first
activation and their object is processed faster than verbs with LH, the objects of which are
delayed due to the costly first activation of the verb.
Inflectional entropy is a continuous measure and can also be defined at a between language
level according to the language’s morphological wealth. In fact, rich inflection allows for
distinct forms per syntactic function resulting in a more uniform distribution of the paradigm
so languages with rich morphology (e.g. Greek) have verbs with higher values of inflectional
entropy and lie on the right side of the entropy continuum. In contrast, poor inflection forces
few forms to be used in more functions than others increasing their frequency and,
consequently, languages with poorer morphology (e.g. Dutch) have verbs with lower values
of inflectional entropy and are positioned on the lower end of the entropy continuum This
means that, all other factors kept constant, in a sentence like (i), the verb praises would be
read faster in Greek (‘epeno’) than in Dutch (‘prijzen’), because Greek is an HH language and
Dutch is a LH language.
The present paper studies the hypothesis that the effects of inflectional entropy can be
generalized between languages. More precisely, it will be demonstrated that, indeed, Greek
verbs are retrieved faster than Dutch verbs as a result of different positioning of the language
on the entropy continuum and that crucially, Μαρία is processed faster than Maria as a result
of the higher amount of available resources in Greek.
Experiments Forty-two native Dutch and forty-one native Greek students read sentences like
(i) and (ii) in two self-paced reading experiments. The words between the verb and the PN
were kept constant across items. The PNs in subject and object position were controlled for
frequency and number of syllables. The inflectional entropy of the matrix verb was varied and
the two entropy ranges between languages did not overlap (the highest value of entropy of a
Dutch verb was lower than the lowest value of entropy of a Greek verb).
Reading times in each word were measured and analyzed with linear mixed effects regression
analyses using a 2-stage model, as described in [5]. In the first stage logarithmic transformed
RTs of each word were regressed against the number of the word’s letters, position of the
word in the sentence and position of the item in the task. The residuals of that model were
used as the dependent variable for the second stage and were crossed with language (Dutch vs
Greek) and object type (PN vs reflexive, which served as a baseline). Analyses with entropy
values were also conducted and yielded the same results, because entropy values correspond
(at least for the present experiments) to a language and vice versa. For the sake of simplicity,
the analyses reported are the one with language.
Results Verb retrieval: The main effect of entropy on RTs was significant (t=-2.758, p<.05) at
the region of the verb. Verbs with higher inflectional entropy were read faster than verbs with
lower inflectional entropy.
Object integration: The main effect of object type on RTs was significant at the region of the
object (t=-2.458, p<.05). Sentences with PNs were read significantly slower than those with
reflexives. There was a significant interaction between object type and language (t=2.476,
p<.05). Simple slopes reveal that high entropy language (Greek) accelerates RTs for sentences
with PN (t=-3.342, p<.01).
Discussion Inflectional entropy is tightly connected to the morphological richness of a
language that in turn affects the speed of sentence processing. Processing of a PN requires
introduction of a new entity in the discourse, which is a costly operation. In a morphologically
poor language, like Dutch, that consumes a lot of resources already at the time of the verb’s
retrieval, it is even costlier. The fact that the reflexive condition was processed equally fast in
both languages provides us with a baseline and enables us to conclude that this delay, in
Dutch PN Maria, is a result of the shortage in available resources, as predicted by the position
of Dutch in the inflectional entropy’s continuum.
[1] KOSTIĆ, A. (1995). Information load constraints on processing inflected morphology. In: Morphological
Aspects of Language Processing. Hillsdale, NJ: Erlbaum.
[2] MILIN, P., FILIPOVIĆ ĐURĐEVIĆ, D. & MOSCOSO DEL PRADO MARTÍN, F. (2009). The simultaneous effects of
inflectional paradigms and classes on lexical recognition: Evidence from serbian. Journal of Memory and
Language, 60(1), pp. 50-64.
[3] MOSCOSO DEL PRADO, F., KOSTIĆ, A. & BAAYEN, R. (2004). Putting the bits together: An information
theoretical perspective on morphological processing. Cognition, 94, pp. 1-18.
[4] MANIKA, S., AVRUTIN, S. & REULAND, E. (2014). The effect of verb’s inflectional entropy on the processing
of reflexive objects. In Selected Papers from UK-CLC4, http://www.uk-cla.org.uk/proceedings/volume_2_36.
[5] FINE, A. B., JAEGER, T. F., FARMER, T. A, & QIAN, T. (2013). Rapid expectation adaptation during syntactic
comprehension. PLoS ONE
Inflectional paradigms and classes in sentence reading
P M*† P H*
R. H B*‡
*Eberhard Karls University, Tübingen, Germany
†University of Novi Sad, Serbia ‡University of Alberta, Canada
{petar.milin, petrus.hendrix, harald.baayen}@uni-tuebingen.de
Milin, Filipović Đurđević, & Moscoso del Prado Martín,  showed that the relative entropy of Serbian nouns correlates positively with response latencies in lexical decision, where
the relative entropy operationalized divergence between the distribution of the inflected
variants of a given word from the distribution of case exponents across all words in that
word’s inflectional class. By running two experiments using masked priming paradigm:
lexical decision and self-paced sentence reading, Baayen, Milin, Filipović Đurđević, Hendrix, & Marelli,  replicated the same effect. To account for the influence of the prime,
the authors used a weighted relative entropy.
e present research question is whether the same effect of relative entropy occurs in
normal reading. Two eye-tracking experiments were conducted. Experimental items were
single-line sentences containing two targeted nouns. In the first experiment the two nouns
belonged to the same inflectional class, while in the second experiment inflectional classes
were different.
Results replicated previous findings but also shed new light on the effect of relative entropy in normal reading. On the one hand, the weighted relative entropy measure was predictive when nouns belonged to the same inflectional class. On the other hand, for nouns
belonging to different inflectional classes a simple, unweighted relative entropy measure
e pattern of results suggests that the second noun is discriminated against the first
noun. Additionally, the necessity of a weight for nouns belonging to the same inflectional
class suggests that in this case more cognitive effort is required to discriminate between the
two words.
Baayen, R., Milin, P., Filipović Đurđević, D., Hendrix, P., & Marelli, M. (). An amorphous model
for morphological processing in visual comprehension based on naive discriminative learning.
Psychological Review, (), –.
Milin, P., Filipović Đurđević, D., & Moscoso del Prado Martín, F. (). e simultaneous effects
of inflectional paradigms and classes on lexical recognition: Evidence from Serbian. Journal
of Memory and Language, (), –.
A functional take on Estonian inflectional paradigms
University of California, Santa Barbara
[email protected]
Inflectional richness is often taken as an indication of linguistic "complexity" (e.g.,
McWhorter, 2011). However, from an informational-theoretical standpoint, it has been noted
that the paradigms of morphologically rich languages are perhaps not all that complex (e.g.,
Blevins, Malouf & Ackerman, 2019; Malouf & Ackerman, 2013). Whereas the view of
inflectional morphology assumes that it is difficult to know how to fill the cells in a paradigm
in an inflectionally rich language, the latter approches show that a few forms in the paradigm,
the classical principal parts, are strongly informative about the rest, and therefore substatially
reduce the actual difficulty or complexity that is entailed by the paradigm. This has led some
researchers to postulate a "low entropy conjecture" (e.g., Malouf & Ackerman, 2013): In
terms of their morphological complexity languages differ less than one might expect.
In this study, I follow this "low entropy conjecture" one step further. One needs to ask the
question of what is the functional role inflectional classes in the first place. Although it is
relatively accepted that noun classes, such as gender, can have a semantic, and thus functional
component to them, it is generally thought that the inflectional classes themselves are of little
semantic relevance. It seems however rather illogical that a system subject to such
evolutionary pressure as is language would indulge in gratuitous levels of complexity in a
systematic fashion. Rather, I show that the variation observed in the forms filling the same
paradigmatic cell has itself a functional justification in reducing semantic uncertainty.
I present a detailed, corpus-based, information theoretical analysis of the Estonian nominal
system. Firstly, following Malouf and Ackerman (2013), I show that the principal parts of
Estonian paradigms can be recovered from corpus data (and I provide a new algorithm to
achieve this automatically). I follow by investigating how the choice of inflected variant to fill
in a paradigm cell is in itself informative about the noun's semantics. This is in slight
contradiction with Blevins (2008)'s statement that "classes are cued by variation in form,
without evident syntactic or semantic correlates". Finally, using word co-occurrence statistics,
I provide a detailed account of mutual implication relations between word meaning,
inflectional class, and inflectional cell in all possible pairings. With this I show that, rather
than orthogonal, these three factors, together with the stem itself, are strongly mutually
informative. The consequence of the mutual informations described above is that, at the end,
the choice of form to fill a paradigm cell, the choice of meaning associated with an inflected
form, or the choice of case for a particular word, displays surprisingly little uncertainty (i.e.,
entropy), in line with the predictions of the low entropy conjecture.
Selected References
• JAMES P. BLEVINS (2008) Declension classes in Estonian. Linguistica Uralica XLIV, 241–267.
• JAMES P. BLEVINS, FARRELL ACKERMAN and ROBERT MALOUF (2009) Parts and wholes: Implicative
patterns in inflectional paradigms. In JAMES P. BLEVINS and JULIETTE BLEVINS (eds.), Analogy in
Grammar: Form and Acquisition. Oxford University Press, 54–81.
• FARRELL ACKERMAN and ROBERT MALOUF. 2013. Morphological organization: The low conditional
entropy conjecture. Language. 89:429-464.
The possible role of Entropy in processing argument dependencies in
Eszterházy College, Eger Budapest U of Technology and Economics
[email protected] [email protected] [email protected]
It is a commonplace of research in sentence processing for at least 40 years
now, since the beginnings of the use of Thematic Roles and different versions
of Frame Theories in processing studies, to find facilitative effects between
verbs and their arguments (Tanenhaus et al, 1995, Kintsch, 1998). Thus
reading CUT would facilitate the reading of instrumental arguments. In a less
trivial way, reading MET or PLAYED also would facilitate reading
instrumentals. This is a rather crucial issue in sentence processing, to the
extent that Bornkessel & Schlesewsky (2006) even developed a full-fledged
crosslinguistic theory of the temporal activation of the verbal argument
frames and the insertion of noun phrases into the slots based on neuronal
processing evidence.
In our research we concentrate on oblique arguments of Hungarian verbs
exploiting the fact that the argument relations are coded by case markers in
Hungarian. Therefore, argument processing is closely tied with
morphological processing. As a matter of fact most of the arguments with
abstract relations are coded by locative case markers as illustrated in the table.
ambiguous in marking
Emlékszik a fiúra
Haragszik a tanítóra.
Készül a versenyre.
Találkozott a lánnyal.
Csókolózott a rendőrrel.
Gondol a lányra.
Gondol valamit.
Gondolkodik a lányon.
Gondolkodik a hajón.
remembers the boy.ON
angers the teacher.ON
Prepares the race.ON
Met the girls.INST
Kissed the cop.INST
thinks girl.ON
thinks something.ACC
thinks girl.ON
thinks boat.ON
Earlier studies on the processing of argument structures in Hungarian have
shown that these assumed interactions between morphology and sentence
processing do indeed hold. Gervain and Pléh (2003), for example, showed that
prenominal verbs facilitate the processing of constructions like ‘Anna thought
of the boat’, and postverbal nouns that are ambiguous between a locativeadjunct and an abstract-argument reading are read slower that arguments.
Compare ‘Anna RUMINATED on the boat’ versus ‘Anna RUMINATED on
the problem’, where in the later case the locative meaning is excluded. Fekete
and Pléh (2011) had shown the relevance of the argument-based approach to
psycholinguistic processing (experimenting in Hungarian with bidirectional
comitative constructions like John was kissing with Mary), compared with
unidirectional comitative such as ‘John was messing with Mary’. They
observed that there was a difference in the processing of anaphors referring
to the Subject. Anaphors in THE singular (He was…) were processed slower
after bidirectional antecedents, showing a ’deep anaphora effect’ in the sense
introduced by Hankamer. & Sag (1976).
All of these studies did try to control for frequency effects. Case ending and
semantic interpretation ambiguities, however, were treated by them in a
categorical manner. In the studies to be presented at the workshop we shall
concentrate on the statistical structure of the possible ambiguities.
1. We would combine the structures illustrated in the table with a consideration
of entropy measures. From the MOKK (2006) and the MAZSOLA corpus
(Sas, 2008) entropy computations would be made for the relations between
a given verb and the noun endings in a ‘plus/minus two content words’
frame. These distributions will be compared to the theoretical propositions.
2. These entropy estimates will be used as predictors in self-paced word by
word reading time experiments for the reading of nouns in sentence
contexts. In this we way we would like to learn whether the entropy relations
between verbs and case endings do have an explanatory power in processing
and verb-noun attachments.
Bornkessel, I. & Schlesewsky, M. (2006) The extended argument dependency model: A neurocognitive approach
to sentence comprehension across languages. Psychological Review, Vol 113(4), 787-821 • Fekete, I. & Pléh, Cs.
(2011) Bidirectional and unidirectional comitative constructions in Hungarian: A psycholinguistic investigation at
the interface of argument structure and semantics. Acta Linguistica Hungarica, Vol. 58 (1–2), pp. 3–23 • Gervain,
J. & Pléh, C. (2004). [Anna is thinking on the ship, or argument expectancies in Hungarian sentence
understanding]. In: Judit Gervain, J. & Pléh, C. (eds): A láthatatlan megismerés, 112–25. Gondolat, Budapest. •
Hankamer, J. & Sag. I. 1976. Deep and surface anaphora. Linguistic Inquiry 7: 391–428. • Kintsch,W. 1998.
Comprehension: a paradigm for cognition. Cambridge University Press, Cambridge. • MOKK (2006): A WEB
based Hungarian frequency dictionary: http://szotar.mokk.bme.hu/szoszablya/searchq.php • Sass, B. (2008): The
Verb Argument Browser. In: Sojka, P. et al. (eds.): Proceedings of TSD 2008, Brno, Czech Republic, , LNCS
5246, 187-192,http://corpus.nytud.hu/vab • Tanenhaus,M.K.,Spivey-Knowlton,M. J.,Eberhard,K., & Sedivy,J.C.
(1995). Integration of visual and linguistic information in spoken language comprehension. Science 268, 632–634.
Implicative organization facilitates morphological learning
University of California, San Diego and San Diego State University
[email protected], [email protected], [email protected]
In word-based morphology, implicative relationships among related wordforms are
used to facilitate the learning of complex morphology (see Blevins, to appear, ch. 7;
Ackerman & Malouf 2013). For example, an Italian singular suffix -a (persona) typically
implies a plural suffix -e (persone), while singular suffix -o (gatto) implies plural suffix -i
(gatti), and vice versa. On this basis, a learner might correctly predict that a novel singular
form rosa has the plural form rose. However, this perspective on morphological organization
is challenged by experimental results in artificial language learning, which find that learners
are unable to acquire these relationships unless they are supported by additional phonological
cues in the stem (Gerken et al. 2009; Brooks et al. 1993, Frigo & MacDonald 1998). We argue
that these results were biased by ecological factors, and that implicative relations alone are
sufficient to enhance paradigm learnability.
We first identify four methodological features of prior experiments that restrict general
learnability: (1) passive exposure and/or rote repetition, rather than active trial-and-error
learning; (2) a requirement to learn both novel lexemes and inflections in a short time; (3) a
requirement to infer relations among abstract linguistic labels rather than referents; (4)
presentation of labels before their referents, which is known to delay association learning
(Ramscar et al. 2010).
We present three novel artificial grammar experiments in which subjects demonstrate
knowledge of formal paradigmatic relationships, despite experimental randomization designed
to avoid redundant phonological or semantic information that would signal class membership.
In Experiment 1, subjects attempted to learn how to inflect familiar nouns for number
(singular, dual, plural) in an alien language. Six nouns were randomly selected from a pool of
30 household objects and divided into two classes which had suffix inflections as shown in
Figure 1. In each training trial, subjects saw one, two, or many pictures of a noun and tried to
guess how the aliens would refer to that set. Subjects were then given feedback and shown the
correct answer.
After 90 training trials, subjects were tested on their knowledge. In each testing trial,
subjects first saw one, two, or many pictures of a new, previously-unseen object, and were
given its label. Subjects were then asked to label a different number of the same noun. In
critical trials, the new noun was first presented in a form that reliably predicted the test form
(e.g., singular -taf implies dual -guk). The results showed that subjects successfully took
advantage of implicative relationships: when subjects could use two compatible suffixes for
the second inflection, performance was significantly better when they were first presented
with a predictive inflection than when they were not (p < 0.001). Experiments 2-3 tested more
complex paradigms with 9 nouns and 3 classes. In these paradigms, performance was better
only for bidirectional relationships, where the given and unseen suffixes both implied each
other (p < 0.01).
Our results indicate that subjects acquired paradigmatic relations without redundant
cues. This finding suggests that ecological factors drove previous results, which helps resolve
the apparent mismatch between prior experimental and typological data. Some researchers
have argued that enriching word classes with additional syntactic or semantic redundancies
could permit acquisition if phonological cues are inadequate (Ouyang, Boroditsky, & Frank
2012; Mintz 2002; Braine 1987). However, our results indicate that learners can acquire
useful paradigmatic relations without any additional supporting cues if the potential influence
of domain-general learning factors is properly taken into account (Frank & Gibson 2011).
ACKERMAN, F., & MALOUF, R. (2013). Morphological Organization: The Low Conditional Entropy Conjecture.
Language, 89(3), 429–464. doi:10.1353/lan.2013.0054
BLEVINS, J. (to appear). Word and Paradigm Morphology. Oxford University Press.
BRAINE, M. (1987). What is learned in acquiring word classes: a step toward an acquisition theory. In
Mechanisms of Language Acquisition (pp. 65–87). Hillsdale, NJ: Lawrence Erlbaum.
BROOKS, P., BRAINE, M., CATALANO, L., & BRODY, R. (1993). Acquisition of gender-like noun subclasses in an
artificial language: the contribution of phonological markers to learning. Journal of Memory and Language, 32,
FRANK, M. C. & GIBSON, E. (2011). Overcoming memory limitations in rule learning. Language Learning and
Development, 7(2), 130–148
FRIGO, L., & MACDONALD, J. (1998). Properties of phonological markers that affect the acquisition of genderlike subclasses. Journal of Memory and Language, 39, 218–245.
GERKEN, L., WILSON, R., GÓMEZ, R. L., & NURMSOO, E. (2009). The relation between linguistic analogies and
lexical categories. In Analogy in Grammar: Form and Acquisition. Oxford University Press.
MINTZ, T. H. (2002). Category induction from distributional cues in an artificial language. Memory & Cognition,
30(5), 678–686.
OUYANG, L., BORODITSKY, L., & FRANK, M. C. (2012). Semantic Coherence Facilitates Distributional Learning
of Word Meanings. In Proceedings of the 34th Annual Conference of the Cognitive Science Society.
RAMSCAR, M., YARLETT, D., DYE, M., DENNY, K., & THORPE, K. (2010). The effects of feature-label-order and
their implications for symbolic learning. Cognitive Science, 34(6), 909–957. doi:10.1111/j.15516709.2009.01092.x
Structural attraction in Croatian:
Effects of inflectional paradigmatic structure in morphosyntactic processing
The Ohio State University
[email protected]
Structural attraction is when a target agrees with a local controller, rather than the proper but
more distant one: How much correctionSG of syntactic errorsPL arePL there, anyway? It is widely
considered to be a locality effect in morphosyntactic processing. Attraction effects are relatively
rare in case-marking languages (Nicol and Wilson 2000), but it is unclear whether this derives
from shallow processing of the noun’s (non-nominative) form, or reflects deeper processing of
morphological feature structure. Croatian offers the opportunity to tease apart these issues.
Attraction effects in Croatian are sensitive to syncretism: attraction is found (rarely) when
non-nominative local controllers are homophonous with nominative; see (1), in which the verb
tjerala su seems to agree with the local accusative NP ova plemena, rather the conjoined
structural subjects. Importantly, ova plemena exhibits systematic nominative-accusative
syncretism, so in a local context it could be construed as a viable controller for the verb.
(1) Vječiti
od uvijek mogućih napada…
perpetual wars(M).NOMPL and dangers(F).NOMPL from always possible attacks
u saveze.
drove.N.PL AUX.3PL these.N.ACCPL tribes(N).ACCPL into alliances
‘Perpetual wars(M).NOMPL and dangers(F)NOM.PL from ever-possible attacks …, droveN.3PL these
tribes(N).ACCPL into alliances.’ (http://hr.wikipedia.org/wiki/Indijanci, March 2008)
Similar facilitation has been elicited through forced production in Slovak (Badecker and
Kuminiak 2007) and German and Dutch (Hartsuiker et al. 2003).
Studies of this syncretism effect in structural attraction tend to assume that it reflects
shallow processing of the morphological form of the local controller, but they fail to distinguish
between superficial homonymy and systematic syncretism, the latter being a deep fact of
morphological structure. This leaves open the question: Does homophony that reflects
morphological organization (syncretism) facilitate structural attraction more than accidental
homophony? We might expect this given evidence that the processing of inflected forms is
sensitive to the amount of information they convey (Milin et al. 2009; Moscoso del Prado Martín
et al. 2004). We might also predict this based on resolution of morphosyntactic feature conflict
by syncretism in which the systematicity of the syncretism matters, e.g., German double
agreement (Groos and van Riemsdijk 1981).
This paper presents a way to quantify accidental vs. systematic syncretism using
information-theoretic measures, and uses it to predict structural attraction effects in Croatian. I
show experimentally that some patterns of syncretism in Croatian (e.g, nom pl - acc pl) induce
attraction effects more than other patterns do (e.g., nom pl - gen sg), and I hypothesize that this
difference is explained in terms of the information that a syncretic form conveys about the
morphosyntactic values it expresses. I quantify the information a given Croatian inflected form
conveys in terms of Shannon entropy. (Since syncretic forms are morphosyntactically
ambiguous, there is on average more surprisal associated with the morphosyntactic values of a
syncretic form than a non-syncretic form.) I also ask the extent to which the distribution of
syncretic forms in a given paradigm is shared across inflection classes and word classes, and
quantify this in terms of Hamming distance calculated over nearest neighbor distributions. In
combination, the two measures offer a way to quantify the extent of syncretism within a
paradigm, and more importantly, the degree of systematicity of a pattern of syncretism.
The results suggest that morphosyntactic processing is sensitive to the difference between
accidental and systematic inflectional homophony in the morphological system. Morphosyntactic
processing and agreement patterns in Croatian thus shed light on the intricate structuring of
inflectional paradigms and the cognitive processing mechanisms underlying structural attraction
BADECKER, W., and F. KUMINIAK. 2007. "Morphology, agreement and working memory retrieval in sentence
production: Evidence from gender and case in Slovak." Journal of memory and language 56(1): 65-85.
GROOS, A., and H. VAN RIEMSDIJK. 1981. "Matching effect in free relatives: A parameter of core grammar." In The
theory of markedness in generative grammar: Proceedings of the 1979 GLOW conference, edited by A.
BELLETTI, L. BRANDI and L. RIZZI, 171-216. Pisa: Scuola Normale Superiore di Pisa.
HARTSUIKER, R., H. SCHRIEFERS, K. BOCK, and GERDIEN M. KIKSTRA. 2003. "Morphophonological influences on
the construction of subject-verb agreement." Memory and cognition 31(8): 1316-1326.
MILIN, P., V. KUPERMAN, A. KOSTIĆ, and R. H. BAAYEN. 2009. "Words and paradigms bit by bit: An informationtheoretic approach to the processing of inflection and derivation." In Analogy in grammar: Form and
acquisition, edited by J. P. BLEVINS and J. BLEVINS. Oxford: Oxford University Press.
MOSCOSO DEL PRADO MARTÍN, F., A. KOSTIĆ, and R. H. BAAYEN. 2004. "Putting the bits together: An information
theoretical perspective on morphological processing." Cognition 94: 1-18.
NICOL, J., and R. WILSON. 2000. "Agreement and case-marking in Russian: A psycholinguistic investigation of
agreement errors in production." In Proceedings of the Annual Workshop on Formal Approaches to Slavic
Linguistics: The Philadelphia Meeting 1999, edited by T. HOLLOWAY KING and I. SEKERINA, 314-327. Ann
Arbor: Michigan Slavic Publications.
Diversity and relative entropy as indices of relational competition in the processing of
English compounds
Department of Psychology, University of Alberta
[email protected] [email protected]
Recent research has indicated that processing compound words involves an attempt at semantic composition of the constituent words (Fiorentino & Poeppel 2007; Kounios et al. 2003;
Koester et al. 2007), and that this meaning construction process involves an attempt to identify a semantic relation linking the constituents (Gagné & Spalding 2009; Ji, et al. 2011). The
current project investigates relational competition during the processing of transparent and
opaque English compounds, using an information theoretic approach. In particular, we examine the impact of relational diversity, which reflects the extent to which the possible interpretations concentrate on a small set of relations, and relational relative entropy, which measures
the relational information gain associated with the particular compound (i.e., relational relative entropy measures how much the relational distribution for a specific item differs from the
distribution based on all items). These measures are derived from the possible relations task
(Gagné & Spalding 2014), which generates, for each item, a distribution of possible relational
interpretations based on participant responses to a question in which they are asked to indicate the most likely literal meaning for a pair of words (e.g., hog wash).
We found that although both semantically opaque and semantically transparent compounds
have morphemic structure, the response latencies in a lexical decision task for these items are
differentially affected by diversity and entropy. Diversity interacted with compound frequency for transparent compounds but not for opaque compounds. Also, when information from
the compound leads to a particular meaning (i.e., when there is low diversity and high entropy), this convergence resulted in faster processing for transparent compounds, but slower
processing for opaque compounds.
These results suggest that relation-based interpretations are computed whenever morphological constituents (and their corresponding semantic/conceptual representations) become available, even in situations where these meanings are incompatible with the established meaning
and, thus, must be rejected.!
FIORENTINO, R., & POEPPEL, D. (2007). Compound words and structure in the lexicon. Language and Cognitive
Processes, 22, 953-1000.
GAGNÉ, C. L., & SPALDING, T. L. (2014). Relation diversity and ease of processing for opaque and transparent
English compounds. In F. Rainer, Dressler, W., Gardini, F., Luschützky, H.C. (Eds.). Morphology and
Meaning: Current Issues in Linguistic Theory (pp. 153-162). Amsterdam: John Benjamins.
GAGNÉ, C. L., & SPALDING, T. L. (2009). Constituent integration during the processing of compound words:
Does it involve the use of relational structures? Journal of Memory and Language, 60, 20-35.
JI, H., GAGNÉ, C. L., & SPALDING, T. L. (2011). Benefits and costs of lexical decomposition and semantic integration during the processing of transparent and opaque English compounds. Journal of Memory and Language, 65, 406-430.
KOESTER, D., GUNTER, T. C., & WAGNER, S. (2007). The morphosyntactic decomposition and semantic composition of german compound words investigated by ERPs. Brain and Language, 102, 64-79.
KOUNIOS, J., BACHMAN, P., CASASANTO, D., GROSSMAN, M., SMITH, R. W., & YANG, W. (2003). Novel concepts mediate word retrieval from human episodic associative memory: Evidence from event-related potentials. Neuroscience Letters, 345, 157-60.
Synchrony and Diachrony of Inflectional Classes:
Theoretical and Empirical Considerations
Petros Karatsareas (Bristol/Nicosia)
Enrique L. Palancar (Paris/Surrey)
Timothy Feist (Surrey)
The inflectional classes of the Rtau verb
A A
G J
, , , 
[email protected]
[email protected]
This paper deals with the verbal inflectional classes of Rtau (locally known as rəsɲəske), a Rgyalrongic language spoken in Rtau country, Sichuan province, China. The data presented here is
based on ongoing fieldwork by the authors.
In Rtau, there are two main verb classes that we can call intransitive and transitive, respectively.
Verbs belonging to the intransitive class have at most two different forms, one with a first person
(singular or plural) reference, the other with a non-first (second or third) person reference. The
first person form has only a limited array of possible rhymes: we never find front or central
vowels, but only open syllable nasal rhymes –ã and –õ, velarized vowels –oˠ and –aˠ or the back
rounded –u. In the non-first person (henceforth 2/3) form however, almost all possible rhymes are
attested, including open and closed syllables (except for non-velarized –o). Vowel alternations
(or the lack thereof) allow us to posit the existence of six inflectional classes for verbs ending
in open syllables, with class 6 including verbs exhibiting no alternations, whose rhyme can be
any of –u, –oˠ, –aˠ, –õ and –ã (Table 1). The alternations can be stated in a straightforward way:
centralized vowels –ə and –ɞ change to –õ, and front and open (unrounded and non-velarized)
vowels change to –ã. In the case of verb stems ending in–r or –v, the first person is always
derived by replacing the entire rhyme by –ã or –õ depending on the main vowel of the rhyme.
Stems ending in –m (the only other final consonant available) are always Tibetan loanwords and
do not present any alternation.
Verbs belonging to the transitive class have at most six different forms, illustrated by the perfective (where the verb is preceded by a directional prefix, which can be neglected for the purposes
of this presentation) paradigm of the verb ‘to kill’ (cf. Table 2). If we disregard the inverse
prefix v–/f–, there are four distinct forms: 1→3; 2→3; 1→3 and 2/3→1 (with the inverse
prefix); and 1→2 and 3→2/3 (with the inverse prefix). As with intransitive verbs, vowel alternations allow us to posit six inflectional classes, depending on the final vowel of the verb stem.
Table 3 presents all six classes (verb forms are shown without the inverse prefix v–/f–.) Class 6
includes all verbs with stems ending in –u, –oˠ, –aˠ, –õ and –ã. In the case of verb stems ending
in a closed syllable, final –v drops in first person forms; in the 2→3 form it stays but there is
vowel fronting as in –zgriv ‘you accomplished’; final –r drops in the first and second person but
is preserved in the third person and in 1→2 forms while final –m is immune to any changes and
verbs ending in this consonant present no stem alternations.
A closer look reveals that in the case of intransitive verbs the first person form is generally
predictable from the 2/3 form except in a few isolated counterexamples. It is therefore possible
to analyze the 2/3 form as the base form and the first person form as being derived from it by
a set of morpho-phonological rules involving fusion with a suffix –ã, which is realized as –õ
when the rhyme is centralized. In the case of transitive verbs, it is possible to consider the third
person form as the base form; the 1→3 and 2/3→1 stems can be analyzed as resulting from
fusion with the first person –ã suffix. The 1→3 form presents rounding of the vowels with an
additional –w glide in the case of mid-low and low vowels. These alternations can be accounted
for by assuming the existence of a suffix whose underlying form is –w. The 2→3 form has vowel
fronting with an additional –j glide for mid-low and low vowels. Here the underlying form –j can
be posited. All these morphophonological rules are summarized in Table 5. A further support
for positing these vowel fusion rules is the fact that they are not restricted to the verbal system,
but also apply to the ergative –w and genitive –j case markers. Table 6 illustrates some examples
of vowel fusion in nouns.
Table 1: Vowel alternations in open-syllable intransitive verbs in Rtau
look at
be full
be ill
be hot
Table 2: fse ‘to kill’
Table 3: Vowel alternations in open-syllable transitive verbs in Rtau
dress up
1→3, 2/3→1
3→3, 1→2
Table 4: Vowel alternations in closed syllable transitive verbs in Rtau
give back
1→3, 2/3→1
Table 5: Vowel fusion in Rtau verbs
PP Suffix
1→3 –w
1 –ã
2→3 –j
Table 6: Vowel fusion in Rtau nouns
base form
hybrid of yak and cow
Reaching out beyond the wordform: Periphrasis and inflectional classes
Surrey Morphology Group, University of Surrey
[email protected]
Surrey Morphology Group, University of Surrey
[email protected], [email protected]
It is common knowledge that inflectional classes are built on inflectional distinctions based
on affixal allomorphy. This understanding is in accord with a view of inflectional
morphology relying on the concept of a morphological word that consists of a phonological
word with morphological internal structure. But across grammatical systems, it is also
common to find periphrastic forms conveying the same kind of inflectional information
realized in wordforms, suggesting that such periphrastic forms could be treated as inflectional
and be thus part of the paradigm (e.g. gradient forms of adjectives in English (more/most
interesting vs. bigger/biggest); the perfect of the passive in Latin (amatus est ‘s/he was loved’
vs. amavit ‘s/he loved’). Against the traditional view that periphrasis is a purely syntactic
phenomenon, approaches to inflection from autonomous morphology or from inferentialrealizational models consider periphrasis as a morphological phenomenon (e.g. Ackerman
and Stump 2004, Vicent 2011, Chumakina & Corbett 2012, etc.).
In this paper, we want to extend this view of periphrasis one step further to cover inflectional
classes too. We claim that the morphological information stored in a lexeme for the selection
of inflectional allomorphs can have scope beyond the phonological word and reach over to
the non-lexical component of a periphrastic expression. We illustrate this with two case
studies on verb inflection in sufficient detail.
One case is the inflectional system of verbs in Otomi; a family of languages spoken in
Mexico that belongs to the Oto-Pamean branch of Oto-Manguean. Verbs in Otomi inflect for
TAM values by means of markers that we call ‘inflectional formatives’ and that often realize
person of subject too. Example (1) illustrates the formative tú of the 1st p. completive realis in
Tilapa Otomi. Otomi inflectional formatives are clitics, e.g. they are obligatorily hosted as
enclitics on certain functional heads, like the negative in (2), or more facultatively to any
preceding phrase, allowing hesitation pauses like in (3).
‘we wanted to follow them.’ [txt]
pe hín=dú
‘but we didn't want.’ [txt]
then=1.CPL.REALIS return=PL.EXCL
‘Then we...returned.’ [txt]
Examples like (2-3) indicate that the entire inflection of Otomi verbs is periphrastic. This
system probably emerged from the grammaticalization of old auxiliary verbs which are
beyond reconstruction at this stage and no longer have a synchronic correlate, making this a
non-canonical type of periphrasis. What is interesting for a theory of morphology is that
verbs of some Otomi languages fall into different inflectional classes because they select
different inflectional formatives for the realization of a set of TAM values in the paradigm.
Examples (4) show this with the subparadigm of completive irrealis, and they serve to show
how allomorphic selection by the verb reaches out beyond the wordform.
Class I
gu ho
gi syo
ta syo
Class II
gutu hi
gugu hi
Class III
giti hóhki
giti hóhki
ti hóhki
Otomi naturally brings us to better-known instances of auxiliary selection that have,
nonetheless, been treated in the literature as motivated by the syntax-semantics interface. For
example, French verbs can be said to fall into two inflectional classes attending to whether
they select the auxiliaries etre ‘be’ or avoir ‘have’ for their past tense. Both classes are openended, but only the former has intransitive verbs -known as ‘inaccusatives’. The class
includes all reflexive verbs together with a small subset of (mainly motion) verbs and their
lexical derivatives. While this class has a semantic core (cf. Mithun 1991), the semantics no
longer helps predicting the verbs of the subset, and they must be learned (i.e. mourir ‘die’
selects etre but expirer ‘expire, die’ or périr ‘perish’ select avoir, similarly monter ‘go up’
selects etre but grossir ‘grow’ or pouser ‘grow (plants)’ select avoir, etc.).
ACKERMAN, FARRELL AND STUMP, GREGORY. 2004. Paradigms and periphrastic expression: a study in
realization-based lexicalism, in Louisa Sadler and Andrew Spencer (eds.), Projecting Morphology, 11158. Stanford, CA: Center for the Study of Language and Information.
CHUMAKINA, MARINA AND GREVILLE G. CORBETT (eds.). 2012. Periphrasis: The Role of Syntax and
Morphology in Paradigms [Proceedings of the British Academy 180. Oxford: OUP/British Academy.
LOPORCARO, MICHELE. 2007. On triple auxiliation in Romance. Linguistics 45.1. 173-222
MITHUN, MARIANNE.1991. Active / agentive case marking and its motivation. Language 67:3, 510-546.
VINCENT, NIGEL. 2011. Non-finite forms, periphrases, and autonomous morphology in Latin and Romance, in
Maria Goldbach, Marc-Olivier Hinzelin, Martin Maiden and John Charles Smith (eds.), Morphological
Autonomy: Perspectives from Romance Inflectional Morphology, 421-439. Oxford: OUP.
Potential and actual variation in morphosyntax
University of Sheffield
[email protected]
This talk will explore variation within two morphosyntactic ‘slots’ of a traditional
inflectional class, and question what that says about the description of this class and its
historical evolution. Although it comes from an emergentist perspective on language,
emergentist enquiry into the structure of language has tended to focus on syntactic variation
(ditransitivity, transitive-causative constructions, adjective suitability, phrasal verbs, see e.g.
Brooks & Tomasello 1999, Ambridge et al. 2008, Boyd & Goldberg 2011). Such data
naturally come to the fore in English, but data from other, morphologically rich languages
present an opportunity to examine choice and variation in slots where ‘overdifferentiation’
has taken place (as in Brown 2007), but the number of variants is very limited and their form
Our starting point is new data from Czech, a language which offers an interesting view into
the nature of paradigmatic relations. Inflection in Czech generally falls into the pattern
described in the workshop call, where sets of morphosyntactic/morphosemantic features are
expressed by different formants across classes of lexical items whose membership is in part
arbitrary. In addition, however, within these inflectional classes, variation or competition
between formants exists in multiple functional slots in paradigms, and the appearance of this
variation also seems unpredictable. Štícha (2009), an extensive survey of one of these places,
notes that while some general phonological and derivational tendencies in its usage can be
observed, hard-and-fast rules dictating the use of one or another formant are absent, and
diachronic developments also show the expansion and contraction of formant usage moving
in mutually contradictory directions.
Our data on variation in the genitive and locative singular of the so-called ‘masculine hard
inanimate’ class of nouns have been collected through trawls in 100-million-token corpora of
written Czech and through large-scale (N=550) experiments probing native speakers’
acceptability judgments of particular forms and their production of them in written forcedchoice tasks. In each of our two cases, there is a historically innovative formant {u} that is
highly productive and forms a marker of this declension class, which competes with a
‘conservative’ formant {a} or {ě} that is far less productive, although not unproductive, and
also serves as a marker of this case in other classes. Some lexemes appear in the corpus
exclusively with one of the two formants (either is possible), while a limited number of highfrequency nouns appear with both formants.
For example, in the SYN2005 corpus, we find lexemes with no (or almost no) variation:
hrad ‘castle’: gen. hradu (n=2602), not *hrada;
les ‘forest’: gen. lesa (n=4336), not *lesu (80 lexemes with this pattern);
zámek ‘stately home’: loc. zámku (n=2078), not *zámce;
prales ‘primeval forest’: loc. pralese (n=192), not *pralesu (isolated lexemes with this
In the SYN2005 corpus, variation in the genitive singular slot of this paradigm is found with
112 relatively common lexemes and in the locative singular slot we find variation with 392
lexemes, e.g.:
rok ‘year’: gen. roku (n=71041) or roka (n=905);
betlém ‘creche’: gen. betlému (n=33) or betléma (n=106)
hrad ‘castle’: loc. hradu (n=122) or hradě (n=1518);
les ‘forest’: loc. lesu (n=5) or lese (n=2845)
Bermel/Knittl/Russell, p. 2
Initial statistical analyses of our experimental data suggest that native-speaker judgments
and choices are sensitive to observed frequencies (Bader and Häussler 2009, Divjak 2008,
Kempen and Harbusch 2008), and the degree of this sensitivity is for forced-choice tasks quite
notable: a regression analysis using primarily these observed frequencies as factors shows
significantly improved predictions vis-à-vis the null model (R2 = 0.30 to 0.68) However, in
contrast to what might be expected from the literature on language acquisition, items with low
(but not zero) relative frequency continue to be produced at low (but not zero) rates and to
enjoy at least middling acceptability, cp. Arppe and Järvikivi 2007. There is some further
evidence for the existence of ‘sub-cases’ distinct to this inflectional class, which have a
morphosyntactic or morphosemantic basis and seem to play a role in maintenance of the lesscommon formant. This is shown in the fact that the data from the two cases under
investigation have different profiles, with the genitive more sensitive to frequency-based
effects and regional variation, while the locative shows more influence of the form’s syntactic
Our findings raise some questions about the way in which concepts from language
acquisition have been used to model language change, i.e. the developments within the
language system shared by a community of (primarily adult native) language users. Some
concepts, such as entrenchment, appear to transfer well from one domain to another, while
others, such as pre-emption (inter alia Boyd & Goldberg 2011), may need further
modification to adequately describe what is happening in morphosyntactic systems.
The data seem to point not to the rationalisation of an inflectional class system as a dead
weight, but to its potential for on-going reanalysis and even elaboration. This might explain
why, in contrast to Brown’s (2007) description of the atrophying of a once live instance of
overdifferentiation, Czech presents a relatively enduring picture of morphological variation in
a similar paradigm.
Touches on the following issues from the CFP: synchronic motivation and function of
inflectional class systems; possible relationships of inflectional class systems with other
grammatical features (phonological, syntactic, semantic); position of inflectional class
features in the grammar; diachronic emergence and decay of inflectional class systems;
acquisition of inflectional class systems.
AMBRIDGE, B. et al. 2008. The effect of verb semantic class and verb frequency (entrenchment) on children’s
and adults’ graded judgements of argument-structure overgeneralization errors. Cognition 106: 87–129.
ARPPE, A. & J. JÄRVIKIVI. 2007. Take empiricism seriously! In support of methodological diversity in linguistics.
Corpus linguistics and linguistic theory 3 (1). 99–109.
BADER, M. & J. HÄUSSLER. 2009. Toward a model of grammaticality judgments. Journal of Linguistics 45. 1–
BOYD, J. & A. GOLDBERG. 2011. Learning what not to say: The role of statistical preemption and categorization
in a-adjective production. Language 87: 55–83.
BROOKS, P. & M. TOMASELLO. 1999. How children constrain their argument structure constructions. Language
75: 720–738.
BROWN, D.. 2007. Peripheral functions and overdifferentiation: The Russian second locative. Russian
Linguistics 31: 61–76.
CZECH NATIONAL CORPUS. Ústav Českého národního korpusu FF UK, Prague 2000–2010. Available on the
web: <http://www.korpus.cz>.
DIVJAK, D.. 2008. On (in)frequency and (un)acceptability. In Barbara Lewandowska-Tomaszczyk (ed.), Corpus
linguistics, computer tools and applications – State of the art, 213–233. Frankfurt: Peter Lang.
KEMPEN, G. & K. HARBUSCH. 2008. Comparing linguistic judgments and corpus frequencies as windows on
grammatical competence: A study of argument linearization in German clauses. In Anita Steube (ed.), The
discourse potential of underspecified structures, 179–192. Berlin: Walter de Gruyter.
ŠTÍCHA, F. 2009. Lokál singuláru tvrdých neživotných maskulin (ve vlaku vs. v potoce): úzus a gramatičnost.
Slovo a slovesnost 70: 193–220,
Third-person realis morphology and verb classes in the Zamucoan languages
Scuola Normale Superiore di Pisa
[email protected]
The Zamucoan family only includes two living languages (Ayoreo and Chamacoco) spoken in
the Northern Chaco, between Bolivia and Paraguay. Both languages are endangered and are
spoken by approximately 4500 and 1600 people, respectively. The linguistic family also
includes an extinct language, Ancient Zamuco, described in the XVIII century by the Jesuit
Father Ignace Chomé.
In the Zamucoan languages the verb expresses neither tense nor aspect, but it still presents
a realis / irrealis opposition, although this only shows a defective paradigm in the currently
spoken languages. The verb has prefixes indicating the person category and suffixes
indicating the number category (specifically, plural). Verb morphology allows the
identification of different paradigms. The most conspicuous identifying feature for the various
paradigms consists in the morphology of the 3-person realis (3-realis), which is characterized
by remarkable morphological variation (Ciucci 2013).
The aim of this paper is precisely to show how, in these tenseless languages, the 3-realis
polymorphism has led to the emergence of inflectional classes. Table 1 illustrates the verb
classification of the Zamucoan languages, as based on the 3-realis marker. One can
immediately note that Chamacoco is the Zamucoan language with the most complex
verb morphology. From a merely descriptive point of view, one can detect the person prefix
(PP), the thematic vowel (TV) and the root: PP-TV-Root. The first two elements may be
absent. In practice, depending on the shape of the 3-realis, verbs can be classified as follows:
(1) ‘radical verbs’, ø-ø-Root; (2) ‘thematic verbs’, ø-TV-Root; (3) ‘prefixal verbs’, PP-TVRoot. In addition, one can identify further subgroups depending on the type of 3-realis PP. In
Chamacoco, one can even identify subgroups depending on the TV. These groups and
subgroups are not related, at least sinchronically, to any semantic or valency property.
thematic verbs
radical verbs
thematic verbs
radical verbs
/ʨ/-verbs /ts/-verbs
/d/-verbs thematic verbs
(/d/-, /j/-) (/i/-, /u/-, /ɨ/-)
Table 1. Verb paradigms as based on the 3-realis marker.
radical verbs
This paper will analyse the causes that might have originated the 3-realis polymorphism
and will discuss the properties of each inflectional class, which may concern, e.g., the
morphological expression of: (1) the 1-person; (2) the 3-person irrealis; (3) the presence of
irregularities or subregularities in the paradigm.
Comparing the verb paradigms of the Zamucoan languages, one can show that some verbs
have changed their inflectional class. Specifically for Chamacoco, one can observe that new
verb groups have arisen (Ciucci 2013). For instance, consider the /ʨ/-verbs paradigm, which
is the most regular and the largest group in all Zamucoan languages. As one can see from the
table, this paradigm has split in two pardigms in Chamacoco (/ʨ/- vs /ts/-verbs):
(1) AZ. ʨ-ise (3-realis) ‘to reach’, AY. ʨ-ise (3-realis) ‘to reach’, CH. ʨ-iɕ (3-realis) ‘to reach’
(2) AZ. ʨ-iasore (3-realis) ‘to help’, AY. ʨ-osõre (3-realis) ‘to help’, CH. ts-osɨr (3-realis)
‘to help’
This is clearly a recent innovation, as shown by the fact that in many Chamacoco verbs /ʨ/and /ts/- can alternate. During the talk, the person markers of the Zamucoan languages will be
compared with the person marking system of other geographically adjacent languages.
Finally, the 3-realis will be compared with the 3-person of possessable nouns. There is an
interesting parallelism between verb and noun morphology: as it happens, in the possessive
inflection, one can distinguish ‘prefixal nouns’, ‘thematic nouns’ and ‘radical nouns’. There is
however a remarkable complementarity in the Zamucoan 3-person morphology: while most
verbs are ‘prefixal’ (see again the Table), most nouns are ‘thematic’.
The only 3-person prefixes common to both verbs and nouns are /d/- and /j/-, to be
observed in a tiny group of Chamacoco verbs. In Ayoreo and Chamacoco the vast majority of
nouns characterized by /d/- are kinship terms: this is the only semantic criterion related to 3person morphology which one can detected in the whole Zamucoan family.
BERTINETTO, PIER MARCO 2009. Ayoreo (Zamuco). A grammatical sketch. Quaderni del Laboratorio di Linguistica
della Scuola Normale Superiore 8 n.s.
BERTINETTO, PIER MARCO 2011. How the Zamuco languages dealt with verb affixes. Word structure 4/2. 215-220.
BERTINETTO, PIER MARCO (to appear). Ayoreo (Zamuco) as a radical tenseless language. Miscellanea in onore di
Alberto Mioni.
CHOMÉ, IGNACE 1958 [ante 1745]. Arte de la lengua Zamuca. Présentation de Suzanne Lussagnet. Journal de la
Société des Américanistes de Paris 47. 121-178.
CIUCCI, LUCA 2007/08. Indagini sulla morfologia verbale nella lingua ayoreo. Quaderni del Laboratorio di
Linguistica della Scuola Normale Superiore 7 n.s.
CIUCCI, LUCA 2009. Elementi di morfologia verbale del chamacoco. Quaderni del Laboratorio di Linguistica
della Scuola Normale Superiore 8 n.s.
CIUCCI, LUCA 2010a. La flessione possessiva dell’ayoreo. Quaderni del Laboratorio di Linguistica della Scuola
Normale Superiore 9,2 n.s.
CIUCCI, LUCA 2010b. La flessione possessiva del chamacoco. Quaderni del Laboratorio di Linguistica della
Scuola Normale Superiore 9,2 n.s.
CIUCCI, LUCA 2013. Inflectional Morphology in the Zamucoan languages. Ph.D. thesis. Pisa: Scuola Normale
HIGHAM, ALICE, MAXINE MORARIE & GRETA PAUL 2000. Ayoré-English dictionary. Sanford, FL.: New Tribes
Mission. 3 vols.
KELM, HEINZ 1964. Das Zamuco: eine lebende Sprache. Anthropos 59. 457-516 & 770-842.
ULRICH, MATTHEW & ROSEMARY ULRICH 2000. Diccionario Ɨshɨro (Chamacoco) – Español / Español – Ɨshɨro
(Chamacoco). Misión Nuevas Tribus Paraguay: Asunción.
Inflection classes and orthogonal conditions
University of Surrey
[email protected]
Introduction: dimensions of generalization
Inflectional systems demonstrate the need for generalizations in the lexical ‘dimension’. I
review such generalizations (inflectional classes) with data from a less familiar language,
namely Burmeso. I then show that we also require orthogonal (cross-cutting) generalizations
in the other dimension, first for Burmeso, and then more generally. I provide a typology of
these generalizations or ‘conditions’; this typology is based on the antecedent of the condition
and the type of paradigm affected.
Dimension 1: inflection classes
Consider the following verbal forms of Burmeso (Donohue 2001: 100, 102). They establish
that we need generalizations over groups of lexemes, that is, inflection classes.
female, animate
mass nouns
banana, sago tree
arrows, coconuts
‘see’ (class I)
‘bite’ (class II)
The prefixal gender-number markers mark agreement with the absolutive argument. The two
examples given each represent a substantial number of verbs. Donohue (2001: 101) states
explicitly that inflection class membership is not predictable from obvious semantic
correlations, and that the classes are approximately equal in size. We need the notion
inflection class to describe Burmeso since (i) the distribution of the affixes can be described
only by reference to the particular lexeme; and (ii) each affix is sufficient to predict every
other within the paradigm. In fact, the inflection classes of Burmeso are close to canonical
(Corbett 2009).
Dimension 2: conditions on inflection
Since we need inflection classes we might try to eliminate generalizations in the other
dimension. However, Burmeso again shows that this is not possible. We need to specify
gender-number syncretisms (such as: [V SG] = [IV SG] = [IV PL] = [III PL]), since they
apply equally to the two inflection classes, but they involve different phonological forms.
Once the need for these orthogonal conditions on inflection is accepted, they deserve full
scrutiny. Arguably, they have received somewhat patchy attention. One type which has
justifiably aroused great interest and detailed study is the morphomic pattern. And that is
precisely the type found in Burmeso, since the syncretism noted has no extra-morphological
justification. But this is just one type of condition, as I demonstrate.
Types of conditions on inflection
We begin with the clearest examples, and then descend a slippery slope. In Russian, a fully
explicit account of noun inflection might have eight different inflection classes. However, we
usually represent four inflectional classes, and specify that each contains both count and
non-count nouns. The condition, then, is this: if a noun is low on the Animacy Hierarchy
(antecedent) it lacks a plural sub-paradigm (consequent). This avoids two types of
duplication: (i) on the formal side, there are four major inflectional classes and each of them
includes count nouns and singularia tantum; (ii) on the semantic side, similarly, the distinction
cross-cuts the four classes.
This condition is convincing because it is fully orthogonal to the inflectional classes.
Its antecedent rests on the lexical semantics of the noun. Its consequent involves the abstract
content of the paradigm (the number of cells, not their realizations). This suggests a typology,
built on possible antecedents of the condition and possible consequents. Going further, the
animacy condition of Russian (as in ACC=GEN student-a ‘student’ vs ACC=NOM zakon ‘law’)
also has a semantic antecedent, but its consequent is different: it determines patterns of
syncretism, hence the form paradigm. Our Burmeso condition has part of speech as its
antecedent (all verbs) and its consequent again involves the form paradigm. Then there is the
(not exceptionless) syllable-counting condition on plural formation in Serbo-Croat, giving
grad-ov-i ‘cities’ versus prozor-i ‘windows’. Here the antecedent is phonological (single vs
multi-syllable stem) and the consequent (augment or none) involves the realization paradigm.
The antecedent of a condition on inflection may invoke four types of information: semantic,
part of speech,
or phonological. antecedent
content paradigm form paradigm realization paradigm
The consequent semantic
Russian number Russ. animacy
may involve the part of speech
Burmeso sync.
paradigm (the phonological
SC plural
type of paradigm), the form paradigm, or the realization paradigm (Stump 2012). This implies
a twelve-member typology of conditions (above), and it proves remarkably well attested, as
demonstrated by further data from a wide range of typologically diverse languages.
Morphological theory requires the availability of generalizations in two dimensions:
inflection classes and orthogonal conditions on inflection. The latter deserve more consistent
study. While I have sketched a typology of these conditions, we still need to investigate their
distribution and relative frequency. Their diachrony is intriguing.
CORBETT, G. G. 2009. Canonical inflectional classes. In: Fabio Montermini, Gilles Boyé and Jesse Tseng(eds)
Selected Proceedings of the 6th Décembrettes: Morphology in Bordeaux, 1-11. Somerville, MA: Cascadilla
Proceedings Project. • DONOHUE. M. 2001. Animacy, class and gender in Burmeso. In: Andrew Pawley,
Malcolm Ross & Darrell Tryon (eds) The Boy from Bundaberg: Studies in Melanesian Linguistics in Honour of
Tom Dutton, 97–115. Canberra: Pacific Linguistics. • STUMP, G. 2012. The formal and functional architecture
of inflectional morphology. In Angela Ralli, Geert Booij, Sergio Scalise & Athanasios Karasimos (eds.),
Morphology and the Architecture of Grammar: On-line Proceedings of the Eighth Mediterranean Morphology
Meeting (MMM8), Cagliari, Italy, 2011, 255-271. • WURZEL, W. 1984. Flexionsmorphologie und
Natürlichkeit. Berlin: Akademie. • ZALIZNJAK, A. 1967. Russkoe imennoe slovoizmenenie Moscow: Nauka.
The singular/plural split and the making of verb inflectional classes
in San Pedro Amuzgo
Surrey Morphology Group, University of Surrey
[email protected]
Surrey Morphology Group, University of Surrey
[email protected], [email protected]
Amuzgan is a small family of languages spoken in Mexico which forms one of the branches
of the Oto-Manguean phylum. The Amuzgan family consists of two languages:
Xochistlahuaca Amuzgo and San Pedro Amuzgo (SPA). Oto-Manguean languages are
renowned for having complex inflectional systems (e.g. Baerman, forthcomming; Finkel &
Stump, 2009; etc.). Amuzgan is no exception. The overall structure of the inflection of the
two languages in the family is very similar, but in this paper we focus on SPA, and we base
our analysis on the data in the large dictionary by Stewart & Stewart (2000) and on the
grammatical description by Buck (2000).
In the inflection of many verbs in SPA, there is a paradigmatic split (i.e. a systematic
differentiation) of singular subject forms and plural subject forms. While the split is a unique
property of Amuzgan (i.e. it is not found elsewhere in Oto-Manguean), what makes it
interesting for a theory of inflection and inflection classes is its complex array of both formal
and distributional properties across the verbal lexicon.
Verbs in SPA inflect for TAM and person of the subject and they fall into two large inflection
classes we treat for convenience as A and B. In (1) and (2) we illustrate each class with the
incompletive (INCPL) and the irrealis (IRR) of the verbs ‘stay’ and ‘break’ (Tone is
represented by numbers, 1 being the highest. Ballistic syllables are indicated by an acute
accent). A-verbs have only two inflected forms for person: one for the 3rd person and another
for anything else (person of subject is encoded by means of enclitics). B-verbs are more
complex. Person and number are realized by a combination of affixal and prosodic formatives
(differences in tone and ballistic stress). Note that the INCPL prefix co2- is used in B-verbs
only in the plural forms. This we take as a first instance of the existence of the singular/plural
split in verbs.
‘stay’ (A)
co -ntho =ja
Furthermore, B-verbs fall into a range of other inflectional classes attending to how the
different TAM distinctions are encoded. For example, the B-verbs ‘break’ in (2) and ‘suck’ in
(3) belong to different classes because they select a different prefix for the IRR, i.e. the former
takes n-, the latter ngi-.
‘break’ (B)
ma -t‹ʔ›án
‘suck’ (B)
ma -tí
In addition to the distribution of the INCPL prefix co2-, it is often the case that a verb has a
distinct stem only used to build the plural forms. This can be seen in the verb for ‘twist, split’
in (4). What makes these verbs remarkable is that they inflect for the plural as if they
belonged to a different inflectional class. This can be seen if the forms of the singular of the
IRR are compared with the IRR of the verb for ‘eat’ in (5), and those of the plural with the
verb ‘suck’ in (3). This phenomenon suggests that the singular/plural split is embedded with
the lexeme itself, as if we had two different verbs in one.
‘twist, split’ (B)
ma -kṓn
‘eat’ (B)
ma -k‹ʔ›íia
co2-kiiaʔ =oʔ3
nt-kiiaʔ =oʔ3
With a phenomenon such as this, a number of questions arise: What is the distribution of the
classes A and B in the lexicon? How many TAM classes are there? How representative or
frequent is the pattern shown by the verb in (4)? What is the status of the plural stem? Is it
basic (given by the lexicon)? Or is it built on the singular? In a given verb whose singular
stem has certain phonological properties, is the class membership of the plural subparadigm
predictable? In our paper, we will address these and other similar questions in the quest to
understand where the cues for inflectional class structure can reside.
BAERMAN, M. Forthcoming. “Inflectional class interactions in Oto-Manguean”, in J-L. Léonard and A. Kihm
(eds.), Issues in Meso-American morphology. Paris: Michel Houdiard. ● BUCK, MARJORIE J. 2000. "Gramática
Amuzga", in Cloyd Stewart and Ruth D. Stewart. 2000, Diccionario Amuzgo de San Pedro Amuzgos, Oaxaca.
[Serie de vocabularios y diccionarios indígenas Mariano Silva y Aceves, 44], 363-473. Mexico City: Instituto
Lingüístico de Verano. ● FINKEL, R AND G STUMP. 2009. “Principal parts and degrees of paradigmatic
transparency”, in J. P. Blevins and J. Blevins (eds), Analogy in Grammar: Form and Acquisition. Oxford:
Oxford University Press, 13-53. ● STEWART, CLOYD AND RUTH D. STEWART (eds.). 2000. Diccionario Amuzgo
de San Pedro Amuzgos, Oaxaca. [Serie de vocabularios y diccionarios indígenas Mariano Silva y Aceves, 44].
Mexico City: Instituto Lingüístico de Verano.
Emergence and decay of inflectional class systems—an evolutionary perspective
WU Vienna
[email protected]
Productivity is a central property of inflectional classes and, as such, decisive in order to
understand both the synchronic organization of inflectional systems and their evolution.
Based on an investigation of the nominal inflectional classes of Latin and Old Italian (Gardani
2013), this paper investigates the role that productivity plays in the dynamics of emergence
and decay of inflectional class systems. It thus offers insights into long-term morphological
change, covering a time of approximately 2,000 years.
On a wider perspective, the paper follows the line of research on the evolution of
morphology recently propelled forwards by Carstairs-McCarthy (2010). More specifically,
the investigation is theoretically couched in the functionalist framework of Natural
Morphology (Dressler 2003), from which also the methodological approach is derived. In the
model proposed, productivity is conceived of as a force of attraction peculiar to inflectional
classes which is reflected both in the morphological integration of loanwords, in the
inflectional class assignment of indigenous neologisms arisen via conversion, and in the
behavior of class shifters. The measurement of productivity is operated on semi-synchronic
cuts, and the diachronic trajectory is outlined by connecting the results of the single cuts. The
data on loanword integration are drawn from the contact languages, Ancient Greek, Germanic
varieties, Arabic, Byzantine Greek and Old French, thus reflecting different chronological
depths in the analysis. The corpora are constructed on a variety of sources, including the
Thesaurus Linguae Latinae, for Latin, and the OVI corpus, for Old Italian.
This paper shows that there exists a link between the lifecycle of inflectional classes and
a universal preference for biunique form-meaning relationships, with respect to the
morphological realization of certain morphosyntactic features. For example, provided that a
given language is sensitive to the feature of number, an inflectional class which realizes the
values of number by means of formatives which are biunique (or rank high on a scale ranging
from biuniqueness to ambiguity), is more likely to undergo an increase in its grade of
productivity than a class whose formatives are ambiguous (or rank low on a scale ranging
from biuniqueness to ambiguity) with respect to the realization of the same feature value(s).
Also, the paper demonstrates that the need for a (progressively more) biunique realization of
number values can promote a process of morphogenesis. In Old Italian, starting from 1200,
three new nominal inflectional classes emerged, viz. poeta poeti ‘poet’, promessa promessora
‘promise’, and nome nomora ‘name’, and were quite productive. As the data show, these
classes arose from extant classes whose formatives were ambiguous with respect to the
realization of both the singular and the plural, via replacement of those formatives by
formatives that realized plural in a unique way.
CARSTAIRS-MCCARTHY, A. 2010. The Evolution of Morphology. Oxford: Oxford University Press.
DRESSLER, W. U. 2003. Degrees of grammatical productivity in inflectional morphology. Italian Journal of
Linguistics 15(1). 31–62.
GARDANI, F. 2013. Dynamics of Morphological Productivity. The Evolution of Noun Classes from Latin to
Italian. Leiden & Boston: Brill.
OVI = Corpus OVI dell’italiano antico. 2005–. CNR. Available at http://gattoweb.ovi.cnr.it/.
TLL = 1900–. Thesaurus Linguae Latinae. Editus auctoritate et consilio Academiarum quinque Germanicarum,
Berolinensis, Gottingensis, Lipsiensis, Monacenis, Vindobonensis. Lipsiae: B.G. Teubner.
Reflexes and remnants of inflectional classes in Maay nominal morphology
Department of Linguistics and Cognitive Science, Pomona College
[email protected]
This paper describes aspects of nominal morphology in Maay, a language of southern Somalia.
Dialects of closely related Somali are described as having inflectional classes, but in Maay these
classes have broken down or been rearranged. In this paper, I focus on two aspects of nominal
morphology – gender and plural marking – that exhibit inflectional class behavior in other East
Cushitic languages. I discuss inflectional class patterns observed in those languages, then
describe the remnants of the corresponding systems in Maay. I discuss how the inflectional
class systems in the other languages, likely inherited, have changed in Maay and propose some
possible mechanisms and explanations for the changes.
The Ethnologue classifies Maay (Paster 2007, Comfort & Paster 2009, Paster 2010) as East
Cushitic. Tosco (2012: 278) advocates for a smaller Omo-Tana group within East Cushitic, and
within this an Eastern (‘Somaloid’) subgroup including Maay, Somali, Rendille, and Tunni,
among others. Eastern Omo-Tana (EOT) languages have gender class systems wherein every
noun is masculine or feminine. Masculine nouns are identified (in the singular) by k-initial
variants of suffixes including the definite marker, demonstratives, and possessive markers;
feminine nouns take t-initial versions of these suffixes. Nominal gender appears to be relatively
stable within the group, except for some interesting changes that appear to be in progress
currently in some Maay nouns, since speakers disagree on their gender. As is widely known,
however, the apparent gender of nouns in these languages varies in the plural in EOT languages.
Rendille (Oomen 1981), for example, has a gender polarity pattern wherein nouns change to
the opposite gender in the plural. In Standard Somali, most nouns switch gender in the plural
(Saeed 1987). In Central Somali (Saeed 1982), plurals in -o show complete polarity, while
plurals in -yal have masculine gender regardless of their gender in the singular (see below for
more on the -o and -yal plural suffixes). The unidentified dialect of Somali discussed by
Lecarme (2002) exhibits the reverse of the Central Somali pattern: plurals in -yal exhibit
polarity, while plurals in -o are masculine. Maay has yet another variation on this pattern: all
plural nouns are masculine, regardless of which plural suffix is used (1). This is apparently true
of Tunni as well (Tosco 1997: 43). It seems likely that true polarity was present in Proto-EOT
and that the masculine plural patterns were innovated, since polarity also exists elsewhere in
Cushitic outside this group, e.g., in Oromo (Andrzejewski 1960).
Languages in this group, especially Somali, are also documented as having declension classes
for plural formation, although as Diriye (2000) notes, there is no consensus on the number of
classes even in a given dialect; cf. Andrzejewski (1964, 1979), Hyman (1981), Banti (1988),
Morin (1991), Saeed (1999). In Maay, the declension classes have been completely
reorganized: there are only two productive plural markers, and their distribution for different
speakers ranges from somewhat to completely phonologically determined. The distribution of
plural markers in Maay appears chaotic when data from multiple speakers are aggregated, but
individual speakers have clearly describable systems of plural formation that suggest a change
in progress to a completely phonologically based system wherein the -o suffix occurs with
consonant-final stems while the -yal suffix occurs with vowel-final stems or optionally with
consonant-final stems in place of -o (Paster 2013). An interesting pattern is documented where
both plural suffixes are used simultaneously on a single noun, with no consistent meaning
change relative to singly-marked plurals (Paster 2010); a further development seems to be
underway where multiple plural marking indicates a larger number (2). I suggest that the
phonologically based pattern in Maay is emergent in response to the breakdown of an earlier
declension class system – perhaps due to the introduction of non-native lexical items that lacked
assignment to a declension class, since Maay has many apparently borrowed nouns that are not
cognate with Somali.
(1) Masculine plurals in Maay (examples simplified to phonemic notation)
a. Feminine nouns in singular
Feminine  masculine in plural
‘my hippo’
‘my hippos’
‘my daughter’
‘my daughters’
‘my knife’
‘my knives’
b. Masculine nouns in singular
‘my grandfather’
bakaile-key ‘my rabbit’
‘my thigh’
Masculine stays masculine in plural
‘my grandfathers’
‘my rabbits’
‘my thighs’
(2) Plural marking in Maay
a. Vowel-final stems take only -yal
buundo-yal ‘bridges’
b. Pattern I: C-final stems take -o ~ -yal ~ -o-yal
mukulal-yal ~ mukulal-o ~ mukulal-o-yal
~ eey-o
~ eey-o-yal
c. Pattern II: C-final stems take -o ~ -yal vs. -o-yal, distinguishing the number of items
~ luk-yal
‘many, many feet’
~ geet-yal
‘many, many trees’
ANDRZEJEWSKI, B. 1960. The categories of number in noun forms in the Borana dialect of Galla. Africa 30: 6275. • ANDRZEJEWSKI, B. 1964. The Declensions of Somali Nouns. London: School of Oriental and African Studies.
• ANDRZEJEWSKI, B. 1979. The Case System in Somali. London: School of Oriental and African Studies. • BANTI,
G. 1988. Two Cushitic systems: Somali and Oromo nouns. Pp. 11-49 in H. van der Hulst and N. Smith, eds.
Autosegmental Studies on Pitch Accent. Dordrecht: Foris. • COMFORT, J. & M. PASTER. 2009. Notes on Lower
Jubba Maay. Pp. 204-216 in M. Matondo, F. McLaughlin and E. Potsdam, eds. Selected Proceedings of the 38th
Annual Conference on African Linguistics. Somerville, Massachusetts: Cascadilla Proceedings Project. • DIRIYE
ABDULLAHI, M. 2000. Le Somali, dialectes et histoire. PhD thesis, University of Montreal. • HYMAN, L. M. 1981.
Tonal accent in Somali. Studies in African Linguistics 12.2: 169-201. • LAMBERTI, M. 1985. Die Somali-Dialekte.
Hamburg: Buske. • LECARME, J. 2002. Gender ‘polarity’: Theoretical aspects of Somali nominal morphology. Pp.
109-141 in P. Boucher and M. Plénat, eds. Many Morphologies. Massachusetts: Cascadilla Press. • MORIN, D.
1991. Marques et relations en Somali: le problème des déclinaisons. Linguistique africaine 6: 75-102. • OOMEN,
A. 1981. Gender and plurality in Rendille. Afroasiatic Linguistics 8.1: 35-75. • PASTER, M. 2007. Aspects of Maay
phonology and morphology. Studies in African Linguistics 35.1: 73-120. • PASTER, M. 2010. Optional multiple
plural marking in Maay. Pp. 177-192 in F. Rainer, W.U. Dressler, D. Kastovsky, and H. C. Luschützky, eds.
Current Issues in Linguistic Theory 310: Variation and Change in Morphology. Amsterdam: Benjamins. • PASTER,
M. 2013. An I-language approach to inter-speaker variation in Maay. Ms., Pomona College. • SAEED, J.I. 1982.
Central Somali: A Grammatical Outline. Monographic Journals of the Near East: Afroasiatic Linguistics 8.2.
Malibu, California: Undena Publications. • SAEED, J.I. 1987. Somali Reference Grammar. Wheaton, Maryland:
Dunwoody Press. • SAEED, J.I. 1999. Somali. Amsterdam/Philadelphia: John Benjamins Publishing Company. •
TOSCO, M. 1997. Af Tunni. Köln: Rüdiger Köppe. • TOSCO, M. 2012. The unity and diversity of Somali dialectal
variants. Pp. 263-280 in N.O. Ogechi, J.A.N. Oduor, and P. Iribemwangi, eds. The Harmonization and
Standardization of Kenyan Languages: Orthography and Other Aspects. Cape Town: Centre for Advanced Studies
of African Society.
Verb morphology and conjugation classes in Dunan (Yonaguni)
[email protected]
Kyoto University
[email protected]
Japanese and other Japonic languages have a relatively simple and transparent verb morphology, with few non-canonical phenomena. They generally exhibit a highly agglutinative structure
with little morphophonology, few conjugation classes, limited stem allomorphy, and very few
irregular verbs. This explains why morpheme-based constructive approaches to morphology have
been popular for the description of these languages. However, the verb morphology of Dunan
(a.k.a Yonaguni, ISO 639-3: yoi), a highly endangered Ryukyuan language spoken by around
500 people on Yonaguni Island (Okinawa pref., Japan), is much more complex and departs in
several interesting ways from the simpler system exhibited by its relatives. In particular, the
existence in Dunan of a rich system of conjugation classes, of a rather high degree of allomorphy,
and of several non-canonical phenomena, is noteworthy. The analysis presented here is based on
original data gathered during fieldwork that fill important gaps in previous descriptions.
The same morphosyntactic features are not marked uniformly for all verbs in Dunan, and
several different patterns partition verbs into conjugation classes. The exact number of these
(more than a dozen) depends upon the amount of morphophonology one is ready to posit, but
in any case it is a function of three mostly independent factors: a) stem allomorphy, b) suffix
allomorphy, and c) tone alternation (metatony). Class membership is historically correlated to
segmental shape, but this has been largely obscured by historical changes, and a synchronic
account needs to refer to several principal parts. A word-and-paradigm (W&P) analysis based on
principal parts has more merits than a morpheme-based approach, which requires recourse to
abstract underlying forms and a set of complicated rules of limited scope. The morpheme-based
approach, far from illuminating the structure of Dunan’s verb morphology, rather obscures its
alternation patterns and its network of implicative relationships between verb forms. A wordbased approach is anyway required for irregular and fully suppletive patterns (e.g. ‘be, exist’ an,
negative stem minu), which escape a cogent morpheme-and-rule treatment. A W&P approach
has thus the merit of allowing a unified treatment of all verbs.
Unlike the typical case of Indo-European languages, in Dunan verbs can be partitioned
and grouped together primarily according to their pattern of stem variation rather than suffix
allomorphy, which is less prominent. A verb can have up to three different stem forms, which
often exhibit a reduction/augmentation pattern, e.g. ‘remember’ ubw ~ ubui ~ ubuir, ‘drop’ ut
~ utu ~ utus. Such alternations are phonologically unmotivated but are the result of historical
change, and they are hardly amenable to a purely phonological treatment. It does not seem
desirable for example to posit a [DORSAL] Ñ [CORONAL] rule which applies in front of the
medial verb suffix -i but not in front of the imperative -i (e.g. ‘pull’ sunt-i vs. sunk-i).
Verb stems in Dunan underlie sets of forms that do not realize any coherent set of morphosyntactic features. Thus no common feature can be found between the imperative and circumstantial
forms, even though they systematically share the same stem in all conjugation classes. Stems are
thus best viewed as purely morphological objects.
Though each verb has only a restricted set of stem forms, the total number of stems needed
to be posited in order to account for all stem alternations across the different classes is far greater.
This is because not only stem alternations depend on conjugation class, but the distribution of
stems within a paradigm is also class-specific. Thus, while the shortest stem ubw of ‘remember’
is used in its perfect forms (e.g. ubwan), for ‘drop’, it is the longest one utus that is found (e.g.
utusyan). The situation is reversed for negative forms (e.g. ubuir-anun vs. ut-anun), and there is
globally no uniform distribution of stems within paradigms. For example the same stem utu is
found in the imperative (utu-i) and the prohibitive (utu-nna) for ‘drop’, but two different stems
ubuir and ubui are found for ‘remember’ (ubuir-i vs. ubui-nna).
Suffix allomorphy is also observed, though it is less developed than stem allomorphy, and it
is not predictable from stem alternation patterns. This concerns the simple present -u ~ -∅, the
circumstantial -uba ~ -iba, or the perfect -ya ~ -a ~ -yu ~ -u. Such variation is not fully correlated
with patterns of stem alternation but further subclassifies Dunan verbs into conjugation classes.
Table 1: Metatony in Dunan conjugations
Dunan has three word-tones (H, L, and F) which
are lexically determined, but some verbs undergo
tonal alternations in some inflectional forms (Tab. 1). Imperative
Six patterns of paradigmatic metatony can be distin- Prohibitive
guished, which can be reduced to four: a single non- Hortative
alternating class and three L ~ F alternating classes. Circumstantial
Simple Present
These tones are melodies which apply to whole word Simple Past
forms, and they cannot thus be segmented into a stem Negative
part and an affixal part. Metatonic forms do not al- Perfect
ways share a common segmental stem-shape, and
putting the burden of metatony on stems would only Sequential
lead to multiply them. On the other hand, metatonic forms do not share a common base of
morphosyntactic features, and therefore metatony itself cannot be considered to be an exponent
but should recognized as a purely morphological feature.
Interestingly, several conjugation classes are non-canonical in the sense that they are barely
distinguished from others. The perfect is sometimes the only form that discriminates between
two classes, with some extreme examples such as ‘boil’ and ‘get boiled’, which are completely
homophonous except for their perfect forms, i.e. nyan vs. nyun. Perfect forms exhibit however
some variation, which threatens the distinction between several classes, as some speakers seem
to be leveling the -(y)a/-(y)u allomorphy to -(y)a. This can be straightforwardly explained by the
fact that perfect forms are less common and that the -(y)u forms are in minority. Contrary to an
approach based on principal parts and implicative relationships, it does not seem that a diacritic
approach to conjugation classes can offer an insightful account of this change.
Many “impostors” are also found, and many forms are ambiguous about class membership
when taken isolatedly. Even the perfect is ambiguous by itself, and a Xsyan form could belong
to two completely different classes. Only reference to other forms, such as the simple present
(e.g. Xn or Xsirun) can resolve the ambiguity. A systematic homophony is also found between
etymologically related transitivity pairs for some forms (e.g. utun present perfect of ‘fall’ or
simple present of ‘drop’), between the medial and imperative forms for most classes (e.g. ‘get up’
that-i), and some perfect forms are ambiguously marked with a suffix -u, which with a different
stem is an exponent of the simple present. The existence of ambiguous forms shows that simply
listing the presence of a stem or a word form for a verb is insufficient: the key information is to
know what paradigm cell such a form realizes.
Selected references
B LEVINS, J. P. 2006. Word-based morphology. Journal of Linguistics 42: 531–573. ‚ C ORBETT, G. G. 2009.
Canonical inflectional classes. In M ONTERMINI, F., B OYÉ, G. & J ESSE, T. (eds.) Selected Proceedings of the 6th
Décembrettes: Morphology in Bordeaux. Somerville: Cascadilla Proceedings. ‚ C ORBETT, G. G. & BAERMAN, M.
2006. Prolegomena to a typology of morphological features. Morphology 16(2): 231–246. ‚ S TUMP, G. & F INKEL,
R. A. 2013. Morphological typology: From word to paradigm. Cambridge: Cambridge University Press. ‚ U WANO,
Z. 2011. Yonaguni hōgen dōshi katsuyō no akusento shiryō (2). Kokuritsu Kokugo kenkyūjo ronshū 34: 135–164.
‚ YAMADA, M. P ELLARD, T. & S HIMOJI, M., 2013. Dunan (Yonaguni)-go no kan’eki bunpō to shizen danwa
shiryō. In TAKUBO, Y. (ed.) Ryūkyū rettō no gengo to bunka, Tokyo: Kuroshio shuppan, 291–324.
Morphomic stems or inflectional classes?
Noura Ramli and Andrew Spencer
Department of Language and Linguistics, University of Essex, UK
{nramli, spena}@essex.ac.uk
Inflectional classes (ICs) are typically defined in terms of affix allomorphy and are generally
taken to require arbitrary (morphomic) labels: [CLASS:{I, II, III, . . . }]. But paradigm complexity
can also result from stem allomorphy and this is not normally described in terms of ICs. Bonami
& Boyé (2003) have argued that French lacks conjugation classes and that instead, complexity
is defined in terms of stem sets. Additionally, it has been argued that stem alternations can serve
as realizations of morphosyntactic property sets (MPSs), without necessarily resorting to nonconcatenative inflectional morphology (Baerman & Corbett 2012). However, Spencer (2012)
has argued that stem must be morphomic in PFM, defined by stem formation rules and selected
by stem selection rules (‘Block 0’), not by realization rules. Stems/root alternants can systematically realize inflections only through non-concatenative or prosodic exponents operating as
morphophemically defined realization rules (‘Block I, II, ...’), e.g. the passive u-i vocalism in
Classical Arabic verbs of the f-‫ع‬-l class (kataba ⇒ kutiba ‘it was written’) or the phenomenally
complex prosodic ‘inflectional series’ found in Chinantecan conjugation (Palancar, to appear)
(and, arguably Nuer, Baerman 2012, which we will briefly re-analyse).
Traditional descriptions of Arabic lack ICs but modern dialects have developed stem alternations which partition the lexicon into classes (Camilleri 2012). We illustrate with Transitional
Libyan Arabic. Verbs inflect for PFV/IPFV series. Inflections are identical for all verbs for a
given series. In ‘weak’ verbs, stem allomorphy distinguishes 3rd person PFV forms from 1/2
(Table 1). This might be analysable as multiple exponence of person/number MPSs (Baerman
& Corbett 2102). However, in ‘sound’ verbs the stem allomorphy distinguishes 3PL/3SG.F from
1/2 and the 3SG.M forms, a morphomic distribution (Table 2). The complexity of the system in
terms of conditional entropy (Ackerman et al. 2009) is similar to that of an IC system defined
by affixal homophony. We provide two analyses within PFM (Stump 2001), one in terms of
ICs, the other in terms of hierarchically organized morphomic stems, lexically defined stem
classes and stem selection rules in both cases treating the stems as morphomic. We (tentatively)
conclude that these give the same results, but since all accounts need stem formation/selection
anyway we should favour the stem-based analysis and reject ICs.
Recently Stump (2012) has argued from Sanskrit conjugation for four types of IC including
a ‘metaconjugational’ class, in which for some lexemes a given IC determines one set of MPSs
(e.g. imperfect) and for others it determines another (e.g. aorist). But the endings are the same
for all lexemes: the variation is entirely defined by stem allomorphy. Stump requires these
stem patterns to be ‘conjugational’ , however, because his rules of paradigm linkage crucially
(it seems) appeal to arbitrary IC labels. We reanalyse his data in terms of stem selection.
We tentatively conclude that stem-based complexity is distinct from exponence-based complexity, offering further support for the autonomy of morphology. Future research will focus on
the search for a clear characterization of the two types.
Table 1: Weak verbs perfective series
2SG. M
ɣa:b ‘be absent’
ʒa:b ‘bring’
ʕadda ‘depart’ (Form II)
Table 2: ‘Sound’ verbs perfective series
2SG. M
laʕab ‘play’
ʕaraf ‘know’
nḍarab ‘be hit’ (Form VI)
Selected references •Ackerman, F et al. 2009. ‘Parts and wholes’. In J. Blevins and J. Blevins (eds)
Analogy in Grammar, 54-82, OUP •Baerman M 2012. ‘Paradigmatic chaos in Nuer’. Language 88:46794 •Baerman M & G Corbett 2012. ‘Stem alternations and multiple exponence’ Word Structure 5:5268 •Bonami, O & G Boyé 2003. ‘Supplétion et classes flexionnelles dans la conjugaison du français’
Langages 152:102-26 •Camilleri, M ‘Morphological Complexity in Maltese’ 2012. Proc MMM8, 92113 •Palancar E (to appear) ‘Revisiting the complexity of the Chinantecan verb conjugation classes’
in Léonard & Kihm (eds) Issues in Meso-American Morphology. Michel Houdiard •Spencer A 2012.
‘Identifying stems’ Word Structure 5:88-108 •Stump, G 2012. ‘The formal and functional architecture of
inflectional morphology’ Proc MMM8, 254-68.
Interrelation of grammatical gender and inflectional class:
A case study of Russian ‘common gender’ nouns
ZAS, Berlin
[email protected]
There are two conflicting claims concerning correlation between grammatical gender and
inflectional class of Russian nouns. Some claim that grammatical gender can be predicted
form inflectional class (Corbett 1982, 1991; Corbett & Fraser 2000). In contrast, others
claim that inflectional class can be predicted from grammatical gender (Crockett 1976,
Thelin 1975). However, there is a class of nouns in Russian ― the so-called “common
gender” nouns ― which cannot be accounted for in any of these proposals.
Common gender nouns denote individuals, like s’irot-á ‘orphan’ and sudj-á ‘judge’ that
can trigger either masculine, or feminine agreement (1a–b). Compare with other Russian
nouns that can trigger only masculine or only feminine agreement (2–3).
(1) a. bol’š-ój
big-MASC.N.SG orphan-N.SG (MASC; CLASS II)
‘big orphan’
b. bol’š-ája
‘big orphan’
(2) a. bol’š-ój
‘big young person (male)’
b. *bol’š-ája
young-N.SG (FEM; CLASS II)
‘big young person (male)’
(3) a. bol’š-ája
‘big nurse’
b. *bol’š-ój
‘big nurse’
nurse-N.SG (FEM; CLASS II)
orphan-N.SG (FEM; CLASS II)
Common gender nouns differ from other Russian nouns, because their grammatical
gender cannot be predicted from inflectional class and vice versa, their inflectional class
cannot be predicted from grammatical gender. For example, in (1a–b), the common
gender noun s’irot-á ‘orphan’ belongs to the inflectional class II, which is evident from
the inflectional suffix –a (only class II nouns have the inflectional suffix –a in Russian).
Nonetheless, it can be either masculine or feminine, which is evident from either
masculine (1a) or feminine (1b) agreement. With this respect, the question arises: What is
so special about common gender nouns, which sets them apart from all other nouns?
I propose that the difference between common gender nouns and other Russian nouns is
that the former are unmarked for grammatical gender, while the latter are marked for
either [MASC], or [FEM] (4a–c).
a. n
b. n[masc]
√j’unoš‘young male’
c. n[fem]
This proposal makes the following predictions. First, in case of common gender nouns,
either masculine or feminine agreement can be used when they refer to an individual
whose sex is unknown. However, when the sex of an individual is known, one or the
other agreement will be used. This prediction is borne out (5a–b).
‘big orphan’
(referring to a male individual)
orphan-N.SG (MASC; CLASS II)
‘big orphan’
orphan-N.SG (FEM; CLASS II)
(referring to a female individual)
Second, in case of other Russian nouns, we predict that it would not matter whether the
sex of an individual they refer to is known or not. If they are marked for [MASC], they
will only trigger masculine agreement, regardless of the sex of an individual they refer to.
Similarly, if they are marked for [FEM], they will only trigger feminine agreement. This
prediction is borne out (6a–b).
‘big nurse’
nurse-N.SG (FEM; CLASS II)
(referring to a male individual)
‘big nurse’
nurse-N.SG (FEM; CLASS II)
(referring to a female individual)
I show in detail how the current proposal works with respect to Russian expressive
suffixes. First, I argue that expressive suffixes cannot be marked for grammatical gender
in Russian. Instead, they are marked for the inflectional [CLASS II]. Second, I argue that
the grammatical gender of a derived word can be predicted from the inflectional class of
an expressive suffix; which gives additional evidence for the first claim discussed above
(grammatical gender can be predicted from inflectional class). Third, I illustrate how this
works with respect to common gender nouns. I argue that since common gender nouns
are unmarked for grammatical gender, their gender cannot be predicted from the
inflectional class of an expressive suffix. Thus, a derived common gender noun with an
expressive suffix will always trigger either masculine or feminine agreement (7), just as it
does without an expressive suffix (1).
‘big orphan (expressive)’
‘big orphan (expressive)’
Corbett, G. (1982). Gender in Russian: An account of gender specification and relationship to declension.
Russian Linguistics, 6(2), 197–232. Corbett, G. (1991). Gender. Cambridge: Cambridge University Press.
Corbett, G., & Fraser, N. (2000). Default genders. In B. Unterbeck, & M. Rissanen (Eds.), Gender in
grammar and cognition (pp. 55-98). Berlin; New York: Mouton de Gruyter. Crockett, D. (1976).
Agreement in Contemporary Standard Russian. Slavica: Cambridge University Press. Thelin, N. (1975).
Notes on general and Russian morphology. Studia Slavica Upsaliensia, 15, 30-32.
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

Download PDF