COYOTE PAPERS Andrea Heiberg Working Papers in Linguistics from A to Z

COYOTE PAPERS Andrea Heiberg Working Papers in Linguistics from A to Z
COYOTE PAPERS
Working Papers in Linguistics from A to Z
Volume 9
1995
Edited by
Colleen M. Fitzgerald
Andrea Heiberg
The University of Arizona
COYOTE PAPERS
Working Papers in Linguistics from A to Z
Edited by
Colleen M. Fitzgerald
Andrea Heiberg
Volume 9
1995
The University of Arizona ®
Tucson, Arizona
Preface
The Linguistics Circle at the University of Arizona is pleased to present
Volume 9 of the continuing series of Coyote Papers, our working papers in
linguistics.
We would like to thank our fellow students in the Linguistics Department
for their encouragement and support. Thanks also to all the contributors
who have made this volume possible. Special thanks to Laura Conway for
the cover artwork and to Pat Perez for sharing her invaluable Coyote Papers
wisdom. For financial support, we thank the Associated Students of The
University of Arizona.
Colleen M. Fitzgerald
Andrea Heiberg
For more information on Coyote Papers or any other publication of The
University of Arizona Department of Linguistics write to:
Coyote Papers
Department of Linguistics
Douglass 200 East
Tucson, Arizona 85721
520- 621 -6897
e-mail: [email protected]
Contents
The Meter of Tohono O'odham Songs
1
Colleen M. Fitzgerald
Arapaho Accent
29
Amy Fountain
Deriving Ternarity
39
Michael Hammond
"PRO Analysis" for Subject -Oriented
Secondary Predicates
59
Hisako Ikawa
Evidence from Modern Greek
for Refinement of the OCP
Diane Meador
79
The Meter of Tohono O'odham Songs *
Colleen M. Fitzgerald
The University of Arizona ®
Department of Linguistics
O. Introduction
The evaluation of poetry and songs has been essential to the progress of generative metrical i
theory. The bulk of work done in generative metrics focuses on English poetry (although
Kiparsky (1968), Maling (1973), and Prince (1989) are some notable exceptions). Limitations of
such a focus become evident when metricists make typological claims, as in Hayes (1989).
Research on the meter of other languages is thus critical for a valid typology of meter and
metrical rules. With such a goal in mind, I examine the meter of Tohono O'odharn songs.
Tohono O'odham (henceforth TO; formerly Papago) is a Native American language of the UtoAztecan family spoken in southern Arizona. An examination of these songs provides a valuable
test for theories of meter, such as that advanced in Hayes (1989). The songs comprise part of an
oral tradition and represent an instance of meter which is Native American, rather than IndoEuropean. These two characteristics unite to produce a metrical system which differs from that
of English poetry.
Here I give both a description and an analysis of the meter of previously unanalyzed Tohono
O'odham traditional songs. The crucial problem is characterizing the line. Song lines are
flexible in the number of syllables and stresses that they allow, but are not completely
unconstrained. The positioning of stressed material is strictly regulated, as stresses are
prohibited from appearing adjacent to another stress, or in the second and final metrical positions
of a line. I argue for binary trochaic feet, built at left edges and wherever stressed syllables
occur; stresses are further prohibited from appearing in weak position. The analysis has four
results: 1) lines are flexible in some traditions, 2) binary feet, with constituency, are necessary,
3) poetic meter is shown to invoke morphology to satisfy constraints on the meter, and 4)
metrical rules for left edges may be strict, contra Hayes (1989).
This study also has relevance beyond meter, specifically for phonological theory and the
study of Native American literature. Recent work in phonology has focused on the nature of
constraints, particularly work in Optimality Theory (McCarthy and Prince 1993, Prince and
Smolensky 1993). Optimality Theory argues that constraints may be violable or undominated
(and inviolable); here I show evidence from meter for two inviolable constraints which govern
the meter (the Edge Constraint and the Binary Foot Constraint). To prevent violations, the
morphology is systematically manipulated. The TO meter provides one example of how a
constraint governed system in meter operates (see Hayes (1993) for specific discussion of
relevant issues for meter and optimization).
*Many people have given helpful comments along the way, including the participants of the 1993 WECOL.
The following have responded to written versions of this paper. Ken Hale, Mike Hammond, Andrea Heiberg, Jane
Hill, Leanne Hinton, Terry Langendoen, Peg Lewis, Diane Ohala, Pat Pérez, Gilbert Youmans and Ofelia Zepeda.
Special thanks to Mike Hammond for extensive input on multiple drafts. Any errors are my own. For surface
forms, I use the official orthography of the Tohono O'odham nation, which was developed by Albert Alvarez and
Ken Hale. This orthography approximates a phonemic transcription. I modify the orthography slightly and mark
primary stress, which is not represented in the official orthography.
1 The term 'metrical' is used in two ways: 1) 'A theory of phonology in which phonological strings are
represented in a hierarchical manner, using such notions as segment, syllable, foot and word (cf. also prosodic
phonology). Originally introduced as a hierarchical theory of stress, the approach now covers the whole domain of
syllable structure and phonological boundaries' (Crystal 1991, 218). 2) The (linguistic) study of versification. as in
poetry and songs. I use the second sense of this term, except when referring to the metrical grid of Hayes (1983,
1989).
1
This study is also germane to work on Native American literature, as it shows that oral
literature, here the songs, may have a system of meter comparable to those found in written
literature, such as the poetry of William Shakespeare. Hinton (1984, 1990) looks at the meter of
Havasupai songs from a linguistic perspective. This paper takes a similar approach. This type of
work furthers recognition of Native American songs, narratives, and speeches as a valid and
important body of literature.
This paper is structured as follows. The first section gives the necessary background on the
theory of meter. This is followed by background on reduplication and stress in Tohono O'odham
and a description of the songs. The second section presents the analysis of the song meter
discussed above. In the third section, I discuss implications of this analysis for the theory of
meter proposed in Hayes (1 989). Specifically, I examine his typological claims for metrical
rules, and show that the typology must be expanded to account for the strict left edge meter
presented here. Finally, I conclude the paper with a discussion of the importance of the study for
metrical theory.
1.0 Background
This section provides the theoretical and descriptive underpinnings necessary for an
analysis of the TO song meter. The first section briefly gives background on meter and
generative metrics. This is followed with a discussion of the necessary descriptive facts, both of
TO phonology and the songs. This is especially important to determine how to characterize this
meter and, just as importantly, to determine how this meter can not be characterized. This
section starts with a look at the relevant facts of TO phonology, then moves into background on
the song format and the songs themselves.
1.1 Meter and Generative Metrics
Meter is a regular pattern of rhythm, where the pattern may be associated with one or more
of several factors: quantity, stress, syllable count, and tone. Quantitative meter is a pattern
based on the arrangement of heavy and light syllables, as in Greek or Latin verse. Stress, or
accentual meter, uses stressed syllables as the basic unit; Old English poetry, such as Beowulf, is
an example of this. A third pattern is characterized by a fixed number of syllables; this pattern
is typical of Romance versification, as in the 12- syllable line of French Alexandrine verse. A
fourth type results with the intersection of stress and syllable count, as in English iambic
pentameter, which consists of a relatively fixed number of both stresses and syllables.2 Meter
may also regulate verse on the basis of tone, as in the Chinese poetry examined in Chen (1979).
Generative meter takes as its basic premise that these styles reflect the unconscious use of
language. One of the goals of generative metrics is to show the principles which underly the
rule- governed behavior of language in meter. Likewise, that is the goal of this paper.
Work in the theory of meter has focused on what constitutes the proper representation of
stress for meter. There are three current theories which each argue for a different representation
of stress: tree -based, grid- based, and Arboreal Grid -based. Much of the debate in generative
meter has centered on the metrical representation of the words (that is, language) to which the
constraints refer. This paper does not address that debate; rather, I am concerned with what the
representation of the line must be for the proper treatment of the meter.
Now I will give examples of grid and tree systems3 and explain some terminology, using
English iambic pentameter as an example. Iambic pentameter is a line of poetry consisting of ten
alternating stressless and stressed syllables. Each syllable constitutes a metrical position in the
line, and these metrical positions may be restricted in what type of material they can contain. An
iamb consists of a stressless syllable followed by a stressed one (W S). Strong positions are
2 There is also free verse, where neither syllables nor stress are regulated.
3Speaking generatively, that is. Actually, there is an additional theory proposed by Halle and Keyser (1971);
see Kiparsky (1975) for arguments against it.
2
those which generally contain stressed syllables, although unstressed syllables are also allowed
here. Weak positions, in contrast, are those which contain unstressed syllables; stressed
syllables only appear in these positions under special circumstances.4
First, there is the tree -based theory of meter found in Kiparsky (1977), which represents a
line of iambic pentameter as below:
(1)
A
A
A
A
A
WSWSWSWSWS
Trees represent prominence relations between two metrical positions5, the S position is
more prominent than the W. Additionally, t---- also allow a representation of constituency;
here two metrical positions are each constituents of a single unit, the foot. More hierarchical
structure may be added, such that two feet comprise a unit.6
A second representation of meter comes from the grid -based theory presented in Hayes
(1983, 1989). Grids indicate stressed syllables with an "X" and stressless syllables with a ". "; the
height of the X column over a stressed syllable indicates its relative level of stress. In this
theory, the underlying meter of a line of iambic pentameter is a grid, as below.
(2)
.X.X.X.X.X
Grids encode the level of stress and their local relations. Stress levels of syllables can only
be compared where syllables are adjacent. The absence or presence of stress is also encoded, as
is the stress level. However, the grid cannot represent constituency.?
In this paper, I will use the grid -based representation for TO. I do so as a formal
convenience, rather than as an argument that the grid constitutes the proper representation of the
language. As I will show later in the paper, it is the tree -based theory of meter which is critical
for representing TO meter.
1.2 Stress in Tohono O'odham
As this paper focuses on TO song metrics, a discussion of the stress system in TO is a
prerequisite to examining the meter. For the purposes of this paper, I use only the primary stress
in words. 8 As Hill and Zepeda (1992) show, primary stress falls on the first syllable of a stem.
This can be seen in forms from Hill and Zepeda (1992: 356, 367) in (3).
(3)
a.
b.
c.
d.
e.
Underlying Form
/ma:ci/
/da:- da:ka/
/da-dagasapa/
/ku:bisi -ce/
/si- da- dapa -ka/
Surface
má:c
dá:dk
dádag sp
kú:bsc
s -dádpk
Gloss
'knowing'
'noses'
'pressing down with fingers repeatedly'
'made smoky, dusty'
'smooth (plural)'
4lambic pentameter also characterizes weak positions (W) as odd and strong positions (S) as even.
5With the addition of more hierarchical structure, trees may indicate other prominence relations as well. See
for example, Kiparsky (1977) and Youmans (1989) for more discussion of these matters.
6Such structure is argued to be binary in Prince (1989).
?However, Hammond (1991) presents arguments for the Arboreal Grid based on the meter of Thomson. The
Arboreal Grid theory of Hammond (1988) is an amalgam of tree and grid theory, representing both constituency and
levels of stress. However, it also allows feet which are not binary. As I will show later, TO meter requires strict
binarity of feet. For this reason, I assume a grid -based theory, rather than the Arboreal Grid.
Primary stress is crucial for the characterization of the meter; however, it is not clear that the same is true
for secondary stress.
3
The form in (3a) shows a monomorphemic form, which receives primary stress on the
initial syllable.9 When words are reduplicated, as in (3b -c, e), primary stress falls on the prefixal
reduplicant. Suffixes, as shown in (3d), do not affect stress assignment. The form in (3e) shows
that the stative prefix, s -, does not receive stress. Other prefixes (e.g., third person pl. obj, ha-)
do not receive stress. These facts are further shown in (4)10:
(4)
a.
b.
c.
d.
e.
f.
Surface
gógs
gógogs
ñéid
ha -ñéid
ñéñeid
ha- ñéñeid
Gloss
'dog'
'dogs'
'seeing'
'seeing (pl. object)'
'seeing (sg. object)'
'seeing them (pl. object)'
1.3 The Song Corpus
This section gives background on the song corpus, the format in which I give the song data,
and the specific phenomena of song lines. As we will see in the following subsections, the
phonology of these songs is rather complex. I try to give song lines in a format which makes the
song phonology more accessible. Also, while work such as Bahr (1980, 1983), Chesky (1943),
Haefer (1981) and Underhill (1938) have dealt with the music and songs of the O'odham, this is
the first theoretical treatment of the meter of O'odham songs. The unanalyzed status of the
corpus thus necessitates the exposition of what the songs look like before presenting the analysis.
The song corpus used in this paper consists of 11 songs totaling 78 lines. The songs used
are traditional 11, where the description of traditional corresponds to a specific purpose and
melodic content, and a specific type of origin. The songs are used for social or ceremonial
purposes, such as traditional round dance songs or healing songs. Musical content is generally
associated with a specific purpose; for instance, a certain type of dance requires a certain rhythm
(much like a polka or waltz requires certain rhythms). Also, traditional songs are those which
are either passed down from one generation of a singer to another or come to the singer via the
inspiration of a dream.
In this study, I analyze songs which all come from one source, Wallace (1981).12 Haefer
(1977) discusses how variation in songs serves both as a way to discriminate between the quality
of singers and as a means to create a new song. By using one source, we isolate characteristics
of a given singer or singers, much as Kiparsky (1977) and others have sorted various English
poets into 'dialects' of English poetry (i.e., Tudor poets, Milton, etc.) by the variation in their
metrical rules.
The example below shows a song line, which I give in the format I use in this paper. Each
song line is noted by SONG; the example given differs from its notation in what I term CITATION
form. CITATION is used to gloss the TO song forms into TO dictionary forms.13 I then provide
9Hi11 and Zepeda (1992) argue for demoraicization of unstressed syllables in certain environments. In the
above figure, I give only sketchy derivations. For more detailed exposition of these arguments, refer to their paper.
At this point, the reader only needs to know where primary stress occurs in TO words.
13As the orthography used for TO does not indicate stress, I have taken the liberty of doing so here. These
forms come from Zepeda (1988).
11My understanding of a definition of 'traditional' comes primarily from Haefer (1977) and Ofelia Zepeda
(p.c.).
12 The description of the songs in Wallace (1981) shows that they can be easily characterized under the
auspices of a single system of meter.
13Citation forms are confirmed in Mathiot (1973) and Saxton, et al. (1989). They appear here in the official
orthography.
4
an English gloss of the O'odham forms in GLOSS .14 The final line, TRANSLATION , represents the
song line in an approximate English translation. The format is given in (5):
(5)
SONG
CITATION
GLOSS
TRANSLATION
Wí- pis -mel
ñé -ñei
wípismel
wa -il
ñéñe'i
a
hummingbird pl -song
AUX
'Hummingbird songs surround me.'
bí je- mi -da.
ñ- bíjemid
1SG -to surround
The notion of a line is referred to above and deserves some attention. Informally, I
consider a line a set of one or more clauses, such that a group of lines constitutes a song. A line
may also consist of just one phrase or may split a phrase between two lines. While clauses and
phrases are important, they are not necessary to the definition of line. For our purposes, LINE is
a string of words which corresponds to a musical phrase in the song.
There are two facts which suggest that the definition suffices in giving a characterization of
line for TO songs. Lines in the songs may be repeated; one common pattern is A A B C B C,
where the first line is repeated twice (A), the next new line (B) alternates with another new line
(C), and the (B C) sequence repeats.
1.4 Reduplication
In this section, I examine reduplication within the songs. Reduplication is used
morphologically in TO to indicate the plurality of nouns, verbs, and some postpositions. Of the
48 instances of reduplication found in the song corpus, only 20 of these are plural reduplications.
This means that there are 28 reduplicated forms which do not have an associated plural meaning.
Here, I will first contrast morphological and nonmorphological (or Vacuous) reduplication.
Following this, I will present an examination of where Vacuous Reduplication occurs. The data
here will show that Vacuous Reduplication is motivated by stress considerations. Finally, I will
give the generalizations of this section, which lay the groundwork for the analysis of meter I
present later in this paper.
As mentioned above, the songs do contain examples of straightforward, morphological
reduplication. An example of this appears below, with the reduplicated word underlined in song
and citation lines. The singular form, gídwul 'swallow' reduplicates as gígidwul 'swallows' (the
reduplicant is underlined).
(6)
E -da
g
eda g
while DET
gígi- dwul -e
ñéi- o -pa -ha
gígidwul
ñé'iopa vOC 15
PL- swallow
PL -to fly
'While the swallows flying...'
However, there are also song lines which contain reduplication without a corresponding
semantic change. That is, singular nouns and verbs reduplicate in song lines, where they would
not reduplicate in citation lines. This can be seen in the examples below, with prefixal
reduplicants underlined.16
(7) a.
Vacuous Reduplication line -initially:
14Ofelia Zepeda, a native speaker of O'odham and a linguist, was an invaluable source of help in both the TO
and English glossing of the songs.
I5Vocables are contentless syllables, as in the final syllable, -ha in the song form ñéi- o- pa -ha. Vocables
(glossed as VOC) consist of adding an entire syllable, not just a vowel, at the end of a word. While I do not deal
with vocables here, Hymes (1981) and Hinton (1980) give evidence that these 'nonsense syllables' have specific
function within a text. Vocables do seem able to appear both line -medially and line -finally, and there are examples
which have two vocables next to each other. Future research may reveal the role played by vocables in these songs.
161 rely on the helpful intuitions of Ofelia Zepeda for where reduplication is vacuous and where syllables are
meaningless, or 'vocables'.
5
Wá -wai
gí- wa -lige
we -co ná- ha -gio
kc
in
in
weco náhagio
kc
Wáw
gíwulk
rock
Cinched
below mouse
CONJ LOC
'The mouse runs around there below Cinched Rock.'
b.
mém- é- li- hi -me.
mémelihim.
to run to repeatedly
Vacuous Reduplication line -finally:
ká- wu -li -ki
yam-e ké -he -ka
oi
i- gáwulk
'am
ké:k
'o
AUX
INIT -to differ LOC
to stand
'God starts to differ standing there.'
Jiós
Jiós
God
c.
Multiple instances of Vacuous Reduplication within a line:
na
so-so
na
són
kú:g
the beginning the end
soon perhaps
'soon perhaps the beginning, the end,'
oi
oi
The first example in (7a) shows two things: 1) Vacuous Reduplication may occur linemedially, 2) Extra vowels or syllables (or both) may also occur where there is Vacuous
Reduplication.
The example in (7b) and the second form in (7c) also show that extra vowels may occur
with Vacuous Reduplication. The line -final vowel in (7b) may occur in speech, while the extra
line -final vowel of (7c) does not occur in speech. These forms also show that Vacuous
Reduplication may occur line -finally. Finally, the two reduplicated forms in (7c) also show that
a given line may have more than one occurrence of this type of reduplication.
At this point, let me also point out the generalizations of Vacuous Reduplication, as
suggested by the data in (7). First, only monosyllabic citation forms (wáw, ké:k, són, and kú: g )
correspond with the reduplicated song forms. Second, these forms are all either nouns or verbs,
and hence, bear lexical stress.17 Third, the forms either precede a stressed syllable (as each
precedes either a verb or noun) or occur line -finally (or both, as in (7d)). The crucial point here
is that all words which reduplicate are stressed monosyllabic words, providing these words do
not precede an unstressed syllable. These generalizations become clearer in the chart below,
where I present a breakdown of the conditions where Vacuous Reduplication occurs:
(8) Distribution of citation line factors where Vacuous Reduplication occurs:
Category
Length in a
Precedes Stressed g
Occurs Line -finally
Nouns
1
7
6
Verbs
1
2
5
The hypothesis, then, is that it is conditions in the citation lines which correspond with
song lines that show Vacuous Reduplication. This is borne out when we consider whether there
are any song lines that allow stressed syllables to be adjacent or occur line -finally. The contrast
obtains, as the corpus has no examples of song lines with either adjacent stresses or line -final
stresses for monosyllables. There are, however, examples of polysyllabic words with stress
appearing adjacent to a stressed syllable or line -finally. These cases do not occur with Vacuous
Reduplication (unless they have a phonological shape, as does jewe d 'earth', which allows
17To simplify things here, I exclude postpositions which are stressed. Fitzgerald (1993) shows that these
postpositions, when monosyllabic, behave the same way as words with lexical stress, as both types of words
reduplicate preceding a stressed syllable. In certain cases, the postposition appears to be receiving stress from
emphatic use. Both Kiparsky (1977) and Hayes (1983) show how contrastive stress or emphasis may change the
normal scansion for stressed syllables.
6
scansion as monosyllables 18 ). In (9), I give these examples (only Vacuous Reduplication is
underlined).
(9) a.
S
s
REP
wa
oa
s wá- pu -si -me
wú- wa -ke.
s- wápusim
CON
STAT-PL -damply
wúwhag
to emerge
Damply it emerges.'
b..
Áw -pa
hió -sig
ga -pe
hí- me- na -ha.
áuppa
híma VOC VOC
ga VOC
hiósig
to walk
cottonwood blossom
over there
'Over there, cottonwood blossoms pass by.'
c.
Pí -si -ne
jé je -wen
jéwe d
earth
mó -ka -me
písin
mó'o -kam
bison
head -one with
'Bison Head (place).'
The song lines in (9a -b) lines allow two lexically stressed words to appear adjacent to each
other, without Vacuous Reduplication. Note that these examples do not have adjacent stresses,
as in the previous examples from (7), as the underlined forms are not monosyllables. This means
that there are intervening unstressed syllables.
The final two stressed words in (9a -b) show that such polysyllabic forms also appear linefinally. Again, the forms do not reduplicate. However, unlike the forms from (7), the stressed
syllables of these words do not appear as the final syllable of the line.
Based on these two sets of facts (the first showing where Vacuous Reduplication does
occur, the second showing where it does not), I propose that Vacuous Reduplication is
motivated by the metrical system of the song.
In order to view the metrical nature of Vacuous Reduplication, I give the song lines from
(7) again below in (10). In this example, I show both citation and song lines in the grid -based
theory of meter used by Hayes (1983, 1989). An examination of these forms shows that
Vacuous Reduplication only occurs wherever either two stresses are adjacent (XX) or wherever a
stress would fall line- finally at the rightmost edge of a line (X).
(10) a.
X
.
Wá -wai
X
X
.
..
gí- wa -lige
X
.
X
.
.
we -co ná- ha -gio
.
X . .
weco náhagio
kc in mém-é-li-hi-me.
...
.
X
Wáw
gíwulk
kc in mémelihim.
cinched
CONJ LOC to run to repeatedly
below mouse
rock
'The mouse runs around there below Cinched Rock.'
.
18Two disyllabic words always reduplicate line -finally: dó'ag 'mountain' and jéwed' 'earth'. The intuition
here is that they are treated as monosyllables by the meter, much as in English, heaven may be scanned as
monosyllabic heav'n; in fact, their medial onsets do allow them to be pronounced as monosyllables in less careful
speech. However, in line -medial positions, they do not always pattern with monosyllables.
7
(10) b. X
.
X
.
.
.
Jiós
oi
ká- wu -li -ki
X
. X .
Jiós
'o
i- gáwulk
God AUX INIT-to differ
'God starts to differ standing
X
C.
Oi
.
.
.
.
yam -e l -he -ka
X
.
ké:k
to stand
'am
LOC
there.'
X
.
so-so
na
X
.
.
kú:-ku:-re.
X
X
kú:g
na
són
perhaps
the beginning the end
'soon perhaps the beginning, the end,'
oi
soon
By aligning the song and citation lines to the metrical grid, the generalization is validated:
Wherever citation lines appear with adjacent stresses or a line -final stress, the song line appears
with the leftmost word of two stressed words or the final word in a line vacuously reduplicated.
These observations suggest two conclusions: 1) Stress clash and line -final stresses are
impermissible in the song meter and 2) Vacuous Reduplication resolves these stress violations
where they would otherwise occur, given the corresponding citation line.
Song lines may include additional vowels. Let me now cover the interaction of the effects
of Vacuous Reduplication with these vowels to show that they are irrelevant to this metrical
process. Below I compare examples of Vacuous Reduplication in song lines with and without
extra vowels. The examples in (1 lb,c) show that extra vowels may also appear in the same
environment (where two stresses are adjacent or where a stress is line- final) where Vacuous
Reduplication occurs. I have underlined adjacent stresses and double- underlined line -final
stresses.
(11) a.
SONG
X .
Wá-wai
X . . .
gí-wa-lige
X
.
.
we-co ná-ha-gio
.
X
.
.
.
kc in mém é-li-hi-me.
.
X .
X . .
X . . .
Wáw
gíwulk
weco náhagio
kc in mémelihim.
rock cinched below mouse CONJ LOC to run to
'The mouse runs around there below Cinched Rock.'
X
CIT
RED
ONLY
VOW
ONLY
X
.
Wá-wai
X
X
Wáw
X . . .
gí -wa lige
.
gí -walk
.
X
.
.
we-co ná-ha-gio
.
X
.
.
we-co ná-ha-gio
8
.
X
.
.
.
kc in mém é-li-him.
X
.
.
.
kc in mém-é-li-hime.
(11)b.
X
SONG
Jiós
CIT
RED
ONLY
VOW
ONLY
X
oi
.
CIT
X
.
ká- wu -li -ki
yam-e
ké -he -ka
X
X .
ké -hek
to stand
Jiós
oi
X .
ká -wulk
X
Jiós
oi
ká- wu -li -ki
Oi
na
oi
X
X
na
són
kú:g
perhaps
the end
the beginning
'soon perhaps the beginning, the end,'
X
.
.
yam
X
.
X
yam -e
oi
of
na
só-so
na
X
són
.
ké -ka
X
.
só-so
.
.
kú: -ku: -rie.
X
X
VOW
ONLY
.
X
ké:k
soon
RED
ONLY
.
X
. X
.
Jiós
'o
i- gáwulk
'am
God AUX INIT -to differ LOC
'God starts to differ standing there.'
c.
SONG
.
X .
kú: -qs
The first two examples show that extra vowels and Vacuous Reduplication may occur in the
same environment; that is, wherever stresses are adjacent or line -final. However, the forms in
(11 a,c) are critical in showing that only Vacuous Reduplication is motivated by this environment.
Wáw in (1 la) and són in (11c) both reduplicate adjacent to another stress, without surfacing with
extra vowels. This indicates that these vowels, regardless of whether they appear in TO speech
or underlying forms, do not affect Vacuous Reduplication. They must be invisible to the meter
in order to meet the conditions for Vacuous Reduplication.
1.5 Characteristics of Song Lines
In this section, I cover three characteristics of song lines. First, I discuss the length of song
lines, and show that there is a flexible number of syllables. Second, I show further flexibility in
the number of stresses per line. Third, I discuss two important metrical positions of TO song
lines, second and final to show that distributional facts reveal these two positions are never filled
with stressed syllables. These facts indicate how TO song lines are rigid.
This section is important for several reasons. First, it will help in determining how to
characterize the underlying metrical pattern of TO song meter. Second, recall from the
discussion of meter above that rhythmic patternings are derived from quantity, stress, or
syllables. Here the discussion shows that it is difficult to place TO song meter in one of these
categories.
First, let us examine line length. Line length in these songs is rather variable; songs may
use lines of anywhere from 7 to 19 syllables, as seen below:
9
(12)
a. S ' a -me g yód -ha -me
s
'am
g ó'odham
yu huwi .
Ó- 'od -ha -me
we -wem -e
jú- ne -kam.
wem
júñkam
with
REP LOC DET people
over here.
to exist -one
'They say over here are the People. With the People is the one who was.'
b.
c.
ó'odham
people
na
so-so
na
§án
perhaps
the beginning
'soon perhaps the beginning, the end,'
oi
oi
soon
S
wa
oa
s- wá- pu -si -me
CON
STAT- PL -damply
kú:- ku : -Oe.
kú:g
the end
wú -wa -ke
wúwhag
to emerge
'There, they say, damply they emerge.'
REP
s- wápusim
The song line in (12a) has 19 syllables, (12b), has 7 syllables and the third example (12c) is
10 syllables.
Songs do not individually cluster about a certain line length either, as each song may
contain lines of varying lengths. This in itself is not unusual, as it is characteristic of Old English
verse, such as Beowulf (cf. Cable 1974, 1991; Russom 1987). However, the variability in OE
meter can be reduced by factoring out resolution.19 This is not true of TO meter. While the
statistics show how long (or short) lines may be, there is little to tell what the limits to the lower
and upper reaches actually are. Therefore, I suggest that the length of a song line is
unconstrained.
The number of stresses per line, and where they appear, is also an important factor in
characterizing the flexible nature of the line. The number of stresses per line appears to cluster
around 3, although like line length, these numbers are relatively flexible. A given line may
contain anywhere from 1 to 5. The actual numbers for the distribution are given here in (13):
(13)
Distribution of lines according to number of stresses:
4
0
1
2
3
4
28
0
33
7
STRESS:
NUMBER OF LINES :
5
2
6
0
The numbers in (13) show that most song lines contain either 2 or 3 stresses. In (14), I
combine information to show the interaction of number of stresses per line with number of
syllables.
i9Resolution is 'whereby a short stressed syllable and the following syllable, long or short, are scanned as
one' (Cable 1974, 7).
10
(14) Distribution of song lines, according to line length and number of stresses
Syllables per song line
011.101
.cw.c-4..."
"77
a
Number
of stresses
per line
2
6
X
.....
2
6
.
1
4
6
7
7
8
,
.,.::
,
T< '+, :X<
-
2 22
.i,...,
2
2
2
3
3
3 .
2
2
2
-
-
This chart shows that as the number of stresses increases, so does the number of syllables.
One stress per line corresponds with song lines of the smallest attested lengths, 7 and 8 syllables.
These lengths have more lines with 2 stresses per line; at 2 stresses per line larger lines also start
to appear, with line lengths of 9 -13, and 18 attested at this point. These line lengths (9 -11
syllables) become more common with three stresses per line; in fact, they are the lower cutoff
point for this range. At three stresses per line, lines of 12 -15 syllables, and 19 syllables begin to
appear. At four stresses per line, the lower cutoff becomes 13 syllables in length; 14 and 18
syllable lines also appear with four stresses per line. Finally, 5 stresses per line, the highest
attested number of stresses, appears in 17 syllable lines.
The crucial generalization here has been with respect to the nature of a song line; it cannot
be characterized consistently in terms of length in syllables or number of stresses. It does appear
that as the number of stresses per line increase, so do the syllables. However, there is no clear
evidence for there being restrictions on either, except for the restrictions on where stresses can
appear in a line.
Now let us explore the distributional facts of where these stresses may appear in a line. As
stated above, the second and final metrical positions of a given line are never occupied by
stressed syllables. The distributional facts are most easily revealed in the chart below, which
gives the distribution of stresses for the first four metrical positions in a line, the final three in a
line, and all other (medial) positions
(15)
First
34
Position of stresses within a song line:
Fourth
Other
Second
Third
18
0
35
65
Antepenult
Penult
39
9
Final
0
The restriction on these two positions does not hold of citation line. The table in (16)
shows that stresses appear in all positions in a (citation) line:
(16)
First
32
Position of stresses within a citation line:
Third
Fourth
Other
Antepenult
Second
7
15
41
16
28
Penult
34
Final
5
A comparison of these three distributional charts suggests that the songs restrict the second
and final position. The latter restriction is noted as typical of trochaic verse, according to
Attridge (1982). The restriction on second position in song lines, when viewed in conjunction
with the high number of stresses which appear in the initial metrical position of a line, suggest
that song lines begin with a trochaic sequence (S W) or two lexically unstressed syllables (W W),
but never begin with an iambic sequence (W S). These elements argue for a trochaic meter, at
least at the line's edges.
11
It is important to note one further point: trochaic meter typically places stresses in the S
positions, which are odd, and avoids placing in W positions, the even ones. While this
characterization is true of the leftmost W position (it is always even and never has stress), the
same is not true of the rightmost position (the final metrical position of a song line may be either
odd or even, and never has stress). I postpone further discussion of this for later, when I propose
an analysis for TO song meter.
These gaps in the second and final positions of the charts in (15) and (16) become even
more intriguing when we note that these positions are occupied in citation lines. The low
number of stresses even in citation lines suggests that avoiding stress in these positions plays a
role in structuring the line, even at the citation level. I compare these facts in (17), where I
compare song and citation lines in the grid.
(17)
b.
Stress appears in second position in citation line, but not in song line:
X .
X . . .
X . .
Wá -wai
gí- wu -lik -e
nó -no -hard
X
X .
X
Wáw
gáwulk
dó'ag
constricted
rock
mountain
'Constricted Rock Mountain'
a.
Stress appears in final position in a citation line, but not in a song line:
X
.
.
X . . .
X .
.
Jiös of
ká wu -li -ki
yam-e
ké -he -ka
X
.
. X
.
X
'am
Jiós
i- gáwulk
'o
ké:k
God AUX INIT -to differ
LOC
to stand
'God starts to differ standing there.'
The contrast in (17) between the song lines and the citation lines shows first that stresses
are prohibited in the second and final metrical positions of a line. The line in (17b) also shows
that the morphology of TO, which uses reduplication to indicate the plurality of nouns and verbs,
may also be employed in order to avoid placing stresses in final position. I showed above that
the Vacuous Reduplication in these lines is used to systematically prevent stresses from being
adjacent from each other or from appearing line -finally. The example in (17a), which triggers
reduplication by virtue of the two adjacent stresses, does show a stress in second position. 20
Interestingly, stressed syllables which occupy second position in citation lines also appear
adjacent to other stresses, violating two restrictions on the meter. This is resolved by Vacuous
Reduplication in the song line. As the reduplication adds another syllable, the stressed second
syllable moves into the third metrical position.
Finally, if we look at the line -final stress which triggers reduplication, we see that both
conditions (line -final and adjacent stresses) may appear in one line. This can be seen in (18).
2QAn alternate account of the form wáwai 'cliff is that the form reflects an archaic singular, rather than
reduplication (thanks to Jane Hill for noting this). Interestingly, the form wáw also appears once, at the beginning of
one song, as wawawai. The scarcity of other forms like this make it impossible to generalize. I do note here that
wáwai may reflect a preserved form; this does not change the fact that stressed syllables do not appear in the second
metrical position. I assume here that wáw is the citation form and wáwai a form reflecting Vacuous Reduplication.
12
(18)
X
.
oi
na
.
só -so
X
.
.
kú: -ku: -rye.
X
X
na
són
kú:g
soon perhaps the beginning the end
'soon perhaps the beginning, the end,'
oi
In this section, I have made several points regarding the absence of stressed syllables in the
second and final metrical positions in a song. Specifically, I have shown that 1) Distributional
facts of song lines reveal these positions to never appear with stressed syllables, 2) Citation lines
may appear with stressed syllables in these positions, and 3) Vacuous Reduplication
strategically allows a stressed final syllable to appear nonfinally, as well as helping to avoid the
placement of such syllables in second position. I conclude this section having shown that as the
morphology may be manipulated to avoid the appearance of stresses in second and final
positions within a song line, it is the case that these positions are systematically devoid of
stressed syllables. That is, it is not a coincidence, but rather reflects a metrical strategy of
avoiding these two positions in the meter.
2.0 An Analysis of Tohono O'odham Song Meter
In this section I will present an analysis of the TO song meter. The crucial generalizations
that this analysis must accommodate are these:
(19) The descriptive generalizations of Tohono O'odham song meter:
A.
The second and final metrical positions are never filled with
stressed syllables as Vacuous Reduplication is used to prevent stresses
from appearing in these positions.
B.
The song's meter is also restrictive in that it prohibits adjacent
stresses. Adjacent stresses trigger reduplication of the leftmost
element to create an intervening unstressed syllable.
C.
Lines are flexible in number of stresses and syllables.
In this section I will first present a proposal which covers the first set of generalizations.
Then I will propose a treatment of the second set of generalizations; following this, I will
discuss the integration and interaction of both constraints.
2.1 The Edge Constraint
There are three sets of related facts which must be dealt with adequately in this section: the
restriction on both the second and final metrical positions; the strictness of both edges; and the
fact that the final position may be either odd or even. In fact, all noninitial stresses may occur in
either odd or even positions. In the first section, we saw the edge effects robustly. Here I will
show the third characteristic as well.
The restriction on stress in these two positions suggests that the meter is trochaic (S W); S
positions are odd and W positions are even ones within a line. In the examples below, I show
one unfortunate consequence of proposing that the meter is trochaic; by aligning each syllable in
the song line with alternating odd and even positions, I show that all noninitial stresses may fall
in odd or even positions, focusing here on the final position. The examples are given below.
13
(20)
a. Song line ending in odd (S):
.
X . .
X
X .
jé je -wen
.
SONG
mó -ka -me
pí -si -ne
WSW
SWS
SWS
jewel
mó'o -kam
písin
head -one
bison
'Bison Head (place).'
b.
Song line ending in even (W):
.
SONG
earth
S
S
REP
wa
W
oa
CON
s- wá- pu -si -me
X
wú -wa -ke
s- wápusim
PL- damply
wúwhag
to emerge
.
X
.
.
.
WSW
SW SW S
'Damply they emerge.'
c.
SONG
Song line ending in even (W):
.
.
X
X .
X
jé -wen -e
ká -ha -ce
Nóligk 'am
jéwed
ká:c
SWSW
SWS
WSW
to turn LOC
earth
'Noligk lies there on earth.'
d.
SONG
.
Nó- lig -kam -e
Song line ending in odd (S):
.
X .
X.
Wi- pis -mel
SWS
to lie over an area
.
ñé -ñei
.
wa -ñ
WS
WS
wa ñ-
wípismel
ñéñe'i
hummingbird PL -song
CON 1SG'Hummingbird songs surround me.'
X
.
bí je- mi -da.
WSW S
bíjemid
to surround
These four examples show two things: 1) that lines do not consistently end in either odd or
even syllables, and 2) if song lines are characterized as trochaic, as above, approximately half of
all stresses appear in Weak positions. Six of the eleven stresses appear in Weak position. This
observation is important, although at this point, I wish to postpone discussion of it until below, as
it does not fit into the current discussion of edge effects. The relevant examples, however, show
that the fluidity of the line length makes it difficult to characterize the meter as underlyingly
trochaic, if such meter is viewed merely as alternating strong and weak positions as for Hayes.
The current problem is this: how do we characterize the fact that stresses are restricted
from these two metrical positions (second and final) on edges of song lines? Is it possible to
characterize, in one way, the similar behavior of these two positions (both on the edge, both
typically iambic and hence weak), despite the dissimilarities (one is always even, while the other
may be even or odd)? Or can it only be characterized as two separate restrictions?
I suggest that it is possible to make a unified characterization of the behavior of these two
positions, and that this characterization is imperative in view of the fluidity of line length in TO
song meter. My proposal is quite simple: (1) Trochaic feet are built on each edge of a song
line; (2) Stresses are prohibited in the Weak positions of these feet. I have formalized the two
parts of this proposal as components of the Edge Constraint in (21):
14
(21)
EDGE CONSTRAINT :
a.
All metrical song lines minimally begin and end
b.
with a trochaic foot.
*Foot
A
SW
[.. X ]
What does each part of (21) do? The statement in (21a) stipulates that the left and right
edges of song lines must consist of trochaic feet. Only feet allow the reference to the second and
final metrical positions; no other construction ensures reference to the restrictions on these
positions. The construction in (21b) prohibits stress in the Weak position of a foot. The Edge
Constraint as formulated will only apply to two metrical positions (second, final), because there
are no other feet in the meter to which (21b) can apply (at this point, I argue for the presence of
feet only at line edges).
Especially striking is the formal reference to the Foot. Without a binary foot, as in (2 1 b),
we cannot unify the reference to both positions. In fact, without a foot, it is otherwise impossible
to refer to the second position in meter. 21
The formalization in (21) accomplishes several goals: 1) It accommodates both edge
restrictions. 2) Edges are restricted without directly specifying the position, but instead by
using a linguistic unit, the foot. 3) It formalizes only a prohibition on stresses in edgemost W's;
recall that the evidence of (20) showed that stresses appear in other W positions. At this point in
the analysis, the constraint in (21) suggests that there are only two relevant W positions which
are evaluated. This can be seen in (22), where I give the song lines in (20) under the Edge
Constraint:
(22)
a.
No violation of the Edge Constraint in a song line ending in odd:
Foot
Foot
A
A
SW
SONG
[x..
pí -si -ne
SW
x.
mó -ka -me
X. .]
.
jé-je-wen
jéwed
písin
mó'o -kam
bison
head -one
'Bison Head (place).'
earth
21The only other possible way to restrict these positions would be if we characterize the line as : [S W ...W),
prohibiting stress in all W positions. This characterization would be possible if there were only edge effects in TO
song meter and the middles of lines were completely unrestricted. However, the restriction on adjacent stresses as a
condition for Vacuous Reduplication suggests that TO song meter consists of more than just edge restrictions. I will
show this in more detail in the next section.
15
(22)
b.
No violation of the Edge Constraint in a song line ending in even:
Foot
Foot
A
A
SW
SW
\
[.
SONG
.
S
wa
wa
REP
CON
x
I
.
.
X
.
.]
.
s- wá- pu -si -me
wú -wa -ke
s- wápusim
PL- damply
wúwhag
to emerge
'Damply they emerge.'
c.
No violation of the Edge Constraint in a song line ending in even:
Foot
Foot
A
A
SW
SW
I
I
X
.
[X .
Nó- lig -kam -e
SONG
.
.
jé -wen -e
Nóligk 'am
jéwe d
to turn LOC
earth
'Noligk lies there on earth.'
d.
.
X
.
.
]
ká -ha -ce
ká:c
to lie over an area
No violation of the Edge Constraint in a song line ending in odd:
Foot
Foot
A
A
SW
SW
I
I
X . . .]
WI-pis-mel
ñé -ñei wa ñ bí-je-mi-da.
wípismel
ñéñe'i a ñ- bíjemid
hummingbird PL -song AUX 1SG -to surround
'Hummingbird songs surround me.'
[X
SONG
.
X
.
.
.
.
The formalization of the Edge Constraint in (21) successfully captures the generalizations of the
data and characterizes song lines.
A similar observation is true with respect to the right edge, as a line with final stress, which
would otherwise violate the Edge Constraint, instead corresponds with Vacuous Reduplication.
The correspondence between the Edge Constraint and line -final cases of Vacuous Reduplication
can be seen in (23), where I give each line in song line form and citation line form (final stresses
are underlined, with violations double underlined):
Foot
(23) a.
Foot
A
A
SW
SW
\
SONG
.
.
[X
X . . .
Jiós
ká- wu -li -ki
yam-e
of
Jiós
'o
i- gáwulk
'am
God AUX INIT -to differ
LOC
'God starts to differ standing there.'
16
x
I
.
.]
ké -he -ka
MA
to stand
(23)
a.
Foot
Foot
A
A
S
SW
[x
CIT
Jiós
b.
.x
.
'am
i-gáwulk
'o
x
I
.
ké:k
Foot
Foot
A
A
SW
SW
I
I
SONG
[X
.
.
X
.
Nó -lig -kam -e
.
X
.
Nó -ligk 'am
jé -wed
to turn LOC
earth
'Noligk lies there on earth.'
to lie over an area
Foot
A
SW
SW
I
x
.
Nó-ligk 'am
CIT
kác
A
[x.
.1
ká -ha -ce
jé -wen -e
Foot
.
x
I
.
jé-we d
kác
In (23) we see that the crucial violations of the Edge Constraint come in those lines where a
stressed monosyllabic word falls line -finally (CITATION) The actual song lines (SONG) are
.
without violation because the final word reduplicates. Vacuous Reduplication clearly prevents a
violation of the Edge Constraint.
Let me now summarize this section. I have isolated the generalizations about edge effects,
and proposed the Edge Constraint to resolve these effects. The Edge Constraint has two parts:
1) It builds trochaic feet at each edge of a song line and 2) It prohibits stresses in the weak
positions of feet. By building feet only at the edges, I have made a proposal which unifies the
edge effects under one treatment. This analysis also has the benefit of allowing a
characterization of the restriction on the final metrical position, which can be either odd or even,
but is never filled with a stressed syllable. Finally, this analysis is compatible with the flexible
song line.
2.2 An Account of Adjacent Stresses
Let me review the relevant facts about Vacuous Reduplication first. Recall from the first
section that a stressed monosyllable will reduplicate when to the left of a stressed syllable or
occurring line -finally. As the Edge Constraint treats only those stresses appearing line -finally,
the behavior of adjacent stresses must be accounted for. Crucially, Vacuous Reduplication
creates an unstressed syllable between two stressed ones.
I propose that Vacuous Reduplication enables the meter to construct a binary trochaic foot
which does not violate the Edge Constraint. Again, the proposal is simple: by adding a
principle which builds binary feet from stresses, the Edge Constraint will rule out any illicit feet.
However, for the Edge Constraint to apply, there must be feet. Recall that from the earlier
discussion, I showed the difficulty of characterizing the final metrical position with the meter
aligned to alternating S and W positions. In (24), I show song lines with Vacuous Reduplication
aligned to this metrical pattern:
(24)
a.
X
.
X
X
.
17
.
SONG
Wá-wai
gí-wu-lik-e
nó-no-haq
X
Wáw
X
Gíwulk
X.
CIT
S WSW S W S
SW
dó'ag
WS
S
SW
constricted
rock
mountain
'Constricted Rock Mountain'
X.
b.
SONG
CIT
Oi
S
di
S
na
só-so
W
S W
na
sws
X
kú:k
X
són
s
W
X .
kú:-ku:-re
w
perhaps
the beginning the end
'soon perhaps the beginning, the end,'
soon
c.
SONG
ha-wui
S
WS
'am
ha -wui
S
LOC
3PL -to
.
CIT
X
.
mé-he-da.
X .
ká-cim
X
méd.
W
to run.
X .
kácim
.
am
.
WS
WSW
SW
SW
sky
X
.
.
ha wu-wui hí-me-da
SWSWSW
.
.
. .
X .
ha- wu'uwui híma
SWSW SW
3PL -DIST.PL -to to move to
'They are running towards it. Towards them, the sky moves.'
For the first and third examples, with two adjacent stresses in a citation line, the first of
these occurs in a strong position and the second in a weak position. Presumably, the Edge
Constraint will rule them out and Vacuous Reduplication will create the legal unstressed syllable
for the weak position.22 Note that the song lines in (24) also place stressed syllables in strong
positions. However, (24c) is problematic for such an account. In this example, both in the
citation and the song line, it is the first stress of two adjacent stresses which falls in the weak
position. The Edge Constraint is violated in both lines, despite the efforts of Vacuous
Reduplication. This shows that the characterization of the meter as a string of alternating S and
W positions is not correct. The correct characterization must consistently place the second of
two adjacent stresses in weak position. I have shown previously that stresses will appear in odd
(S) or even (W) positions if the song meter is characterized as alternating S and W positions.
Here, I propose that the proper treatment of the metrical pattern is one which allows all
stresses to appear in S position. This gives an adequate formal characterization of the prohibition
on adjacent stressès; these stresses head binary feet. Binary feet consist of two positions, and
the Edge Constraint will rule out any foot made up of two stressed syllables.
To achieve this effect, the meter requires one more constraint, which will effectively build
binary feet wherever stresses appear. In (25), I formalize this rule:
22The nature of Vacuous Reduplication deserves a comment. As it is prefixal, it actually allows the stress to
move from one syllable to the newly created one. This has the ultimate effect, however, of creating a weak position
in the meter. Also, the phonological changes found in songs, especially those related to features, nasalization, and
lenition, are often reflected in the reduplicated forms.
18
(25)
BINARY FOOT CONSTRAINT:
a. All stresses appear in the strong position of a foot.
b. All feet are binary (following Hayes 1987).
With the proposed Binary Foot Constraint (BFC) of (25), we also require a formalization of
Vacuous Reduplication to reduplicate and prefix syllables that result in the eventual filling of the
necessary weak positions by the BFC. I give this formalization in (26) (following McCarthy and
Prince (1990)). The formalism creates a prefixal reduplicant wherever a vacant position appears
in a foot in the meter.
(26)
Vacuous Reduplication: (In this rule, On typifies the forms which
participate in this process; that is, stressed monosyllables.)
= E-( són: (1)) *má: s /'
:cto (On)
= E- (són1)
Boson
where
Foot
A
SW
x
The rule in (26) will reduplicate monosyllables wherever they are not followed by an
unstressed syllable. Note that stressed syllables cannot fill these positions, as this violates the
second half of the Edge Constraint (21b). In the data in (27), I show the effects of the proposed
BFC and Vacuous Reduplication. Note the empty weak positions (underlined) that result from
the ban on a stress in weak position:
(27)
a.
Foot
Foot
A
A
Foot
A
SW SW
X. x
1
SONG
1
.
.
.
.
x.
.
SW
i
.
.
we -co ná- ha -gio kc in
mém- é-li-hi-me
gíwulk
weco náhagio
wáw
kc 'in
mémelihim
rock
cinched
below mouse CONJ LOC to run to rpt
'The mouse runs around there below Cinched Rock.'
Wá -wai gí- wa -lig -e
Foot
Foot
Foot
A
A
A
S W SW
1
CIT
A
SW
I
1
Foot
X
wáw
1
1
x
I
gíwulk
A
SW
1
.
Foot
.
.
SW
i
i
x
.
.
weco náhagio
19
.
kc in
X
..
1
.
mémelihim
(27)
Foot
Foot
b.
A
SONG
.
.
.
No -lig -kam -e
Nóligk 'am
SW
I
I
I
X
A
SW
SW
I
Foot
A
x
I
.
.
I
x
.
.
jé- wen -e23
ká -ha -ce
jéwe d
ká:c
to turn (place) loc earth to lie over an area
'Noligk lies there on earth.'
Foot
Foot
Foot
A
A
A
SW
SW
I
I
I
x
X .
Nóligk
CIT
I
'am
c.
.
Foot
Foot
A
A
SW
SW
I
1
x
I
oi
na
kú -ku: -D e. (20:6)
só-so
són
oi
na
kú:g
the beginning the end
soon perhaps
'soon perhaps the beginning, the end,'
Foot
Foot
A
A
SW
I
CIT
I
ká:c
I
SONG
I
jéwe d
X.
.
SW
x
oi
na
I
X
són
SW
I
I
X
kú:k
The lines in (27) show several consequences of the second part (BFC ; Vacuous
Reduplication ) of my proposal: 1) The second of two adjacent stresses cannot fill the weak
position of the foot, following the Edge Constraint. 2) Line -final stresses cannot build binary
feet to satisfy the BFC, for they have no material to fill the weak position. 3) All stresses will
necessarily appear in strong positions. 4) The BFC and Vacuous Reduplication make redundant
part of the Edge Constraint, specifically, building feet on right edges.
It is this final consequence which suggests that the Edge Constraint should be revised to
build feet only at left edges. I give the simplified Revised Edge Constraint below in (28).
23This example also shows the ambiguity in how the meter treats jéwed The form must be scanned as a
disyllable here as it does not reduplicate.
20
REVISED EDGE CONSTRAINT : Tohono O'odham song meter is governed by the
(28)
following:
a. The left edges of song lines minimally begin with a trochaic foot.
b.
*Foot
A
SW
[..X]
To conclude this section, the significance of the lines in (27) are that they confirm the
intuition that Vacuous Reduplication is a metrical effect; the significance of the Binary Foot
Constraint is that it explains why the extra syllable of Vacuous Reduplication is generated: to
build satisfactory trochees. An additional benefit is that the BFC unifies the right edge effects,
the prohibition on adjacent stresses, and the effects of Vacuous Reduplication under the auspices
of constraints that govern the meter. Thus, the BFC augments the Edge Constraint, and a highly
complex, rigid system of meter emerges in the O'odham songs.
2.3 Consequences of the Constraints
Here, I will briefly demonstrate why each part of the two constraints is necessary to rule
out nonoccurring lines. There are three types of lines which never occur in the corpus; I give
these below, followed with sample lines that are acceptable.
Unacceptable Lines:
a.
* . X
(29)
b.
c.
*
*
.
.
X
XX
Acceptable Lines:
d.
X
e.
X
f.
X.X
g.
X.X.
Which lines are judged acceptable and unacceptable by only application of the Revised
Edge Constraint? Those in (29b -g) are all acceptable, while only (29a) is unacceptable. The line
in (29a) is exactly the type of line the Revised Edge Constraint is meant to rule out. How do the
lines fare in a comparable treatment with the Binary Foot Constraint? By the Binary Foot
Constraint, the acceptable lines are (29a,d -g) and the unacceptable ones are (29b -c). The figure
below shows this in more detail:
Binary Foot Constraint
(30) Revised Edge Constraint
*Foot
-4 Foot
A
A
SW
SW
a.
*
.
X
a.
*
*Foot
Foot
A
A
SW
b.
*
X
SW
X
b.
21
*
X
*Foot
4 Foot
A
A
SW
SW
(30)
c.
XX
*
c.
4 Foot
4 Foot
A
A
SW
SW
d.
*..X
X
d.
X
4 Foot
4 Foot
A
A
SW
SW
e.
.
.
X
.4 Foot
.4 Foot
A
A
SW
f.
X.X
f
4 Foot
Foot
A
SW SW
X. X
-4 Foot
Foot
A
A
A
S W SW
SW
g.
X
e.
.
.
.
X
.
X
g.
.
X
.
X
.
As this figure shows, the Revised Edge Constraint and the Binary Foot Constraint are both
independently needed to account for different types of ungrammatical lines. Each rules out a
different subset of the unacceptable lines. Note that for the acceptable lines, there is only overlap
in lines like (30d,f), which begin with a stressed syllable followed by an unstressed one. Here,
both of the constraints build acceptable feet. This marginal overlap, however, is a consequence
which is insignificant in view of the work both constraints do separately on unacceptable lines.
Let me quickly review the analysis presented in this section. My analysis consists of three
parts, the Revised Edge Constraint, the Binary Foot Constraint and Vacuous Reduplication. The
Revised Edge Constraint consists of two parts: 1) All metrical song lines minimally begin with
a trochaic foot, and 2) Stresses are prohibited in the weak positions of feet. This filter constrains
the meter by restricting the left edge position which does not allow stress, the second metrical
position. The Revised Edge Constraint allows us to refer to this metrical position by using the
Foot, as there is no other category which covers the set of distributional facts. I argue that the
Edge Constraint in its earliest version is further confirmed by the behavior of monosyllabic
stressed words line -finally. For the Revised Edge Constraint, the manipulation of the stressed
syllables by the morphology shows that the avoidance of the two positions is significant, and not
that they were merely overlooked.
I further argued that the critical effect of Vacuous Reduplication is to create an extra
syllable which results in separating two stresses by an unstressed syllable. This effect can also
be characterized under the Revised Edge Constraint, if we make one crucial modification to the
analysis. I proposed that the Binary Foot Constraint effectively makes a binary foot for each
stressed syllable. Here again we see how the Revised Edge Constraint rules out any foot with a
stress in the weak position. Binary feet ensure that stressed monosyllables will never appear
line -finally, as they create degenerate feet. 24 Thus the Binary Foot Constraint allows both a
24The meter does show surface dactylic (S W W) effects. This is seen at the right edge with respect to the
extra vowels, as pointed out to me by Leanne Hinton and Gilbert Youmans. Forms such as ké:k surface as a dactyl
when line -final: kéheka. Note that a dactylic account fails when we consider the fact that trisyllabic words, such as
bíjemid, to surround' ends with an extra vowel in the songs: bí je -mi -da (36:1). Also, vowel -final disyllables may
22
characterization of the distributional facts (no adjacent stresses, stress may fall in either even or
odd metrical positions), as well as an account of Vacuous Reduplication which coheres with the
entire metrical system of the songs.
This analysis of Tohono O'odham song meter enriches our understanding of metrical
theory: There is no other way in which we can characterize the data presented here; the metrical
categories of foot and line are strongly motivated, and indeed, we are able to construct a metrical
theory using rather simple theoretical entities. Further, the system argues for strict binarity in the
representation of the foot in meter.
3. On Meter and Universals
In this section I will discuss Hayes' theory of meter with respect to the analysis presented in
the previous section. Hayes (1989) claims that metrical requirements are lax line -initially. The
analysis of TO requires modification of the typological claims made in Hayes (1989), as
O'odham metrics requires strict left edge metrical constraints.
Let me review Hayes' rule typology. Three rule types make up 'an exhaustive typology of
the ways in which metrical rules may refer to bracketing. A given rule may belong to more than
one type' (Hayes 1989: 246). In (31), I list and define these three types:
(31)
a.
BOUNDING RULE: 'considers only those peaks25 that are defined within a
given peak in a snapshot of that category.' (p. 245)
b.
RIGHT EDGERULES: 'apply to rule out structures of the following
form:
[D... Peak]
W
where 'D' is a specified prosodic domain, 'Peak' is a peak in metrical W
position defined within D, and '...' is material included in D that the rule
may optionally specify. The claim here is that the right edges of prosodic
categories are often scanned with special strictness.'(p. 245)
c.
LEFT EDGE RULES: 'apply to configurations of the form
[D Peak...]
W
where 'Peak' and 'D' and '...' are defined as before. The difference here is
that left edge rules, rather than forbidding a specified cadence, may
overrule other metrical rules, licensing cadences that would otherwise be
ill -formed.' (p. 245)
An interesting situation arises when we consider the Revised Edge Constraint. If we allow
a loose definition of domain here, such that Foot and Line are allowed, the Revised Edge
Constraint (28a) contradicts Hayes' claim for left edges as lax, as it provides that a foot must be
built on the left edge of a line (to prevent a stress in weak position). This is never violated in the
songs, robust evidence that left edge rules for TO song meter are strict. The more general part of
the Revised Edge Constraint (28b), which prohibits stresses in metrical W, is consistent with
receive no additional syllables, as when kak -ke (20:3) to ask', surfaces line -finally without change. The
formalization of Vacuous Reduplication could easily be changed to produce dactyls for these types of words, as well
as for adjacent and line -final stresses. The effects of Vacuous Reduplication clearly support a binary analysis.
Finally, the form káidaghim, 'resounding noise -CONT' truncates into a disyllable when it ends a line: káim -he
(24:6). These facts are incompatible with a dactylic analysis of the meter.
25PEAK: any syllable with a higher grid column than AT LEAST ONE of its neighbors (Hayes 1989, 227).
23
Hayes' formulation of right edge rules. But crucially, it is the instantiation of this for the initial,
left edges of song lines which contradicts his typological claim of left edge rules as lax. The
necessity of building a strict trochaic foot line -initially presents the argument for a revision to the
typology of rules to allow strict left edge rules.
Finally, I would like to discuss the theoretical role played by the categories motivated in
this paper: Foot and Line. Hayes notes that the bracketed units Line, Colon, and possibly Foot
are thus supported by the metrical rules that must refer to them' (1989: 256). A central point of
this paper is the motivation of the Foot as a category referred to by metrical rules. Kiparsky
(1977), 26 Prince (1989), and Youmans (1989) (and others) also provide evidence for the Foot.
The second comment is that while Hayes acknowledges that the Metrical Hierarchy (made up of
Line, Colon, Foot) plays a role in meter, he observes that the metrical rules referring to
categories of the Metrical Hierarchy follow the typology of metrical rules which he proposes.
Again, however, I argue that the study here shows that line -initial strictness must be allowed.
4. Conclusion
In this paper, I have argued for a number of points. Let me review them, starting with
those of a more descriptive nature. The data involved in this study is very complex, and I have
attempted to organize the various phenomena in the songs according to their relevance to the
meter, as that is the focus of this paper. Tohono O'odharn songs are made up of lines which are
flexible in some ways, and rigid in others. The line is flexible in allowing a relatively
unconstrained number of stresses and syllables.
Vacuous Reduplication shows itself to be metrically motivated. I have shown that it
involves the systematic manipulation of the morphology to avoid adjacent or line -final stresses.
The descriptive facts of Tohono O'odham songs reveal the dual nature of the line: it is
flexible in terms of the number of stresses and syllables; it is rigid in terms of regulating where
stresses may appear. The closest parallel can be found in Old English poetry, such as Beowulf
The meter of Beowulf has typically been described as a line which consists of two verses, which
in turn are made up of two main stresses and an unspecified number of more weakly stressed
syllables.27 However, unlike TO song meter, the meter of Beowulf can be characterized by the
regularity of the number of stresses.
In O'odham song meter, the regularity is derived differently; it is derived through
restrictions on where main stresses may appear. Unlike other meters which regulate this
characteristic, Tohono O'odham song meter does not regulate line length. Like Beowulf, this
results in a meter which allows an indeterminate number of syllables of lesser prominence. The
characterization of the O'odham song meter may exemplify a novel system of meter. This
finding alone is an important one for the typology of versification systems.
The metrical system can be captured in an analysis which relies on the foot. The foot based analysis also provides for the line's flexibility. The O'odham meter shows that these
components are necessary for generative metrics.
I have made two specific proposals the cornerstone of my analysis. First, I argued for the
Revised Edge Constraint, which proposes that the prohibition on stressed material in the second
metrical position, can only be handled by building a foot at the beginning of each song line and
restricting stresses from appearing in the weak position. Second, I showed that the additional
restrictions on the line, namely the prohibitions on stresses appearing adjacent to each other or
line- finally, can be derived by the Binary Foot Constraint, which states that all feet are binary
trochees and all stresses must appear in strong position of feet.
Under my analysis, the flexibility in song lines comes from the fact that the meter only
regulates stressed syllables; weak syllables are relevant only when they are incorporated into
26However, see Hayes (1983) for arguments against Kiparsky's analysis.
27However, both Cable (1974) and Russom (1987) argue against this 'textbook' characterization of the meter,
and propose their own alternatives.
24
feet by the Revised Edge Constraint or the Binary Foot Constraint. All other weak syllables in
the line are unrestricted. This results in the variability of line lengths seen in the songs.
Three central theoretical points have been developed here. First, I show that the line may
be flexible in some poetic traditions. Second, I argue that binary feet, with constituency, are
needed in this line. Finally, I argue that Tohono O'odham songs attest to the existence of strict
metrical rules for left edges. The necessity of beginning each line with a foot, to ensure the strict
enforcement of no stresses in the second metrical position, provides the impetus to revise the
typology of meter presented in Hayes (1989): He argues for a typology of metrical rules that
states left edge rules are lax. The importance of this study from a typological perspective is
evident.
In conclusion, this study has isolated certain problem areas in generative metrics, as well as
providing evidence for a novel system of versification. By using data that from Tohono
O'odham, a Native American language steeped in the oral tradition, we can see not only where
metrical theory is lacking, but also recognize the unique properties of the system of meter used in
these songs. The metrical system is rooted in three simple principles (the Revised Edge
Constraint, the Binary Foot Constraint and Vacuous Reduplication) which produce the intricately
organized metrical pattern, the meter of Tohono O'odham songs.
25
References
Attridge, Derek. (1982). The rhythms of English poetry. Longman House: Essex, United
Kingdom.
Bahr, Donald. (1980). Four Papago Rattlesnake Songs. Anthropological Research Papers, No.
20, p. 118 -126. Tempe: ASU.
Bahr, Donald. (1983). A Format and Method for Translating Songs. Journal of American
Folklore 96: 380, p. 170 -182.
Cable, Thomas. (1974). The Meter and Melody of Beowulf Illinois Studies in Language and
Literature. Urbana: University of Illinois Press.
Cable, Thomas. (1991). The English Alliterative Tradition. Philadelphia: University of
Pennsylvania Press.
Chen, Michael. (1979). Metrical Structure: Evidence from Chinese Poetry. Linguistic Inquiry
10, 372 -420.
Chesky, Jane. (1943). The Nature and Function of Papago Music. Unpublished M.A. thesis,
Tucson: The University of Arizona.
Crystal, David. (1991). A Dictionary of Linguistics and Phonetics. Cambridge, Mass.: Basil
Blackwell.
Fitzgerald, Colleen. (1993). Too Many Vowels: The Phonology of Syllables in Tohono
O'odham Songs. Paper delivered at WECOL XXIII.
Haefer, J. Richard. (1977). Papago Music and Dance. Occasional Papers, Vol. 3 (Music and
Dance Series) No. 4, Navajo Community College Press.
Haefer, J. Richard. (1981). Musical Thought in Papago Culture. Ph.D. dissertation, University
of Illinois, Urbana, Illinois.
Halle, Morris and Samuel J. Keyser. (1971). English Stress: Its Form, Its Growth, and Its Role
in Verse. New York: Harper and Row.
Hammond, Michael. (1988). Constraining metrical theory: A modular theory of rhythm and
destressing. New York: Garland.
Hammond, Michael. (1991). Poetic Meter and the Arboreal Grid. Language 67: 2.240 -259.
Hayes, Bruce. (1983). A grid -based theory of English meter. Linguistic Inquiry 14.357 -93.
Hayes, Bruce. (1987). A Revised Parametric Theory. Proceedings of the Northeastern
Linguistic Society 17.
Hayes, Bruce. (1989). The Prosodic Hierarchy in Meter, in Rhythm and Meter, P. Kiparsky,
and G. Youmans, eds. 210 -260.
Hayes, Bruce. (1993). Metrics as an Optimization Problem: The Case of English Folk Songs.
Rutgers Optimization Workshop.
Hill, Jane and Ofelia Zepeda. (1992). Derived words in Tohono O'odham. IJAL 58.355 -404.
Hinton, Leanne. (1980). Vocables in Havasupai music. C.J. Frisbie, ed., in Southwestern
Indian ritual drama. Albuquerque: University of New Mexico Press.
Hinton, Leanne. (1984). Havasupai Songs: A Linguistic Perspective . Tubingen: Gunter Narr
Verlag.
Hinton, Leanne. (1990). Song Metrics. Proceedings of the Berkeley Linguistic Society,
Parassession on Native American Languages 17b.
Hymes, Dell. (1981). 'In vain I tried to tell you.' Philadelphia: University of Pennsylvania
Press.
Kiparsky, Paul. (1968). Metrics and Morphophonemics in the Kalevala, in C. Gribble, ed.,
Studies Presented to Professor Roman Jakobson by His Students. Cambridge, Mass.:
Slavica.
Kiparsky, Paul. (1975). Stress, Syntax and Meter. Language 51.576 -616.
Kiparsky, Paul. (1977). The Rhythmic Structure of English Verse. Linguistic Inquiry 8, 189441
Maling, Joan. (1973). The Theory of Classical Arabic Metrics. PhD Dissertation, MIT.
Mathiot, Madeleine. (1973). A Dictionary of Papago Usage. Indiana University: Bloomington.
26
McCarthy, John and Alan Prince. (1990). Foot and word in prosodic morphology: the Arabic
broken plural. NUIT 8: 209 -284.
McCarthy, John and Alan Prince. (1993). Prosodic Morphology I. University of Massachusetts,
Amherst and Rutgers University ms.
Prince, Alan. (1989). Metrical Forms, in Rhythm and Meter, P. Kiparsky, and G. Youmans, eds.
45 -80.
Prince, Alan and Paul Smolensky. (1993). Optimality. Rutgers University and University of
Colorado ms.
Russom, Geoffrey. (1987). Old English Meter and Linguistic Theory. New York: Cambridge
University Press.
Saxton, D., Saxton, L. and S. Enos. (1989). Dictionary: Papago/Pima- English, English Papago/Pima. The University of Arizona Press: Tucson.
Underhill, Ruth. (1938). Singing for Power: The Song Magic of the Papago Indians of
Southern Arizona. Berkeley: University of California Press.
Wallace, Margaret (transcriber). (1981). O'odham Ha -Neñei (Papago Songs). San Simon
School, Coppei House Publisher: Waitsburg, Washington.
Youmans, Gilbert. (1989). Milton's Meter, in Rhythm and Meter, P. Kiparsky, and G. Youmans,
eds. 45 -80.
Zepeda, Ofelia. (1988). A Papago Grammar. Tucson.: University of Arizona Press.
27
28
Arapaho Accent'
Amy Fountain
The University of Arizona ®
Department of Linguistics
1.
Introduction
Arapaho is an Algonquian language spoken by a population of about 3500 in Wyoming and
Oklahoma (Salzmann 1983). The accent system of Arapaho is quite complex and presents a
challenge to any theory of stress/accent which attempts to account for these phenomena in a
derivational manner (Salzmann 1965, Tsay 1989). In this essay it is argued that Arapaho accent
involves both lexical and derivational aspects. In section 2, the phonetic characteristics of Arapaho
accent are outlined. Section 3 briefly overviews Idsardi's (1992) theory of the computation of
stress. In section 4, the Arapaho data are presented and the crucial generalizations are stated.
Section 5 contains an analysis of these facts, utilizing Idcardi's theory. An alternative analysis is
offered in section 6, and finally in section 7 the theoretical implications of the Arapaho facts are
discussed.
2.
Arapaho Accent
Prominence in Arapaho is realized phonetically through both pitch and fortition of
articulation. Some Arapaho words contain a single accent (i.e., he:sno.hni 'efamine')2, others are
characterized by seemingly binary foot structure (i.e., niacó nemí 'lantern'). Accents may be
realized with a falling tone pattern, as in ci:té: 'foam', and these accents are marked with a hacek.
Non -falling accents are realized primarily through fortition (Salzmann 1983).
3.
Idsardi's Theory of the Computation of Stress3
Idsardi (1992) provides an algorithm for the construction of metrical grids which allows for
the parsing of these grids into domains which may be bounded or unbounded. In this system, five
parameter values are set for a given language. First, stress- bearing elements (moras or syllables)
project markers (x's) onto line 0 of the metrical grid. Second, a left or right parenthesis may be
projected onto line 0 at the left or right edge of certain kinds of elements (e.g., heavy syllables).
Third, a left or right parenthesis may be projected at the left or right edge of the left- or rightmost
element This third step is referred to as "edge -marking ". Edge - marking settings are listed in
abbreviated form where the first letter represents the choice of parenthesis type, the second the
choice of element edge and the third the choice of domain edge. The edge -marking setting RLR is
to be read "project a right parenthesis to the left of the rightmost element in some domain ".
Edge - marking is the mechanism which allows Idsardi's system to account for
extrametricality effects, as well as pre- and post- stressing cases (where an unaccented domain
i Special thanks to Jane Tsay and Mike Hammond for their comments and suggestions in the preparation of this
manuscript, and to the editors of this volume for their additional suggestions. All errors are the responsibility of the
author.
2The official orthography of the Arapaho tribe is adopted here. Phonetic correspondences are as follows.
Consonants: [c] is a spread -glottis palatal affricate; [3] is a spread -glottis interdental fricative. All other consonant
symbols are used in the standard manner. Vowels: [e] is a mid, front, lax vowel; [i] is a high, front, lax vowel; [o]
is a mid, back, tense vowel; [u] is a high, back, lax vowel.
3The interested reader is referred to Idsardi (1992) for a full explication of this theory, and motivation for the
parameters and constraints suggested within it. All aspects of Idsardi's theory which are necessary to this analysis are
overviewed here, however.
29
seems to force an accent to appear on an adjacent, accentless, domain). Edge marking can occur
on line i as well as on line O.
The fourth parameter, "iterative constituent construction" (henceforth, ICC), allows
iterative binary parsing, either from the right edge leftward or vice- versa. Under the ICC, the type
of parentheses projected onto line 0 is constrained by the directionality of the parse. A left-to-right
parse always places right parentheses on line 0, whereas a right -to-left parse always places left
parentheses.
Finally, languages may mark heads on line 1 either at the right or the left edge of the
domains identified by parenthesis projection/insertion. These parameters are applied to a simple
stress -type system (Warao) in (1) below. In Warao, stress falls on even numbered syllables
counting from the right edge of the word. Main stress falls on the penultimate syllable (Idsardi
1992:6).
(1)
The Warao Parameters:
Line 0:
Line 1:
Example:
yapurùkitàneháse
Project x, Edge:RRR
x x x x x x x x)
yapurukitanehase
ICC:R
(x x(x x(x x(x x)
yapurukitanehase
Head:L
x
x
x
x
(x x(x x(x x(x x)
yapurukitanehase
Line 1: Edge:RRR
x
x
x
x)
(x x(x x(x x(x x)
Edge:RRR
Edge:RRR
ICC:
R
Head:L
Head:R
yapurukitanehase
Head:R
x
x
x
x)
x)
(x x(x x(x x (x x)
yapurukitanehase
Idsardi introduces a number of constraints on the construction of the grid. Constraints take
the form of "avoid" clauses, and these disallow the construction of certain configurations on the
grid. Ternary systems like Cayuvava, for example, are subject to "avoid (xx( ". This causes the
ICC to place a parenthesis one marker to the left of the marker that would have been selected under
binary construction. An example of the application of "avoid (xx(" is included in (2). In
Cayuvava, stress falls on every third mora counting from the right edge of the word, and on the
initial mora in shorter words ( Idsardi 1992:27).
30
(2)
Edge:RLR
The Cayuvava Parameters:
Line 0:
Example:
maráhahaéiki
Project x; Edge:RLR
x x x xxx)x
marahahaeiki
ICC:R
x(x xx(xx)x
marahahaeiki
Head:L
x
ICC:R Avoid (xx(
x
x(x xx (xx) x
marahahaeiki
With this brief explication of Idsardi's theory in mind, section 4 will examine the Arapaho
data.
4.
Arapaho Nouns
The data analyzed here include only unaffixed nouns. Arapaho nouns show a wide variety
of patterns of accent Monosyllables are never accented, but all polysyllabic words contain at least
one accent. Two- syllable words can exhibit an accent on the first or second syllable, or both, and
the locus of accent is unpredictable according to syllable weight or mora count. This is illustrated
in (3) below.
(3)
Mono- and disyllables:
Monosyllables: Unaccented
ho3
'arrow'
Disyllables: Initial Accent
yó:kox
bétson
'willow'
'elbow'
Disyllables: Final Accent
wonót
'abdomen'
bi:11í3
'dung'
Disyllables: Initial and Final Accent
363
'cartilage'
It is always the case that adjacent nonfalling accents occur on a CV.VC sequence. Falling
accents only occur on long vowels which immediately precede an accented, consonant- initial
syllable. Examples are given in (4).
(4)
Adjacent Accents:
3é13
célto:
cîrté:
nih'êmó'
'cartilage'
'earring'
'foam'
'blackbird'
31
Three- syllable words contain one or two accents, and all of the possible accentual patterns
are attested. The data in (5) contain examples of accentual patterns in three- syllable words.
(5)
Three- syllable Words:
Initial Accent
né:so:tox
nóxkuhu:t
'eight'
'button'
Medial Accent
ni:zi:wo:
tecéno:
'handkerchief
'gate'
Final Accent:
he:netí:t
'language'
Two Accents:
célto:
sé'temé3
nih'êmó'
'earring'
'blood hound'
'blackbird'
Forms longer than three syllables show all possible arrangements of accents, subject to the
following restrictions. First, all four -syllable spans have at least one accent. Second, only long
vowels can receive falling accents, and these may only surface when they precede accented short
vowels. Third, all adjacent non -falling accents occur on CV.V sequences. Example forms are
given in (6) through (9) below.
(6)
Four- Syllable Words:
ni:s13oó
'job'
tébexóno:
ni:36:te:'e:
wohomóno:k
ninôacúwut
31'o:ku:36:'
k&lhuyo:'
'chainsaw'
'braid'
'thread'
'governor'
'fencepost'
'honey'
32
(7)
Five -Syllable Words:
tóno'wfl:hóe
hé:tese:'éi:t
touhóalo'ó
cinóu:tono
hehi:sé3oxio
hi:hó:cebíte:
séí'ko:fi'i
(6)
Six -Syllable Words:
he:téci:'éíhi:
bí:se:ni3ó:teyo:
woníseine1ú:s
wonóhno'kútemó:
(7)
'cellar'
'bald eagle'
'boot'
'grain'
'laundry'
'butter'
'cabbage'
'gull'
'chrysalis'
'burr'
'han drake'
Seven -Syllable Words:
ce:céíbecéinó:
hó:wohóo:htéíhi
'candy'
ko:kóízni:sóúhu'
'dime'
'centipede'
A number of generalizations have emerged from these data. First, there is no span of
unaccented syllables longer than 3 syllables. Second, falling accent only surfaces on long vowels
which precede accented syllables. Third, adjacent non - falling accents only occur on CV.V
sequences. Fourth, both alternating (bounded) and non -alternating (unbounded) accent patterns
occur. Fifth, neither syllable weight nor position in a word predict placement of accent. Finally,
monosyllabic forms are never accented. An adequate analysis of Arapaho accent must account for
each of these generalizations.
5.
The Analysis
The generalizations discussed above can be accounted for by an analysis positing that
certain components of the accent system must be lexical, while others may be derivational in
nature. It is assumed that lexical entries contain only material that is idiosyncratic about a form.
Once these lexical properties are defined, the form can be realized through the application of
language -invariant parameter settings.
That monosyllables are unaccented indicates a binary minima condition on accentable
spans. Since bimoraic monosyllabic words do not receive an accent, it is hypothesized that the
domain of stress is the syllable and not the mora. That binarity plays a critical role in the accentual
system of Arapaho is evidenced further by the fact that there must be at least one accent on every
four - syllable (two-foot) span.
Idsardi's theory allows two unique options for the analysis of Arapaho. First, the
positions in which parentheses can be inserted through edge -marking also seem to be the positions
in which lexically marked parenthesis must exist in Arapaho nouns. Since the data considered for
this analysis consist only of unaffixed nouns, the pre- and post -stressing settings (RLL and LRR)
are not attested here. Second, in the forms exhibiting alternating accents (binary feet), the
placement of accents is sensitive to the placement of constituent boundaries from lexical marking
33
(either edge -marking or adjacent accent marking). In the forms exhibiting seemingly unbounded
feet, accent always appears at least once in every bipodal string. This suggests that in some forms,
accents created by the ICC may be "bled" or deleted at some stage in the derivation.
Third, adjacent stresses are unpredictable in terms of their position within a word, but not
with regard to the segmental structures on which they can occur. The first of a pair of adjacent
accents may fall on a long vowel or the second may fall on an onsetless syllable. Note that falling
accents are always followed by non -falling accents, but there is no way to predict which nonfalling accent will be preceded by a falling one. This suggests that lexical marking of adjacent
accents is accomplished through the projection of a left parenthesis (since left parentheses force an
accent to occur on a constituent to their right). It is therefore hypothesized that marked syllables
project two parentheses. Each is a left parenthesis, and one is projected on either side of a marker
in the following configuration: (x(.
A short vowel will never surface with a falling accent. If an accented short vowel is
immediately followed by an accent, it is also immediately followed by an onsetless syllable. In
other words, CV.CV strings will never receive two accents, but CV.V strings may. This situation
seems analogous to the falling accent case in that adjacent accents are licit in this configuration, and
are illicit in all other configurations except the falling accent case just described. Interestingly, the
same form of lexical parenthesis insertion posited for the falling accent cases will account for these
sequences as well. For this reason, it is asserted that all cases of adjacent accent are lexically
marked, and these are the only lexical accents which are not subject to placement by edge- marking
settings.
Words four syllables or longer which contain medial accents always show placement of
these accents on alternating syllables rightward of (and sensitive to the placement of) lexically
marked constituent boundaries. This suggests that the ICC is binary; rightward, and is sensitive to
the placement of lexical parentheses. In order to illustrate this, let us look at the four -syllable
words with penultimate accent.
If the ICC operates from left to right, it will place a right parenthesis after the second
marker on line 0, resulting in the pattern xx)xx. If it operates from right to left, it will place a left
parenthesis to the right of the second marker on line 0, xx(xx. In either case, we must assume that
there is some method for limiting such a form to a single accent, and therefore to a single metrical
constituent. This will be accomplished by an edge -avoidance constraint on the ICC. In order to
get a penultimate accent, a constituent containing the penultimate marker must be created. Only
ICC leftward accomplishes this. Further, in order for the accent to appear on the penultimate
syllable, the constituent must be left-headed. Thus the ICC must operate from right to left, and feet
thus constructed must be left-headed.
If the ICC is subject to the constraint "avoid #(x ", a number of patterns will result, and
these patterns are all attested in Arapaho words. These patterns are exemplified in (8) below. I
assume that edge -marking parameters are lexically determined, and limited to just those Idsardi
allows. I assume further that no lexical edge- marking need occur. Finally, I assume that adjacent
accents are always lexically determined, and not subject to edge -marking parameters. All forms are
subject to ICC:L
(8)
Patterns Attested in Words Four Syllables and Longer.
Four -Syllable Casesx
Edge:None
Adjacent Accents:None
xx(xx
cele ibes
'box'
x x
Edge :.None
x(x( x x
nino:xuwut
AdjacentAccents:2nd/3rd
34
'governor'
x
Edge:LLL
Adjacent Accents:None
x
x(x x
tebexono:
(x
x
x x
Edge:LLR
Adjacent Accents:None
'chainsaw'
(x (x x(x
ho u3o o
'clothesline'
x
x
x (x x(x x
hi:ho:cebite:
'butter'
x
x
x) x x (xx
he:tese:'ei:t
'bald eagle'
Five -Syllable Cases:
Edge:None
AdjacentACCents:None
Edge:RRL
AdjacentACCents:NOne
Six -Syllable Cases:
x
Edge:RRL
AdjacentACCents:None
x
x
(x x (xx (xx
be'e3einoo:
'cedar'
Seven -Syllable Cases:
x
Edge-LLL
AdjacentACCents:5th/6th
(x
x
x x
x(xx (x (x x
ho:wohoo:hteihi
'centipede'
This analysis enables us to capture all of the attested patterns except those in which footing
does not appear to be iterative. Specifically, the remaining patterns are listed in (9) below.
(9)
Non - iterative Patterns (in five - syllable or longer words):
(a)
(b)
(c)
(d)
initial and ultima accent
second syllable accent
second and ultima accent
penultimate accent
If the ICC is operative on all Arapaho forms, then it may be that medial accents are lost through a
process of tier conflation. One possibility is the operation of conflation posited by Idsardi
(1992:38), in conjunction with line 1 edge -marling. Conflation is an operation in which heads at
line 0 are deleted while heads derived on line i are retained. If conflation applies to lexically
marked forms after line 1 edge-marking, then the patterns in (9) can be accounted for. Derivations
for these patterns are given in (10).
35
(10)
Derivations Including Conflation After Line -1 Edge- marking:
x
Conflation
Edge:LLL-Linel
AdjacentAccents:None
0
x
x (xx (x x
ci:nou:tono
(x
x
Conflation
Edge:LLRandLRR-Linel
AdjacentAccents:None
(x
'grain'
x
0
x
(x
x (xx (x x (x
woniseine:hi:s
x
Conflation
Edge:LLLandLRR-Linel
AdjacentAccents:None
x
0
x
(x
(x
(x x(x x(x
sicene:woxu'
Conflation
0
x
Edge:LRL -Linel
AdjacentAccents:None
'burr'
'yucca'
x
(x
x(x x (xx
wo3onohoe
'paper'
In sum, the following information is argued to be lexically marked in Arapaho: (i) Long
vowels and vowels preceding onsetless syllables my be lexically marked with a left parenthesis to
the right and to the left of their markers on line O. Otherwise, lexical markers can occur in all and
only those positions corresponding to Idsardi's edge - marking parameter. (ii) Certain forms are
marked to undergo conflation.
Beyond this lexical information, Arapaho has selected the following parameters for the
assignment of accents: (i) The ICC operates leftward. (ii) Constituents are left-headed. (iii)
Markers are projected from syllables. This analysis predicts that for forms not lexically marked,
there should be no more than eleven distinct patterns. Critically, all of the formal possibilities are
attested in the data, and all patterns not predicted can be accounted for by lexically marked adjacent
stresses or conflation.
6.
An Alternative Analysis
Problematic forms for any analysis undertaken using Idsardi's approach are those which
fail to exhibit medial accents on syllables where it would seem that the ICC should place
constituent boundaries. One way of dealing with these problem forms is to argue that the ICC
simply fails to apply to them. Thus the lexical mark on such forms would instruct the ICC not to
apply at all, rather than invoking tier conflation. This analysis misses a number of generalizations,
however. First, it cannot account for the role of binarity in the failure to assign accent to
monosyllabic words. Second, it cannot account for the limit of three unaccented syllables in a
string; importantly this limit is not violated even in forms which do not demonstrate binarity of
footing. These facts indicate that conflation is a more adequate mechanism for explaining the lack
of medial accents in such forms.
7.
Implications
Arapaho Accent provides a rigorous test for any theory of stress or accent systems.
Although the Arapaho system appears chaotic, it is more constrained than a purely lexical analysis
would suggest. It is clearly, however, too complex for a purely derivational theory to
accommodate. The system provides interesting support for Idsardi's approach to metrical grids,
especially with regard to edge -marking parameters. If accent systems generally make use of the
possibilities provided by edge-marking, then the power of Idsardi's theory is in some part
36
justified. This analysis surely suggests that further study of systems such as the Arapaho one will
be a fruitful area of research into the diversity of stress and accentual systems in the languages of
the world.
References
Idsardi, W. (1992). The Computation of Prosody . PhD Dissertation, MIT.
Salzmann, Z. (1983). Dictionary of Contemporary Arapaho Usage . Wyoming: Wind River
Reservation.
Salzmann, Z. (1965). "Arapaho: Noun," International Journal of American Linguistics 31,3949.
Tsay, J. (1989). 'Stress' in Arapaho, (ms) University of Arizona.
37
38
Deriving Ternarity*
Michael Hammond
The University of Arizona
Department of Linguistics
0
.
Introduction
Ternary stress patterns have posed a problem for a parametric metrical theory for some
time. In this paper, it is argued that ternary systems can be derived in an explanatory fashion from
binary systems. The basic idea is that ternary stress systems can be analyzed as binary stress
systems if the theory of extrametricality is enriched. Two specific proposals regarding
extrametricality are made. First, extrametricality must be tolerated not just at the edge of
morphological and syntactic constituents, but also at the edge of phonological constituents.
Second, extrametricality can be lost if adjacent feet are subminimal.
The organization of this paper is as follows. First, the foot typology is briefly reviewed.
Then the theory of extrametricality is presented. It is argued that regardless of the analysis of
ternary systems, the theory of extrametricality must be enriched as outlined above. Four metrical
systems are then considered: Cayuvava, Chugach, Winnebago, and Estonian. Each of these
systems provides arguments for deriving ternarity as proposed here.
i
.
Foot typology
For convenience, the metrical theory proposed by Hayes (1987) is adopted.' Hayes
maintains that there are three metrical constituents: the syllabic trochee, the iamb, and the moraic
trochee.
(1)
syllabic trochee
iamb
moraic trochee
[[a]ß]
HI-Li µl
The syllabic trochee is a left- headed constituent with syllables as terminals. The iamb is a right headed foot where the left terminal is monomoraic and the right terminal is a syllable or a mora.
The moraic trochee is left -headed and takes morae or monomoraic syllables as terminals.
There are also systems that exhibit superficially ternary iteration. Halle & Vergnaud (1987)
deal with these systems by supplementing their foot typology with an amphibrachic foot. This is a
foot with three terminals and the head in the middle. It will be shown that such a foot cannot
capture the range of ternary iteration and misses central generalizations about ternary systems?
2.
Extrametricality
Extrametricality excuses an element from metrification. An element that has been made
extrametrical need not be included in the metrical tree to be pronounced and escape stray erasure.
This device is constrained by the Peripherality Condition, which stipulates that extrametricality is
only available at the edge of a domain (Hayes 1981; 1982; Archangeli 1984-1985).
Thanks to Diana Archangeli, Stuart Davis, Bruce Hayes, Mike Kenstowicz, Adrienne Lehrer, James Myers, Pat
Perez, and Curt Rice for useful suggestions. All errors of data or analysis are the author's.
1See Hammond (1990) for an alternative to this system.
2There are also systems where the stresses can fall at potentially unbounded distances from each other. These
systems are not discussed here and the reader is referred to Hayes (1981) or Halle and Vergnaud (1987) for a traditional
treatment of such systems. See Prince (1985) or Hammond (1990) for another proposal.
39
English nouns provide an example of extrametricality. Oversimplifying, stress falls on a
final long vowel. Else, stress falls on a long or closed penult. Else, stress falls on the antepenult.
(2)
long ultima
kangaróo
Tennessée
tirade
repúte
brigadóon
chimpanzée
light penult
América
cinema
asparagus
metrópolis
javelin
vénison
long penult
aróma
balalaika
hiatus
horizon
thrombósis
coróna
closed penult
veranda
agénda
consénsus
synópsis
amalgam
uténsil
Secondary stresses fall on alternating syllables leftward regardless of syllable weight.
(3)
Mississippi
Apalàchicóla
Cònestóga
dèsignation
sèrendípity
Srìràngapátnam
hàmamèlidánthemum
còmpensátion
To account for the basic pattern of stress, the final rhyme is made extrametrical if it is not
long. Then a single moraic trochee is built on the right edge of the word followed by syllabic
trochees from right to left.3 Some sample derivations are given below .4
(4)
XXXX
x
x x x<x>
America -> America
X
x(x x)<x>
->
meri ca
X
(x) (x x)<x>
->
A meri ca
X
X X
(x)(x) <x>
x X <x>
x (x) <x>
x x x
consensus -> consensus -> consensus -> consensus
x
x x (x)
x x x
anecdote -> em. n/a -> anecdote ->
x
x
(x x) (x)
anecdote
Final extrametricality captures the fact that ternarity is exhibited only at the right edge of the
word. The moraic trochee captures the fact that only the rightmost stress is sensitive to syllable
weight.
The restriction against making a long vowel extrametrical is a natural one. There are a
number of languages where heavy syllables of various sorts are immune to extrametricality.5 The
account of ternary footing to be offered below hinges on the fact that extrametricality may be
blocked by syllable weight.
The account to be presented here depends on several other properties of extrametricality.
First, the domain peripherality refers to can be smaller than the domain footing applies to. Second,
extrametricality can be assigned before or during the footing process. Third, extrametricality can be
3There are a number of ways of effecting this differential sensitivity to syllable weight. See, for example, Halle and
Vergnaud (1987) and Hammond (1990). Since the precise mechanism used to achieve this effect is irrelevant, a
simple analysis, which allows feet to be constructed noniteratively, is given in the text. (See Kager, 1989 for a
different view.)
The notation in the text is adopted for typographical convenience. In relevant respects, it is a notational variant of
the "lollipop" notation developed in Hammond (1984/1988). See Hammond (1987) for discussion.
5See Hayes (1981) and Halle and Vergnaud (1987) for examples.
40
lost if an adjacent foot is subminimal. Fourth, extrametrical elements are invisible to clash. Each of
these points is considered below.
It is suggested that the domain of peripherality can also be the foot. That is, extramerical
syllables can satisfy peripherality merely by occurring at the appropriate edge of a foot. That the
domain of peripherality should differ from the domain of scansion is not a novel proposal. Prince
(1985) proposes this in his treatment of English compounding. He suggests that the right sister of
each compound constituent should be extrametrical. In the domain of syllabification, Rubach and
Booij (1990) have proposed that an extrasyllabic element can satisfy peripherality at the edge of a
medial syllable.
Second, the account to be presented here depends on the possibility of assigning
extrametricality at various points in the derivation. This too is not a novel proposal. The normal
picture, of course, is that extrametricality is constructed on the fly, as feet are built. However, there
are a number of systems that have been analyzed with extrametricality assigned before footing, e.g.
Yawelmani (Archangeli 1984 -1985) and English (Hayes 1981). In both of these cases, lexical
extrametricality is presumed to be present in the underlying representation of certain forms.
The third assumption that is critical for this proposal is that extrametricality can be lost if an
adjacent foot is subminimal. Prince (1991) argues that just such a process is involved in the Latin
phenomenon known as Brevis Brevians, whereby in a disyllabic form consisting of a short open
syllable followed by a long syllable, the long syllable shortens, e.g. ego: -> ego, cito: -> cito, etc.
Prince argues convincingly that this is a consequence of the loss of final extrametricality and
incorporation of the final syllable in a moraic trochee when the foot would otherwise be
subminimal.
(5)
x
x
(X) <X>
e go: ->
(x X)
ego
It will be argued below that analogous processes affect extrametricality when it is assigned at the
edge of a foot.
A final property of extrametricality is that syllables are invisible with respect to rules of
rhythm and destressing. It will be shown in the following that the English Rhythm Rule is
considerably simplified on the assumption that extrametrical material is invisible to rules of rhythm
and destressing.
Consider the following facts concerning the English Rhythm Rule. As noted by Hammond
(1984/1988), forms like the following readily undergo a shift of stress.
(6)
Tènnessée
Kàngaróo
Ténnessèe Tim
Kángaròo Cárl
Hayes (1984) notes the following contrast. The forms in (7a) undergo rhythm much more readily
than the forms in (7b).
(7)
a.
Mìssissíppi
Míssissìppi Mábel
Analytic thóught
the Pássamaquòddy vérb
?Mínneàpolis Míke
?Analytical thóught
?the Pótawàtomi vérb
Analytic
Pàssamaquóddy
b.
NTinneápolis
Analytical
Pòtawátomi
The contrast falls out automatically given that extrametrical syllables are invisible to metrical
structure and that the shift of stress is triggered by clash. Representative input representations to
the Rhythm Rule are given in (8) where clashes are marked with hyphens. (A clash exists at some
41
level n of the representation if columns are adjacent at level n and level n -1.) Notice that the cases
where rhythm is more likely are characterized by having more clashes .6
(8)
X
X
x - - -x
(x
(X
x)-(x)
X)(x) (x)
(x
(x
Tennessee Tim
x)---(x)
x) (x) <x> (x) <x>
Mississippi Ma bel
X
(x
(x
but:
x
X
(X)
)
X) (X X) <x> (X)
Minne
apo lis Mike
This line of explanation requires the uncontroversial assumption that word -edge extrametricality is
not lost when words are concatenated syntactically.
Hayes deals with this by maintaining that the shift of stress is caused by a rule that
endeavors to place stresses at four -syllable intervals- -"the Quadrisyllabic Rule ". Shifting stress in
the cases in (6) alters the interval between primary stresses from one to three syllables. In (7a), the
interval shifts from two to four syllables. In (7b), however, the interval shifts from three to five
syllables bringing it no closer to the desired interval of four syllables. In support of this approach,
Hayes cites the following contrasts where rhythm is preferred in (9a), but less likely in (9b).
(9)
a.
b.
Alabáma
Européan
Oklahóma
Alabáma
Européan
Oklahóma
Alabàma rélatives
Europèan history
Oklahòma cóngressman
?Alabàma connéctions
?Europèan histórian
?Oklahòma congréssional dist.
There are three problems with Hayes' proposal. First, there is an alternative analysis
available for the dispreference of rhythm in the forms in (9b). These forms have all undergone
initial destressing which would presumably leave a stranded 'x' which would interrupt the lower level clash, which renders the forms in (9b) analogous to the forms in (7b).
x
(10)
(x
x)
(x x) (x) <x>
Ala
ba ma
x
x
(x)
x (x) <x>
connections
(x
(x
x
x
)
(x)
x) (x x)<x> (x)
Minne
apo lis Mike
The second problem for Hayes is that there is a contrast in the following forms as well.
Rhythm is also dispreferred in (1 lb).
(11)
a.
b.
Kàngaróo
Tènnessée
Kàngaróo
Tènnessée
Kángaròo Kim
Ténnessèe Tim
?Kángaròo connéctions
? Ténnessèe congréssional dist.
6Myers (1987) and Halle and Vergnaud (1987) maintain that the adjectival suffix -ic is an exception to
extrametricality. This accounts for the fact that the preceding syllable attracts stress and undergoes Trisyllabic
Laxing, e.g. phone/phonic. If this were so, it would predict that adjectives with -ic should pattern like (7b). This
does not seem to be the case. Phrases like phílosòphic Fréd undergo rhythm readily. There are, however, other
analysis of this phenomenon that do not require that -ic be an exception to extrametricality, e.g. Yip (1987). Yip
proposes that the vowel of -ic is underlyingly absent and that shortening with -ic is an instance of closed syllable
shortening.
42
By the Quadrisyllabic Rule, the shift of stress in (11b) should actually be preferred to the shift in
(11a). In (11 a), the shift alters the distance from one to three syllables. In (1 l b), the shift would
alter the distance from two to four syllables, the optimal target by the Quadrisyllabic Rule.
Finally, the Quadrisyllabic Rule is to be avoided on theoretical grounds. It has generally
been assumed that grammatical principles do not count past two. If there is a reasonable alternative,
as has been shown above, the Quadrisyllabic Rule is to be eschewed.
Summarizing, extrametricality is subject to peripherality with respect to a stipulated
domain. In addition, it exhibits three other properties. First, it can be blocked from applying to
heavy syllables (English right -edge syllable extrametricality). Second, it can be lost to make a
minimal foot (Latin). Third, extrametrical elements are invisible with respect to clash (English
rhythm).
There is a paradox brewing here in terms of how extrametricality affects the representation.
Extrametrical elements must be available for metrification to account for the Latin facts, but
extrametrical elements are not available in the computation of what constitutes a clash environment.
This problem disappears, however, if a different representation of extrametricality is adopted.
Rather than marking it "positively" with angled brackets, assume that it is marked "negatively" by
the absence of an 'x' on line 1 of the grid. A word like Minnesota is then represented as in (12).
(12)
proposed here:
x
Halle & Vergnaud:
x
(x
x
(x
x
x) (x)
x x xx
x) (x) <x>
Minne sota
Minne so to
Such a representation provides an account for why extrametrical
Brevis elements
Brevians, but not visible for English rhythm. Rhythm depends on clash and clash depends on
adjacency at two levels of the grid (Liberman & Prince 1977). Extrametrical elements are not
represented at a sufficiently high level of the grid to matter for the determination of clash. The new
representation of Mississippi Mabel is given in (13).
(13)
x
(x
X
x
x) (x)
x
(x)
XXXXX
Mississippi Mabel
This new representation will be used in the following sections (except when discussing previous
work).
Four languages are now considered: Cayuvava, Chugach, Winnebago, and Estonian. It is
argued that each of these is more explanatorily treated in terms of foot -edge extrametricality.
3.
Cayuvava
The relevance of Cayuvava to metrical theory was first noted by Levin (1988). The data are
from Key (1961, 1967). Stress in Cayuvava falls on every third vowel counting from the right end
of the word (Levin 1988; p.101 -2).
(14)
a.
'hand'
'leaf'
'still'
dáru
éne
néA
43
b.
sákahe
ribera
óene
úhia
báau
'stomach'
'leg'
'capywara'
'you (sg.) go'
'Brazil nut'
c.
kihíBere
takáasi
soísoi
'I ran'
'old man'
'good spirits'
d.
arikájahi
Bariékimi
ariúuca
'he has already fallen'
'seed of squash'
'he came already'
e.
pópohecéBaka
Bádacaóai
ráibirínapu
'inside of cow'
'my younger brother'
'dampened manioc flour'
f.
aBárericákaA
maráhahaéiki
hiBújuruéine
'palate (Px)'
'their blankets'
'I burn it also again'
g.
ikitáparerépeha
tiBiBíoaíine
iepétiBiBóai
'the water is clean'
'she spank me again'
'they not spank me'
h.
cáadiróboBurúruce (caadáirobóirohíine)
(Bururuce) medárucecéirohíine
'ninety -nine'
'fifteen each'
The most direct way to accommodate these facts would be to build dactylic feet from right to left
(Spring 1989). Dactyls are ternary feet where the head falls on the left.
(15)
xxxxxx
x
x
(xxx)(xxx)
popoheceBaka -> popohe ceBaka
This fails to account for forms with other than 3xn syllables. For example, this analysis
would produce the following derivation for a word like ikitúparerépeha, incorrectly predicting
initial stress.
(16)
xxxxxxxx
x
x
x
(xx)(xxx)(xxx)
ikitaparerepeha -> *iki tapare répeha
To deal with these, the dactylic analysis needs a special rule to remove nonmaximal feet (feet with
less than three terminals) in words with more than one stress.
(17)
Dactylic Destressing
Remove a nonmaximal foot in a word with more than one stress.
The derivation in (16) would then continue as in (18).
44
(18)
x
x
x
x
x
x x(x x x) (x x x)
(x x) (x x x) (x x x)
-> ikitapare repeha
iki tapare repeha
Levin (1988) rejects this analysis. She proposes instead that the final syllable is
extrametrical and that amphibrachs are constructed from right to left. Halle and Vergnaud (1987)
adopt this analysis as well.
(19)
X
X
X
x x x x x x x x
(x) (x x x) (x x x) <x>
ikitaparerepeha -> *í kitápa rerépe ha
Again, a special rule is needed to remove degenerate feet (feet with only
with more than one stress.?
(20)
Amphibrachic Destressing
Remove a degenerate foot in a word with more than one stress.
A problem for both analyses is the unnatural destressing rules (17) and (20) which do not
remove clashes. Hammond (1984 /1988) argues that all destressing rules remove clashes.
This problem for the dactylic and amphibrachic analyses of Cayuvava is solved if it is
assumed that superficial dactylic feet are derived via foot -edge extrametricality. Cayuvava actually
involves trochaic footing, but a stray syllable is licensed at the right edge of each foot.
( Extrametrical elements must be at the right periphery of a foot.) In (21), a sample derivation is
given that can be compared with (16) and (19) above.
(21)
x
XXXXXX
popoheceBaka ->
(X X)
x
(X X)
XXXXXX
popoheceBaka
This approach to ternarity solves the destressing problem above. The destressing rule can
now be treated as a straightforward rule of destressing under clash. An initial binary foot under the
primitive dactyl analysis is an initial monosyllabic foot in a clash environment under the derived
dactyl analysis presented here. A sample partial derivation is given in (22).
(22)
XXXXXXXX
ikitaparerepeha ->
X - - -X
(X) (X
X
x) (x x)
XXXXXXXX
ikitaparerepeha
Cayuvava is thus best analyzed in terms of trochaic feet plus extrametricality, where
peripherality is relativized to the foot. This accounts for destressing in a natural way. While this
analysis requires a weakening of the traditional Peripherality Condition, it is in conformity with the
general principle that extrametrical elements, relativized or not, are invisible to operations sensitive
to
clash.8
Notice that if foot -edge extrametricality is assigned on the fly in Cayuvava, then relativized
peripherality is not satisfied until the next foot to the left is constructed. This is a general property
Levin proposes a special rule; Halle & Vergnaud propose their recoverability condition.
8Suggestive support for this analysis comes from the prosodic morphology of Cayuvava where there is support for a
disyllabic foot. First, the minimal word is disyllabic. Second, Key (1967) cites instances of disyllabic reduplication,
but no cases of trisyllabic reduplication have been found anywhere in the data.
45
of extrametricality, however, and not specific to this extension of peripherality. Compare the
analyses of English adjectival suffixes (Hayes 1981) and the Yawelmani suffix -zoo (Archangeli
1984- 1985). In both cases, extrametricality is in place early in the derivation, but peripherality is
not checked until later in the derivation.
4.
Chugach
Let us now consider the complex facts of Chugach, a dialect of Eskimo. The data are from
Leer (1985). There are recent metrical descriptions by Rice (1988, 1990a), and Halle (1989).
Word -initial closed syllables and all syllables containing long vowels count as heavy. In a word
with all light syllables, stress falls on the second syllable from the left and every third syllable
thereafter. Rice (1988) cites the following forms.
(23)
pa.lá.yaq
qe.ná.wik
qa.yá.kun
a.tún.'ir.túq
ta.gú.ma.lu.ní
qa.ngá.te.ra.mék
a.kú.tar.tu.nír.tuq
sa.rá.ni.wa.kár.tuq
'rectangular skiff'
'hospital'
'by boat'
'he stopped singing'
'apparently getting done'
'from a porcupine'
'he stopped eating akutaq'
'he is too sleepy'
Heavy syllables interrupt this pattern in two ways. First, heavy syllables always receive
stress. In (24) are examples of long vowels receiving stress; in (25) are some examples of initial
closed syllables receiving stress. Noninitial closed syllables do not count as heavy. In the
following examples, heavy syllables are underlined.
(24)
tuá.ta.gá
taá.taá
mu.lúk.'uút
naá.'uq
náá.qu.ma.lú.ku
mu.lú.kuút
pa.lát.kaáq
pi.lú.liá.qa
(25)
11 .'uq
ú1.1uá
án.ci.quá
án.ci.qu.kút
íßc .11u.nir.túq
q. ,.y .kun
ún.yuár.te.qu.té.ka.gá
'my father'
'her father'
'milks' (noun plural)
'it's burning'
'apparently reading it'
'if you take a long time'
'tent'
'the fish pie I'm making'
'it flooded'
'its tongue'
'I'll go out'
'we'll go out'
'he stopped lying'
'by his boat'
'I am thinking about it'
Second, a light syllable following a light syllable and preceding a heavy syllable receives stress.
(The relevant syllables are marked with double underlining.)
(26)
náá.ma.cí.quá
ig.ku.tár.tuá.nga
ág.n guá.qu.tár.tuá.nga
'I will suffice'
'I'm going to go'
'I'm going to dance'
This pattern is accounted for directly on the analysis presented here. First, following Hayes
(in progress), the fact that only word -initial closed syllables count as heavy is accounted for by
46
restricting Weight -by- Position (Hayes 1989) to word -initial syllables. Noninitial closed syllables
are treated as light by the stress system because Weight -by- Position does not apply to these
syllables. Second, iambs are built left to right. Third, each iamb licenses a (relativized)
extrametrical light syllable to its right.
The analysis is given in (27).
(27)
a.
Word -initial Weight -by- Position;
b.
c.
Build iambs left to right;
Foot -edge extrametricality of a light syllable on the right.
Some sample derivations are given below. Derivation (28) is of a word containing all light
syllables.
x
x
(28)
X)
(X
x)
(x
X X X X
X
X X X X
ga.ngá.te.ra.mék -> ga.ngá.te.ra.mék
X
Derivation (29) contains medial heavy syllables.
x
(29)
X
XXX
X
(x)
X
X
x
(x)
(x
x
X
x)
(x)
X
X X X X X
úm.yuár.te.qu.té.ka.gá
X
gm.vuár.te.qu.té.ka.gá ->
Derivation (30) shows how the light syllable effect illustrated in (26) requires some
additional machinery.
(30)
X
X X X
X
x
Ag.nctuá.qu. tár. tuá.nga ->
X
X
(x)
(x)
X
X
X
(x
X X
x)
X
X
g.ncruá.qu. tár. tuá.nga
To complete such derivations correctly, it is proposed that an extrametrical syllable is adjoined to a
following foot, subject to the condition that the resulting feet must be bimoraic.
(31)
Adjunction
(X
X
X
->
(X X)
XX
The restriction that feet be bimoraic produces the correct results. There are basically four
situations to consider, diagrammed in (32). (Heavy syllables are marked with an underposed hat.)
(32)
a.
x
X
X
(x x) -> (x x) (x)
XXX
XXX
n
A
b.
x
(X X) -> * (x
x x x
XX
x) (x)
x x x
47
C.
X
X
(x) -> * (x X)
XX
XX
A
A
d.
X
X
-> (x x)
(x)
XX
XX
Adjunction succeeds only in (32a) and (32d) because only in these cases can it produce bimoraic
feet. Adjunction in (32b) and (32c) would result in non -bimoraic feet. In (32d), adjunction has no
effect on the stress pattern; in (32a), it results in an additional stress on the medial syllable.
Additional cases are cited below where adjunction results in an automatic shift of stress.
Adjunction can be taken as an argument for deriving ternarity as proposed here. Superficial
ternarity is eliminated just in case well -formed bimoraic feet can be created. On the analysis
presented here, binary feet are part of the basic stress assignment procedure. If ternary feet were
basic in Chugach, the appearance of binary feet via adjunction would be completely mysterious.
With rule (31), the derivation in (30) continues as follows, exemplifying case (32a).
(33)
X
X
ág ncruá .
.
x
X
qu
X
X
X
. tár tuá . nga ->
.
x
(X) (X
(x)
x
x
x)
(X)
XXX
X
X
X
X
(X)
(x)
X
XX
X
(x
X)
X
X
X
ág.nauá.qu.tár.tuá.nga
X
-> ág.ncruá.qu.tár.tuá.nga
Rule (31) has additional consequences as well. Derivations like (29), with a monomoraic
foot on the right edge, undergo (31) vacuously, illustrating case (32d).
(34)
x
x
(x)
(x)
X
X
(x
X
X
x
x
x
x
x)
(x)
(X)
(x)
X
X
XXX
úm.vuár.te.qu.té.ka.gá
(X
x
x
x) (x
X
x)
XXX
X
-> íun.vuár.te.qu.té.ka.gá
Support for this analysis comes from fortition, a process whereby certain consonants
become fortis (Leer 1985). Fortition applies to the onset of a medial long syllable or a medial light
syllable immediately preceding a stressed light syllable. Fortis consonants are marked with a
prefixed ' +' in (35).
(35)
a.lí.+káa
án.ci.+quá
án.ci.+qu.küt
a.kú.+ta.mék
'she is afraid of it'
'I'll go out'
'we'll go out'
'a food'
Under the analysis presented here, these are precisely the medial foot -initial consonants.
48
(36)
(x
x
x
x
x
X)
(x)
(x)
x
x
x
X
(x
x)
X
X
X
X
X
X
X
(x)
x
x
án.ci.+quá
a.1í.+káa
(x)
x
án.ci.+qu.kút
(x
x)
(x
X)
X X
X X
a.kú.+ta.mék
Under previous approaches, ad hoc readjustment rules were required so as to alter the feet
required for stress to produce appropriate foot structure for fortition. On the analysis presented
here, no such machinery is required. The pattern of fortition is an automatic consequence of the
feet constructed, the distribution of relativized extrametricality, and the independently required
adjunction process.
Basic amphibrachic feet would fail on a number of fronts. Halle (1990) proposes an
analysis of Chugach that includes amphibrachic feet and the following machinery. First, heavy
syllables must be marked so that they occur at the left edge of a foot (a left square bracket).
Second, heavy syllables must be treated as bipositional so that the head of the amphibrach falls on
the heavy syllable (two asterisks at line 0). Finally, there is a special rule that readjusts foot
boundaries to get fortition to work out correctly:
(37)
line 0: x x)(x) -> x)(x x)
A sample derivation is given below.
(38)
x
x
x
[xx
x
[xx
xx
x
x)(x)
maa.ma.qa -> maa.ma.qa -> maa.ma.qa
x
[xx)(x
x
x)
-> maa.ma.ga
This analysis is undesirable for a number of reasons. First, it requires a number of
supplementary devices to get stress in the correct place. The derived ternarity analysis requires only
adjunction (31), which is analogous to machinery independently required in Latin. The amphibrach
analysis requires idiosyncratic boundaries and bipositionality, which are otherwise unnecessary.
Halle cites Cairene Arabic, but as shown by Hayes (1987), no such device is necessary if moraic
trochees are adopted. Moreover, enriching the theory by including bipositionality and idiosyncratic
boundaries predicts that the two devices should be able to operate independently of each other, yet
no cases of this sort occur. Including these two devices in the theory would also predict that they
should be able to cooccur with the more orthodox accent rules of Halle and Vergnaud (1987).
Again, they do not.
Second, the amphibrach analysis requires the rule (37) solely to get the fortition facts to
work out. On the analysis presented here, the fortition facts follow automatically. Finally, the
amphibrach analysis fails to account for the fundamental binary aspect of the Chugach system. The
representation of heavy syllables is enriched because they must attract the heads of ternary
constituents. Rule (37) is also required because the constituents are ternary. As argued above, with
basic binary constituents, far less machinery is required.
Notice that Chugach makes the distribution of relativized extrametricality more symmetric
in two respects. First, while relativized extrametricality in Cayuvava occurs at the right edge of a
foot built from right to left, in Chugach, relativized extrametricality occurs at the right edge of a
foot built from left to right. Foot -edge extrametricality does not correlate with the direction of
iteration. Moreover, while the feet in Cayuvava are syllabic trochees, the feet in Chugach are
iambs.
49
5.
Winnebago
Winnebago (Hale and White Eagle 1980) provides an example of foot -edge extrametricality
at the left edge of the foot. Main stress in Winnebago falls on the third syllable from the left.
Secondary stresses fall on alternating syllables to the right of the primary stress.
(39)
wajé
wijúk
hochichínik
hakirújikshàna
haakítujìk
haakítujìkshanà
'dress'
'cat'
'boy'
'he pulls it taut'
'I pull it tauf(plain)
I pull it taut'(decl.)
This system requires word -initial extrametricality and iterative iambic footing. A sample
derivation of the analysis so far is given below.
(40)
X X X
xx
xx
X X X X
xx
X X X X
hakirujikshana -> hakirujikshana ->
x
x
(x x) (x
XXXX
x
x) (x)
X
X
-> hakiru jiksha na
Winnebago is also subject to a rule of destressing which removes the final degenerate foot in a
stress clash.
Winnebago exhibits a process of epenthesis which has long posed a problem for metrical
theorists. In fact, under the derived ternarity analysis to be presented here, the facts of epenthesis
in Winnebago require nothing special. Rather, they confirm the analysis presented so far. The
epenthesis rule breaks up obstruant-sonorant clusters by inserting a copy of the immediately
following vowel. Halle and Vergnaud (1987; p.31) formalize the rule as in (41).
(41)
[-son] L+sónJ V
1
2
3
->
1323
In the following forms, epenthetic vowels are underlined.
(42)
hosháwazhá
harakíshurujìkshánà
maashárach
wakiripáras
hirakrohò
wakiripóropòro
'you are ill'
'you pull taut'
'you promise'
'flat bug'
'you dress, prepare'
'spherical bug'
What is crucial here is the effect epenthesis has on metrical structure. Sometimes the
epenthetic vowel behaves as if it were inserted before stress assignment, as in [hirakóroho], and
sometimes it behaves as if it were inserted after stress assignment, as in [hosháwazhá]. This
apparent paradox prompted Halle and Vergnaud to posit an otherwise unmotivated principle to
account for these facts: the Domino Condition. The Domino Condition is global in that it requires
the metrical structure of the entire string to be rebuilt under specified circumstances. In contrast, the
analysis below relies only on independently required machinery.
50
Let us assume simply that epenthesis precedes stress assignment and that epenthetic vowels
are marked as foot- extrametrical (i.e. subject to relativized peripherality).9 As such, they are
invisible to footing and must ultimately occur at the (left) periphery of feet. If, in the course of
footing, epenthetic vowels end up in the right place, nothing happens. If, on the other hand, an
epenthetic vowel should end up in an illegitimate position, then extrametricality marking on the
relevant syllable is not interpreted.lo
Here are some sample derivations showing how this works. In (43), epenthesis applies
inserting vowels marked as extrametrical. Next, initial extrametricality applies removing the first
syllable from the domain of the scansion. Iambs are then built.
(43)
X X X
xx
X
X X X
X
X X X X X X
X X
X X
harakishrujikshna -> harakishurujikshana ->
x
x
x
x
(x x)
XX
(x x)
(X)
X X X X X X
x x
X X X X X X
x x
-> harakishurujikshana -> harakishurujikshana
XX
Notice that unlike in Cayuvava, a relativized extrametrical syllable does not seem to be
transparent for the determination of clash. To get this result without stipulating that extrametricality
is only occasionally invisible, it is proposed that destressing is preceded by Left- Adjunction.
(44)
Left- Adjunction
(x) -> (x x)
xx
xx
Notice that (44) has no direct effect on the distribution of stresses. Additional support for (44)
comes from a process of Right- Adjunction to be motivated below.
The derivation in (43) is then completed as follows.
(45)
x
x
(x x)
(x x)
xxxxxx
x
(x)
xx
harakishurujikshana
x
(x x)
x
x
(x x) (x x)
xxx xxx x x
-> harakishurujikshana
The derivation in (46) shows that there must be an additional adjunction.
(46)
x
x x
x x
x x
X X X X X
x x x x x
hirakroho -> hirakoroho -> hirakoroho ->
xx xx
x
(x x)
x
(x x)
X X X X X
-> *hirakoroho
This is termed Right- Adjunction and precedes Left- Adjunction (44).
9This idea can be implemented by assuming that the rule responsible for epenthetic vowels inserts vowels
specifically marked as extrametrical or that epenthesis is followed by a rule marking empty morae as foot -edge
extrametrical.
10Cf. the interpretation of extrametricality in Yawelmani (Archangeli, 1984- 1985).
51
(47)
Right- Adjunction
(x)
xx
-> (x x)
xx
Since feet are right- headed, (47) does shift stress to the right.
Derivation (48) shows a complex case involving multiple instances of Right- Adjunction
and destressing in clash.
(48)
X
X
X
X
X
X
X
X X X X X X X
X X X X X X X
X X X X
wakripropro -> wakiriporoporo -> wakiriporoporo ->
x
x
x
(x)
(x)
(X)
X
X
X
(x x) (x X) (X)
X X X X X X X
X X X X X X X
-> wakiriporoporo -> wakiripo ropo ro ->
x
x
(X x) (x x)x
X X X X X X X
-> wakiripo roporo
This analysis needs no new mechanisms. It requires only that extrametricality be marked
before footing. Left- Adjunction and Right -Adjunction are of a form that has been seen in the
analyses of Latin and Chugach already. In contrast to Halle and Vergnaud's analysis, no global
reapplication of footing must take place halfway through the derivation.
Halle and Vergnaud formalize the Domino Condition as follows.
(49)
Domino Condition
The introduction of an additional position inside a bounded constituent destroys that
constituent and all constituents to its right if the Constituent Construction rule
applied from left to right, and all constituents to its left if the Constituent
Construction rule applied from right to left. Constituent structure is reimposed on
the affected substring by a subsequent reapplication of the Constituent Construction
rule.
This produces derivations as follows. Derivation (50) shows what happens when epenthesis does
not intrude in a foot.
(50)
x
x
<x> x(x x)
<x> (x x)
x
x x
hoshwazha -> hoshwazha -> hoshawazha
Derivation (51) shows what happens when epenthesis does intrude in a foot.
(51)
x
x
<x>x(x x x)
<x> (x x)
x x x
wakripras -> wakripras -> wakiriparas ->
x
X
x
<x>x(x x)x
<x>x(x x) (x)
<x>x(x x x
-> wakiriparas -> wakiripa ras -> wakiriparas
52
There are two problems with this analysis. First, it is devastatingly global. Intrusion into a
foot entails the destruction of that foot and all feet to the right. In addition, feet must be
subsequently reassigned.l1
A second problem with the Domino Condition is that it is empirically inadequate. Miner
(1990) cites data showing that epenthesis applies in initial syllables as well. This epenthesis does
not interrupt a foot, but must apparently trigger the Domino Condition.
(52)
'you mash hard'
'Black Hawk'
'in formation'
'hollow'
shawazhókji
kerejúsep
páragúchge
xorojíke
A sample derivation showing how the Domino Condition fails is given in (53).
x
(53)
<x> (x
x)
shwazhokji
x
x)
x x (x
-> *shawazhokjí
Under the analysis presented here, these cases are straightforward. The epenthetic vowel is
inserted and marked as extrametrical. Word -edge extrametricality applies vacuously (since the
leftmost syllable is already extrametrical), feet are built, and adjunction is inapplicable.
(54)
X
X
X
XX
XXX
X
X
(x
XX
XX
X) (X)
XX
shwazhokji -> shawazhokji -> shawazhokji
Thus the analysis in terms of derived ternarity is preferred to the analysis incorporating the
Domino Condition.
6.
Estonian
Finally, consider stress in Estonian. Estonian further instantiates the system of derived
ternarity developed here. The analysis to be presented here basically recasts the analysis of Prince
(1980) into the terms of this framework. Main stress falls on the first syllable of the word.
Secondary stresses fall on every second or third syllable thereafter. Prince (1980; p. 518) cites the
following data.
i1Rice (1990b) proposes an analysis in which the Domino Condition is argued to follow from other principles. This
analysis too is subject to the same objection of globality, however.
53
(55)
a.
'glove' part. sg.
'piece' part. sg.
'blinding'
'cunning' part. sg.
'worse' part. sg.
'ladder' all. sg.
kínnast
pálatt
pímestav
kávalátt
páhemáit
rételíle
p í mestáva l e -pí mestavá l e
'blinding' ill. sg.
pímestávasse- pimestavásse
'blinding' ill. sg.
hílisémattéle- hílisemáttele
'later' all. pl.
b.
áa:s:tátt
'year' part. sg.
káu:kéle- káu:kele
'far away'
júl:késse
'bold' ill. sg.
j61:kétest- j6l:ketést 'track' el. pl.
tóö:s:tústesse-tóö:s:tustésse
'industry' ill. pl.
téo:t:táttuttéltt
'supporter' abl. pl.
The stress pattern of Estonian is intimately tied up with quantity. There are three degrees of
quantity in Estonian: Q1, Q2, and Q3. Prince schematizes them as follows.
(56)
Ql
CV
Q2
CVV
CVVC
CVC
Q3
CVV:
CVV:C:
CVC:
If word -final consonants are extrametrical, the distribution of stress with respect to quantity
can be characterized as follows. Column (57a) shows the distribution of quantity when stresses fall
three syllables apart. ('X' stands for a syllable of any weight.) Notice how while the first two
syllables can be either Q1 or Q2, the third syllable must be Ql. Column (57b) shows what happens
when the two stresses are two syllables apart. Here there are basically two cases. The first four
possibilities are a mirror of the first four possibilities of (57a) for the first two syllables. The last
two cases involve Q3 on the first syllable and Q 1 or Q2 or the second. Column (57c) only gives
one possibility with Q3.
54
(57)
b.
a.
X
X X
Q1 Q1 Q1
Q2 Q1 Q1
Q1 Q2 Q1
Q2 Q2 Q1
X
X
X
XX....
X
X
X
X
Q1
Q2
Q1
Q2
Q3
Q3
c.
XX
XX
Q3 X
X
X X ....
Q1
Q1
Q2
Q2
Q1
Q2
X
X
X
X
X
X
....
_
This analysis is conceptually close to Prince's, but is cast in terms of the theory developed
here. Word -final consonants are extrametrical. Left- headed binary feet are built from left to right.
That stress recurs on every second or third syllable is captured by allowing any foot to be either a
syllabic or moraic trochee. A light syllable is optionally licensed as extrametrical at the right edge of
the foot. Q3 results in a nonfinal degenerate foot. Some sample derivations are given below.
(58)
a .
Xx
x
(X
X
kinnast -> kínnast
Q2 Q2
x x
x
x
(x) (x)
X
X
X X
aastatt -> aastatt -> áa:s:tátt
b.
c.
X
X
x
XXx
XXx
Q1 Q1 Q1
d.
(X X)
xX x
pimestav -> pímestav
xx
Xx
x
(x) (x x)
X X X
X X X
Q2 Q1 Q1
e.
x)
Q2 Q2
X
kaukele -> kaukele -> káu:kéle
XXx
xXXX x
XX
X
X
(X X) (X X)
XX X X x
Q1 Q2 Ql Q2 Q1 pimestavasse -> pímestávasse
f.
xx xx x
xx xxx
X
(x x)
xx
X
(x
x)
x x x
Q1 Q2 Q1 Q2 Q1 pimestavasse -> pimestavasse
There is an additional glitch involving heavy syllables after Q3 medially. In contrast to the
normal case, a heavy Q2 syllable can occur after a Q3 syllable if it is medial.
55
(59)
'industry' ill. pl.
tóö:s:tustésse
áu:sattéle
vét:mettéka
tóö:k:kailéki
'honest' all. pl.
'key' com. pl.
'hardworking' all. pl.
This can be accounted for if it is assumed that Q3 can optionally be assigned in a medial binary
foot.
(60)
x -> Q3 /
x)(
Estonian provides an argument for the system proposed here because the restrictions on the
relativized extrametrical syllable are different from the restrictions on the weak position in the
trochaic foot (57). If Estonian were analyzed with a dactylic or amphibrachic foot, this fact would
be much more difficult to account for. If the foot imposed no requirements on the weight of the
first two syllables, then it should impose no requirements on the weight of the third syllable. This
is not true as is apparent from (57a).
7.
Summary
To summarize, analyses of four systems exhibiting superficial ternary iteration have been
presented: Cayuvava, Chugach, Winnebago, and Estonian. All four of these were shown to be
more perspicuously analyzed in terms of foot -edge extrametricality.
A theory of extrametricality emerges which includes the following criterial properties. First,
extrametricality can occur at the edge of a word -medial foot. Second, extrametrical elements are
invisible to rhythm and destressing rules. Third, extrametricality may be lost to augment feet.
8.
Appendix: Weak Local Parsing
Hayes (in progress) develops a system that is similar in some senses to the system
proposed here. He suggests that languages can stipulate whether iteratively constructed feet are
adjacent or are separated by a greater distance. These options are Local Parsing and Weak Local
Parsing. For a language like Cayuvava, syllabic trochees would be constructed weak -locally.
(61)
x
x
x
X X X X X X X X
(x) x (x x) x (x x) x
ikitaparerepeha -> *íkitáparerépeha
Notice that this allows a straightforward treatment of the destressing problem in Cayuvava.
While it is difficult to consider in detail work that it is still in progress, it is perhaps
appropriate to indicate how the approach presented here differs from the Weak Local Parsing idea.
First, relativized extrametricality, as argued by Rubach and Booij (1990), is independently
required to treat syllabification. Weak Local Parsing is not independently required.
Second, relativized extrametricality can appear at either end of a foot. For example, in
Cayuvava and Winnebago, the extrametrical syllable appears closer to the origin of iteration, while
in Chugach and Estonian, the extrametrical syllable appears on the far side of the foot from the
origin of iteration. The latter is not a possibility for Weak Local Parsing.
Third, Weak Local Parsing is a property of the scansion, not a property of syllables. Hence
the analysis of Winnebago presented here would not be possible because relativized
extrametricality is marked on syllables before the scansion.
Thus, it appears that relativized extrametricality is distinct from Weak Local Parsing and
superior to it in explanatory and descriptive terms.
56
References
Archangeli, Diana (1984- 1985). Extrametricality in Yawelmani. The Linguistic Review 4. 101120.
Hale, K. and J. White Eagle (1980). A preliminary account of Winnebago accent. International
Journal of American Linguistics 46. 117 -132.
Halle, Morris (1989). The exhaustivity condition, idiosyncratic constituent boundaries and other
issues in the theory of stress. Manuscript, MIT.
Halle, Morris (1990). Respecting metrical structure. Natural Language and Linguistic Theory 8.
149 -176.
Halle, Morris and Jean-Roger Vergnaud (1987). An Essay on Stress. Cambridge: MTT Press.
Hammond, Michael (1984 /1988). Constraining Metrical Theory: A Modular Theory of Rhythm
and Destressing. 1984 UCLA doctoral dissertation, revised version distributed by IULC.
1988 published by New York: Garland.
Hammond, Michael (1987). Accent, constituency, and lollipops. CIS 23/2. 149 -166.
Hammond, Michael (1990). Metrical Theory and Learnability. Manuscript, U. of Ariz.
Hayes, Bruce (1981). A Metrical Theory of Stress Rules. 1980 MIT doctoral dissertation, revised
version distributed by IULC. Later published by New York: Garland.
Hayes, Bruce (1982). Extrametricality and English stress. LI 13.227 -276.
Hayes, Bruce (1984). The phonology of rhythm in English. LI 15.33 -74.
Hayes, Bruce (1987). A revised parametric metrical theory. NELS 17.274 -289.
Hayes, Bruce (1989). Compensatory lengthening in moraic phonology. LI 20.253 -306.
Hayes, Bruce (in progress). Principles of Metrical Stress Theory. Manuscript, UCLA.
Inkelas, Sharon (1989). Prosodic Constituency in the Lexicon. PhD dissertation, Stanford.
Kager, René W.J. ( 1989). A Metrical Theory of Stress and Destressing in English and Dutch. PhD
dissertation, Utrecht.
Key, Harold (1961). Phonotactics of Cayuvava. International Journal of American Linguistics 27.
143 -150.
Key, Harold (1967). Morphology of Cayuvava. The Hague: Mouton.
Krauss, Michael (1985). Yupik Eskimo Prosodic Systems: Descriptive and Comparative Studies,
Alaska Native Language Center Research Papers 7. University of Alaska.
Leer, Jeff (1985). Prosody in Alutiiq. In Michael Krauss (ed.). 77 -134.
Levin, J. (1988). Generating ternary feet. Texas Linguistic Forum 29.97 -114.
Liberman, Mark and Alan Prince (1977). On stress and linguistic rhythm. LI 8.249 -336.
Miner, Kenneth L. (1990). Winnebago accent: the rest of the data. Manuscript, University of
Kansas.
Myers, Scott (1987). Vowel shortening in English. Natural Language and Linguistic Theory 5.
485 -518.
Prince, Alan (1980). A metrical theory for Estonian quantity. LI 11.511 -562.
Prince, Alan (1985). Improving tree theory. BIS 11.471 -490.
Prince, Alan (1991). Quantitative consequences of rhythmic organization. Manuscript, Brandeis
University.
Rice, Curtis (1988). Stress assignment in the Chugach dialect of Alutiiq. Paper presented at CLS.
Rice, Curtis (1990a). Pacific Yupik: implications for metrical theory. Presented at the Third
Arizona Phonology Conference. To appear in the proceedings.
Rice, Curtis (1990b). Deriving the domino effect in Winnebago. Manuscript, University of
Arizona.
Rubach, Jerzy and Geert Booij (1990). Edge of constituent effects in Polish. Natural Language
and Linguistic Theory 8.427 -463.
Spring, C. (1989). Dactyls in Cayuvava. Manuscript, University of Arizona.
Yip, Moira (1987). English vowel epenthesis. Natural Language and Linguistic Theory 5.463484.
57
Michael Hammond
Department of Linguistics
University of Arizona
Tucson, AZ 85721
[email protected]
58
"PRO Analysis" for Subject- Oriented Secondary Predicates*
Hisako Ikawa
The University of Arizona
Department of Linguistics
O. Introduction
The central aim of this paper is to argue that subject -oriented (depictive) secondary
predicates' such as those in the sentences in (1) have PRO as their subjects, contrary to the
prevailing "non -PRO analyses" in recent literature (Williams 1980, Rothstein 1983, Chomsky
1986, McNulty 1988, Roberts 1988, Nakajima 1991, Koizumi 1992, and others), which do not
consider the predicate phrase to be constituted as a Small Clausen.
(1) a. John married young.
b. John left the hospital healthy.
c. They left the medical school doctors.
In this paper, I will endeavor to show both empirical and theoretical reasons that we
should adopt a "PRO analysis." The governing idea is that the relationship between the matrix
subject NPs and their predicates is not theta -marking, but control.
Regarding the structure of the predicate phrase (Small Clause), I will present the
possibility that the predicate phrase is AgrP, based on the phenomenon of agreement in English
and Spanish. I will also deal with the inevitable question of how PRO is licensed.
*I am very grateful to Andrew Barss, Dick Demers, Eloise Jelinek, and Simin Karimi for their invaluable
suggestions and warm encouragement. I would also like to thank Richard Bernat, Tom Craig, Willem De Reuse,
Amy Fountain, Chip Gerfen, Terence Langendoen, Shaun O'Connor, and Pilar Pillar, who kindly and patiently
acted as consultants, and who provided me with many inspiring suggestions. I greatly appreciate the helpful
comments and probing questions of Masaaki Fuji and Naoki Nakajima. Thanks are also due to my peer reviewers,
especially Jen -i Jelina Li and Nayla Yateem.
t In this paper, we will deal only with the subject- oriented (depictive) secondary predicates. The term
"secondary predicate" stems from its adjuncthood. The italicized predicates below are adjuncts that are not selected
by any head
i) subject -oriented depictive
a) John left angry.
b) They left the medical school doctors.
ii) object -oriented depictive
a) John ate the meat raw.
2To the best of my knowledge, there is little literature on a PRO- analysis for the Small Clause (Chomsky
1981, Stowell 1981, and Hornstein and Lightfoot 1987, etc.), and only a small part of that literature focuses on
secondary predicates. Hoshi (1992) is one of the exceptions.
59
The paper is organized as follows. Section 1 provides a brief review of previous studies
based on theta -role assignment, with special emphasis on McNulty (1988). In section 2, I will
first present the data and will then show that these data cannot be accounted for by a theta marking analysis. Also I will present major reasons to choose a PRO analysis over theta- marking
analysis in the light of the VP- Internal Hypothesis. In section 3, I will consider what structure
should be given to the predicate phrase (Small Clause) under my analysis. I will also try to
delineate a mechanism of licensing PRO within Chomsky's (1992) framework. I use negation data
in both English and Spanish to support and develop my hypothesis. In the final section, the
remaining problems with the predicate NP will be discussed.
1. Theta- marking Analysis - McNulty 1988
According to McNulty (1988), "predicate of" is understood to mean "theta -role
assignment by an XP ". That is, "B is predicated of A" is understood as B assigns a theta role to
A, where B =XP. Consider the sentence in (2).
(2)
John left angry.
Since the predicate left assigns a theta -role to John, left is a predicate ofJohn. This
relationship is called primary predication and left is the primary predicate. By contrast, angry in
(2) is termed the secondary predicate because (2) would be a grammatical string even if the AP
(angry) were eliminated. This is not the case with, for instance, VP, a primary predicate.
McNulty applies the above type of theta -role assignment requirement to secondary
predication as well. In her theory, the AP angry is required to assign theta -role to John to
establish a predication relationship.
McNulty elaborates her theta -marking analysis and proposes the following condition
governing the distribution of secondary predicates.
(3) Locality Condition on XP Theta Role Assignment (LCXP)
A assigns a theta -role to B iff A mutually m- commands B and there is no Z such
that Z mutually m-commands A, where A, Z = theta -assigning XP
(McNulty
1988)
Notice that in McNulty's framework the maximal projection XP assigns a theta -role to its
subject. The AP angry can assign the theta -role to John with no violation of her LCXP3. (see
3As we will see in 2.2.1., McNulty (1988) assumes that the subject -secondary predicate is adjoined to the VP
node. Consider the configuration she gives in McNulty (1988). Since she accepts Theta -Criterion (Chomsky
1986a) (and does not adopt the VP Internal Hypothesis), the AP angry meets LCXP. Also notice that she has to
take the definition of "m- command" in (ii) below to meet her LCXP, following the basic idea of May (1985) and
Chomsky (1986b).
60
McNulty 1988 for more details.) If we were to consider the theta- assigner to be X° (or X'), the
mutual m- command relations between the predicate and the subject would break down under the
configuration McNulty assumes (See note 3), and LCXP might become a meaningless condition.
We will also refer to this problem in 2.2.1.
2. Arguments for PRO analysis
2.1. Data
In this section we will look at sentences that cannot be explained by a theta- marking
analysis.
2.1.1. Raising Adjectives
It is striking that raising adjectives such as likely or certain (sure) can occur as secondary
predicates. According to my consultants, the following sentences are perfectly normal as regular
predicates, with no commas before the predicates. Consider the following sentences;
(4) John left the room likely to find his mother.
(5) John left the room certain/sure to win a race.
As we have seen in section 1, theta -marking analyses assume that the whole sentences in
(4) and (5) constitute one unit. Thus, likely is required to assign a theta -role to John in (4) and
certain/sure is required to assign a theta -role to John in (5). However, these raising adjectives are
one -place predicates which take a clausal complement. The subject position is not assigned a theta
role. John has no thematic relation with the adjective likely in (4) and with certain (sure) in (5).
This means that such raising adjectives cannot assign any external theta- roles; John cannot be
theta- marked by likely in (4) or by certain (sure) in (5). Theta -marking analyses cannot provide
any explanation for the above data.
The issue we have to consider is whether adjuncts have a PRO subject or not. The data I
raise here is one of the strongest pieces of evidence that this type of secondary predicate actually
has a subject position, as is discussed below. We have to assume that the above sentences have
the structures shown in (6) and (7), that is, the PRO subject is generated as an argument of find in
(i)
NP
3oiin
\I'
I NP
VP
AP
P
t
left
angry
(ii) M- command
A m- commands B iff no segment of A dominates B and no segment of B dominates A and every G
(= maximal projection) that dominates A dominates B.
61
(6), and that it is generated as an argument of win in (7). In (6) and (7), PRO has no thematic
relation with the adjectives likely and certain (sure), respectively, but it certainly occupies the
subject position of the infinitival complement.
(6) John left the room [ likely PRO to find his mother].
(7) John left the room [ certain/sure PRO to win a race].
I claim that PRO in the above sentences is controlled by the matrix subject John.
2.1.2. Implicit Argument
Roeper (1987) raises the examples below as grammatical sentences;4.
(8) The game was played dnmk/nude /sober /angry.
This kind of sentence is problematic for a theta -marking analysis, because the predicates
(drunk, nude, sober, angry) do not assign a theta -role to the non- agentive subject (the game).
Therefore, an explanation based on a theta -marking analysis does not work here. Notice that the
following sentence is impossible.
(9) *The game was drunk/nude/sober /angry.
Roeper uses the example in (8) to show that predicate adjectives allow control by "implicit
agents." This is similar to the arguments that Manzini (1983) referred to as "phonologically null
agents." Roeper's assumption is that predicate adjectives, adverbial participial clauses, and
rationale clauses all share a common property: a PRO subject. His examples are as follows;
(10) a. <the predicate adjective>
The game was played [PRO drunk].
b. <the adverbial participial clause>
The game was played [PRO wearing no shoes].
c. <the rationale clause>
The boat was sunk [PRO to collect the insurance].
Roeper (1987) argues that PRO (including "PRO- arbs ") can be controlled by implicit
arguments'. However, he does not delve into clarifying the structure of secondary predicate
constructions.
4Some of my consultants state that the present tense is more comfortable, as in the following examples.
(i) The game is played drunk/nude/sober.
cf. (ii) The game is played wearing no shoes.
5PRO that has an arbitrary reference is often called "PRO -arb" in the literature.
'Implicit arguments have been much discussed in recent literature (e.g., Jaeggli 1986, Safir 1987, Williams
1987, Baker, Johnson and Roberts 1989).
62
What I would like to point out in this section is that in order to explain the sentences in (8)
we have to posit that secondary predicates have PRO as their subject.
2.1.3. Parallelism with Adverbial Participial Construction
Unlike 2.1.1. and 2.1.2, this subsection will not provide new, direct evidence for a PRO
analysis. Rather, it serves to support the discussions in 2.1.1. and 2.1.2.
Let us consider what categories are allowed as secondary predicates. We will deal with
this question in section 4. However, I will point out here that verbs cannot occur in bare
infinitival form as secondary predicates. Consider the following sentences.
(11) *The boy came run out of the house.
(12) *Father sat read the newspaper.
(13) *They stood look at the exciting game.
If the italicized verbs are replaced with their participial (V -ing) forms, the sentences
become grammatical as in (14) -(16).
(14) The boy came running out of the house.
(15) Father sat reading the newspaper.
(16) They stood looking at the exciting game.
This construction has not been discussed in detail in generative grammar to date'. At
present we must say that it is not clear what explanation should be given to the construction.
However, it would be possible that this construction is also regarded as a VP secondary predicate,
because X' -theory predicts that all categories can occur as secondary predicates, as long as an
independent principle does not block it. The V -ing form in (14 )-(16) is not NP but VP, as is clear
from the following data.
(17)
*The boy came the running out of the house.
(18) a. *Father sat the reading the newspaper.
b. *Father sat the reading of the newspaper.
(19)
*They stood the looking at the exciting game.
I propose that the adverbial participial clause is a Small Clause having the structure,
roughly, [PRO [VP]]. If this is the case, my secondary- predicate analysis can explain Roeper's
example (l Ob). If one claims that the participle in question is a kind of secondary predicate, we
can view the examples below as raising/unaccusative -verb secondary predicates. As is well
known, raising/unaccusative -verbs have only one theta -role to assign, namely, an internal -theta
role. The raising predicate seeming cannot assign an 'external' theta -role to John in (20) and
neither can the unaccusative arriving in (21). Theta -marking analyses break down when it comes
to this type of sentences, as they do with the examples given in 2.1.1.
7Williams (1975) is one of the exceptions. He refers to the participial clause as Small Clause, which is
smaller than the "that" clause or the infinitival clause in that it does not include a S' node.
63
(20) John left the room seeming to be sick.
(21) John felt embarrassed arriving late.
Of course, we have to consider why this VP must be a V -ing form, and why it does not
work as a bear V- infinitival form or other inflected form (e.g., PAST -form). And we have to
delineate the exact structure of the Small Clause.
2.2. VP-Internal Hypothesis and Theta -Role Assignment Capability
In this section, I will discuss the VP- Internal Hypothesis and theta -role assignment. The
VP- Internal Hypothesis (supported by Kitagawa (1986), Koopman and Sportiche (1988), Kuroda
(1988), and others) proposes that subjects are base -generated VP- internally - namely, within the
VP- nodes. I will show that a PRO analysis for secondary predicates is compatible with the VPInternal Hypothesis, while a theta -marking analysis is not. If one theory is compatible with a VPInternal Hypothesis and another theory is not, then the compatible theory would definitely be
preferred theory.
Taking this step further, if a PRO analysis is proved to be valid for secondary predicates,
then this will support the VP- Internal Hypothesis. (The empirical data discussed above make a
strong case for the advantages of a PRO analysis.) Once the VP- Internal Hypothesis is adopted,
it seems clear that a PRO analysis must also be adopted. Many of the theta -marking analyses have
incorporated the VP- Internal Hypothesis. However, these attempts are vulnerable in terms of
category neutrality and theta -role assignment capability.
Let us begin by discussing whether theta -assignment can be done properly in a theta marking analysis. A theta -marking analysis ought to show that the Theta- Criterion is met
between the predicates and their subject (the matrix subject) because, under this view, the whole
sentence is considered to be one unit. It proves useful to examine theta -assignment under both a
theta -marking analysis and a PRO analysis.
2.2.1. Theta- Criterion
As is well known, Chomsky's Theta- Criterion actually has two versions. Let us discuss the
earlier version first.
(22) Theta- Criterion
Each argument bears one and only one theta -role, and each theta -role is assigned
to one and only one argument.
(Chomsky 1981, 36)
The Theta- Criterion of this version rules out the sentence in (2), as long as secondary
predication is considered to be licensed by theta- marking. I will repeat (2) here as (23).
The VP- Internal Hypothesis successfully explains various properties of languages, such as the behavior of
floating quantifiers and word -order parameters. (See Koopman and Sportiche (1988).)
64
(23) John left angry.
By definition, angry assigns the theta -role to John, and left also assigns the agent theta -role to
John. As a result, John receives two theta -roles in this sentence. This is clearly a Theta- Criterion
violation, and there is no way to avoid it, unless we either consider a very different mechanism
from theta - marking to establish the predication relationship (as in Williams 1980), or posit that the
secondary predicate angry assigns the theta -role to PRO, which the matrix subject John controls.
Now we will consider the later version of Theta- Criterion, a weaker version than (22).
(24) Theta- Criterion
Each argument a appears in a chain containing a unique visible theta -position P,
and each theta- position P is visible in a chain containing a unique argument a
(Chomsky 1986a, 97)
Notice that (24) allows an argument to receive more than one theta -role. This formulation
requires that every chain contain one and only one argument and one and only one theta- marked
position.
If we adopt the (1986a) version of the Theta -Criterion and do not take the VP- Internal
Hypothesis, then the following sentence can be ruled grammatical.
(25) John left the room angry.
John is a single member chain. John can be assigned two theta -roles - one by left and
one by angry - with no violation at all, because the chain contains only one argument and only
one theta -marked position. The sentence in (25) clearly satisfies the Theta- Criterion (1986a).
Whether or not this is a clever solution will not be discussed here. In either case, the problem is
resolved.
If we take the VP- Internal Hypothesis (and consider it to be category- specific only for the
sake of this discussion), complicated problems may arise.
(26) [John; [vp t; [vp [v left the room]][,p angry]]]]
In (26), left theta -marks t with no problems, and angry must theta -mark t to establish
secondary predication. The chain (John, t) contains one theta -marked position: the position of t.
If this theta -assignment were done properly, there would be no problems. If we take
McNulty's LCXP, and assume that AP is the theta -assigner, we see that AP angry can assign the
theta -role to t in her framework, because the AP angry mutually m- commands t. McNulty
assumes that angry is adjoined to the VP. The structure in question is as follows:
65
(27) IP
NP
John
\
r
I
VP
t
-V'
AP
A'
A
NP
left the room angry
V
However, under the VP- Internal Hypothesis, if we consider A° (or A') (angry) to be a
theta -assigner to t, then angry cannot theta -mark t because the A° angry does not mutually mcommand t.
A PRO analysis does not face these problems under either version of the Theta- Criterion.
This is another strong argument for a PRO analysis. And most importantly, if we adopt the VPInternal Hypothesis, it becomes essential that we also adopt a PRO analysis, as I will explain in
the next subsection. The VP- Internal Hypothesis and a theta -marking analysis have been used
together in the literature (e.g., Hasegawa 1991), but this combination seems problematic. In the
next subsection, I will explain in detail why the VP- Internal Hypothesis requires a PRO analysis,
and how a PRO analysis supports the VP- Internal Hypothesis.
2.2.2. Category Neutrality
Stowell (1981, 1983) argues for structural uniformity across categories in terms of the
"X'- theory," which was originally proposed in Chomsky (1970) and later elaborated upon in
Chomsky (1981). Stowell proposes that all major syntactic categories contain a structural subject
position, conforming to a general pattern determined by principles of X'- theory.
In the VP- Internal Hypothesis theta -role assignment of both the subject and the object is
done uniformly within the V- projection. Committing the specifier of the VP guarantees the null
hypothesis that all categories have their own specifier positions. It follows that the verb assigns
the 'external' theta -role to the subject in [Spec VP], and that the adjective assigns the 'external'
theta -role to the subject in [Spec AP].
That is, an adjective assigns the theta -role AP- internally within this hypothesis. More
generally, all theta -marking by X is done within the projection of X. The VP- Internal Hypothesis
is not category- specific but category- neutral.
Consider the following sentence. I will repeat (26) here as (28).
(28) [John; [vr t; [vp [v, left the room]][A,angry]]]]
In (28) angry cannot assign the theta -role to John (or its trace), which is outside the AP.
Instead angry must assign the theta -role to the Spec of AP. This is only compatible with a PRO
analysis. This idea is fully compatible with Chomsky's recent framework9 (e.g., Chomsky 1992),
9Chomsky and Lasnik (1991) adopt the VP- Internal Hypothesis. According to their assumption, in a sentence
such as "John met Bill. ", John is considered to be generated in [SPEC, VP], and raises to [SPEC, IP] to receive
66
because, informally speaking, the subjects are too far away from A (A- projection) (A °, A', AP) to
be theta -marked, as long as we subscribe to the notion of highly - differentiated functional
categories, such as AgrsP , TP, and AgroP. The category neutrality of the VP- Internal Hypothesis
is a strong argument for PRO analysis. And PRO analysis for secondary predicates can be a
contribution to establish VP- Internal hypothesis:
3. The Predicate Phrase as AgrP
We now turn to two related questions. First, what is the exact internal structure of these
predicates? And what licenses PRO?
3.1. Why AgrP?
In this section, I will argue that the predicate phrase (Small Clause) is AgrP.
The empirical evidence for this proposal is that the predicate must agree with the PRO,
which is controlled by the matrix subject. Consider the following sentences. (29) is from Spanish,
and (30) is from English.
(29) Thomas y Pedro se fueron [PRO enfadados]
Tom and Peter left
angry
(+plural, +masculine)
(30) They left the medical school [PRO doctors/ *a doctor].
In (29) the matrix subject, Thomas y Pedro, is [ +plural, +masculine] and the secondary
predicate adjective, enfadados, is also [+plural, +masculine]. The agreement is fully realized
morphologically. English does not have such a rich inflectional system. However, (30) shows
that even in English the secondary predicate noun agrees with the matrix subject in number.
These facts tell us that the predicate has to get agreement within its clause. This means we
have to assume that the predicate clause includes an agreement mechanism within itself Although
Small Clauses have been broadly analyzed as a [- Tense], the possibility that they are [ +Agr] has
been explored in the literature10.
For the present purpose, I will tentatively propose the following internal structure for the
secondary predicate phrase. The tree in (31) is for (29).
Case and produce a visible chain.
10Belletti (1990) analyzes Small Clauses as AgrP based on Italian data.
67
(31) [PRO enfadados]
AgrP
PRO
Agr'
NA
t
A'
A
enfadados
'and
[+pL, +masculine]
PRO is base -generated in Spec AP in (31). The Spec AP is the subject position of the
Small Clause. This is derived from the VP- Internal hypothesis. The next step is that PRO moves
to Spec AgrP. The question that arises is why PRO can't stay in its base -generated position. If
PRO moves to Spec AgrP, is PRO properly licensed in this configuration? This question will be
dealt with in detail in the next subsection.
Notice that the movement of PRO also triggers subject -predicate agreement. The
predicates (enfadados in (31)) must get agreement in the given structure. We have to consider the
mechanism of this agreement. I propose that, (presumably) at LF, the predicates (e.g., the A in
(31)) raise to Agr solely for agreement, not for Case or other requirements. (However, I will not
go into the further discussion at present.)
3.2. How is PRO Licensed?
3.2.1. The Ungovernment Requirement
In Chomsky's (1981) theory, the PRO theorem requires that PRO be ungoverned. In Dstructure, enfadados governs PRO in (31). So we are forced to conclude that PRO moves to an
ungoverned higher position, namely, the Spec position of the (closest) XP (maximal projection
phrase) distinct from its own projection. In the case of secondary predicates, we must consider
the XP to be the AgrP, as was proposed in section 3.1. Thus, PRO is motivated to move to Spec
AgrP in order to be ungoverned. However, a problem arises. The head Agr is generally taken to
be a governor for PRO; PRO is governed by Agr in this movement. In Chomsky's (1981) theory,
there must be a non -governing head. This requirement is hard to meet in the configurations I
proposed.
Also, in the Pre - Minimalist theory, we have to consider the possibility that PRO is
governed from outside of the Small Clause (AgrP). Since the Small Clause in question is an
adjunct, it may be an island. In fact, by the definition of Cinque (1990, 42)", AgrP is a barrier
that prevents PRO from being governed from outside.
I 'Cinque defines a barrier for government as follows:
Every maximal projection that fails to be directly selected by a category nondistinct from [ +Vl is a barrier
for government.
68
In the Pre - Minimalist theory, the XP does not have to be AgrP for PRO to evade
government. There is no restriction that states the XP should be a functional category, because
the only condition for XP is that it be a barrier for the predicate (enfadados in (29) and doctors in
(30)). However, in Chomsky's (1992) framework, PRO must be Case - checked via a head -Spec
relation, and only a functional category can be a Case Checker. As a consequence, XP must be a
functional category. In this respect, the argument in Minimalist theory is more constrained and
restrictive, and, as a result, stronger than that in Pre- Minimalist theory.
For this reason, we do not further explore the "pre- Minimalist PRO licensing mechanism."
3.2.2. The Case -Checking Requirement
In this section, we consider how PRO is licensed - that is, how PRO is Case - checked following Chomsk's recent works.
Chomsky and Lasnik (1991) claim that PRO has Case like other arguments, observing that
PRO is forced to move from a non -Case position to a position where its Case can be checked.
However, they assert, the Case that PRO bears is different from the familiar ones like nominative
Case, accusative Case, etc. They regard PRO as a "minimal" NP argument, lacking independent
phonetic, referential or other properties. Their proposal is that PRO has a minimal Case called
"null Case," and is checked via a head -Spec relation. In their framework, nominative Case is
checked by INFL, the head of IP, where I involves the features of tense and agreement, while null
Case is checked by the minimal "INFL," where I lacks tense and agreement features. They
assume that null Case is checked by the infinitival element (with null agreement) and the head ING
of gerundive nominals, since PRO typically appears in such constructions as:
(32) a. PRO to VP (to be sick)
b. PRO ING VP (being sick)
However, even if this is the case, there is no independent reason to assume that only to
and ING can be null Case Checkers. In the next subsection, I will explore the possibility that
another functional head can Case -Check PRO under the Spec -head configuration.
3.2.3. Proposal - Agr as a Null Case Checker
As we have discussed in 3.1, PRO is generated in Spec AP in (31). This position is the
subject position. Then PRO moves to Spec AgrP in order to be Case -Checked. For PRO to be
licensed, this movement is obligatory.
Now we can say that the null Case Checker is the head Agr. Notice that Agr is the only
possible Case -Checker within the predicate phrase. According to Chomsky (1992), only the
functional category can check Case. As shown in (31), Agr meets the Case -Checking condition
from the viewpoint of syntactic configuration. It would be desirable to analyze the predicate
phrase as AgrP, since this would provide a way for PRO to do its Case -Checking without
stipulation.
3.2.4. Is a TP Node needed?
As is well known, the Small Clause has no overt Tense manifestation. Thus, we have two
logical possibilities available regarding the Tense Phrase: 1) the predicate phrase (AgrP) contains
69
a Tense node, and 2) it does not12.
First, we will explore the possibility that the predicate phrase contains a Tense node.
According to Chomsky (1992) (and Lasnik (1993)), nominative Case is licensed via the head Spec relation with the functional head AGRs -Tense complex created by the raising of Tense to
AGRs. Nominative Case is checked by the N- feature of T [ +Tense]. In a similar way, accusative
Case is licensed by the AGRO -V complex created by the raising of the Verb to AGRO. Chomsky
assumes the following basic structure:
(33)
CP
C'
SPEC
AGR"
C
\
SPEC AGRs'
AGRs
T
TP
G
SPEC
ÁG°'
AGRO VP
The assumption is that the Case properties depend on characteristics of T and the V head
of VP.
Following Chomsky's view, it is possible to assume that null Case can be checked by the
N- feature of T [- Tense]. Thus, we might have the following structure for a predicate phrase
having the [- Tense] feature. Consider, for example, the sentence John left angry.
(34)
AGR"
ÁGRs'
T AGILs
PRO
TÁ
SPEC
"
AGRo'
(NP, PP, VP)13
AGRO
A'
I
A
angry
I propose, for the time being, that the N- feature of T [- Tense] triggers null Case Checking.
12Belletti (1990) and Grimshaw (1991) take the latter position.
13The question of what categories are allowed as secondary predicates will be discussed in section 4.
70
This proposal is consistent with cases where PRO appears in other configurations. Consider the
typical occurrence of PRO: the subject of the infinitival clause and that of gerundive nominals.
In the next subsection, we will observe negation data from Spanish and English, and then
consider whether this data supports our hypothesis. The relevant questions here are 1) whether
the negation data proves the hypothesis that the predicate phrase is AgrP, and 2) whether the
negation data suggests the existence or absence of the Tense node in Small Clauses. The
significant issue here is whether the TP node is actually needed in the predicate phrase.
3.3. Negation
In this section, we will consider the interaction of negation and AgrP. Consider the
sentences below. (35) is from English, and (35') is from Spanish
(35) a. John left the room not angry.
b. ? John married not young.
c. They parted not good friends.
(35') a. * Juan sallo de la habitación no enfadado
John got out of the room not angry
b. * Juan se casó no joven.
John married not young
According to my consultants, the English sentences in (35) are grammatical, or at least
acceptable, while the Spanish sentences in (35') are not.
How can we explain this difference in grammaticality between Spanish and English? Why
can English secondary predicate phrases be negated, while those in Spanish cannot ?14
According to Laka (1990), Spanish NegP has the structure shown in (36), while English
has the structure shown in (37) (Pollock (1989)). (For the sake of convenience, the structures are
simplified.)
t4The examples in (35) and (35') seem to be sentence negation rather than constituent negation. It may be
more appropriate to use sentences of the following type to show that the examples truly are sentence negation.
i) John left the room not convinced that he'd done the right thing.
The following examples can be regarded as constituent negation. When regarded as constituent negation, the
sentences are more acceptable in Spanish, according to my consultants.
(ii) a.
John left the room not very angry.
John married not very young
(ii') a. 7? Juan salio de la habitación no muy enfadado.
b. ?? Juan se casó no muy joven.
b.
(iii) John married not particularly far into his thirties.
(iii')Juan se casó no muy entrado en los treinta.
71
(36) [Spanish]
(37) [English]
TP
grP
In Spanish, NegP occupies the highest position in the structure (Laka 1990). In English,
on the other hand, NegP is located between TP and AgrP. Notice that TP dominates AgrP. That
is, TP is higher than AgrP in Pollock's theory". I will adopt this analysis.
If we postulate that the AgrP does not dominate a Tense node, the following
straightforward explanation is possible: In Spanish, the predicate phrase (AgrP) cannot include
Neg beyond TP, but in English it can. This seems plausible.
If we assume that AgrP dominates a Tense node, following Chomsky's (1992) basic idea,
we must consider another configuration. In Chomsky's framework, the two functional roles of
AGR -= AgrsP and AgroP - are distinguished. In Chomsky (1992), NegP is located between
TP and AgroP in English in the following way:
(38) [English]
,s1
/P
The Spanish structure is arguably the following:
(39) [Spanish]
mu
grsP
TP
AgroP
In (38), NegP is evidently located under AgrsP, which accounts for the fact that secondary
predicates can be negated in English. The Spanish case is more complicated. In (39), NegP is
John got married not very entered in his thirties.
15Note, however, that Belletti (1990) argues that AgrP dominates TP.
72
still in the highest position. However, if AgrP can incorporate the higher NegP as a constituent
- just as we assumed in explaining the English data in the configuration (37) - Neg must occur
within Agrp. This prediction contradicts the fact that Spanish does not allow negation within the
predicate phrase. This shows that Chomsky's (1992) Case - checking mechanism using Tense
features (or the clause structure he assumes) does not work as it is in secondary predicate phrases.
As long as we use Tense features to check the null Case, we must have the structures shown in
(38) and (39), and as we have seen, these structures are problematic for Spanish.
To explain the Spanish data, I will not take the configuration in (39). Instead, I will
assume that TP dominates AgrP, namely, that TP is higher than AgrP. I will apply Pollock's
analysis to recent models of functional categories - highly differentiated Agr phrase
configurations. Thus, I will propose the structures shown below. Note that I am turning to the
position that the predicate phrase (AgrP) does not dominate a Tense node.
(40) [English]
P
(41) [Spanish]
AgroP
In (40), since NegP is lower than AgrsP, the predicate phrase can include Neg. This
explains the fact that, in English, the predicate phrase can be negated. In (41), TP intervenes
between NegP and AgrsP. For this reason, AgrsP cannot incorporate NegP beyond TP. This
explanation agrees perfectly with the Spanish data.
To conclude, the predicate phrase is AgrP, and it has no Tense node16. The assumption is
161f we assume that the secondary predicate phrase (AgrP) has no Tense node, there is a possibility that we
can explain the semantic constraint in subject -oriented secondary predicates.
The constraint is that secondary predicates must be stage -level predicates (cf. Carlson (1978) and Kratzer
(1989)). Consider the sentence below.
(i) John carne to the party blue -eyed.
The predicate blue -eyed is normally an individual -level predicate. However, in this construction, we have to
interpret the predicate as a state -level predicate. Therefore, the interpretation is that John came to the party
wearing blue contact lenses (in order to surprise the guests).
According to Diesing (1992), in the case of individual -level predicates, the subjects are base- generated in
[Spec, IP], and the subjects of stage -level predicates are base- generated in [Spec, VP]. If we interpret Diesings 1P
73
that TP dominates AgrP. This idea comes originally from Pollock (1989), and is developed for
Romance in Laka (1990).
4. Primary Predication and Secondary Predication
In the rest of the paper, I will deal with the difference between primary predication and
secondary predication. I will first address the question of what categories are allowed as
secondary predicates.
4.1. What categories are allowed as Secondary Predicates?
What categories are allowed as secondary predicates? McNulty(1988) observes that
English has AP, NP, and PP secondary predicates, as in the following sentences.
(42) John left the hospital healthy. [AP]
(43) John left the hospital a healthy man. [NP]
(44) John left the hospital in good health. [PP]
If we add a VP to the above categories, we can have a more complete paradigm. The X'schema predicts the occurrence of every category. Thus, if we are faced with the absence of a
category, it follows that we must filter it out by certain independent principle(s).
McNulty (1988) points out that NP secondary predicates do not exist in Spanish, but she
does not provide any explanation for it. Some languages seem to allow NP secondary predicates,
while others do not. The reason is worth exploring. Notice that the NP predicate is allowed in
primary predication universally, but not in secondary predication. In theory, we should expect a
NP to occur as a secondary predicate, as long as the NP is licensed. Are there any crucial
differences between primary predication and secondary predication? We will discuss this problem
in the next subsection.
4.2. Identifrcational be and Predicational be
Now let us consider the differences between primary predication and secondary
predication. Descriptively speaking, we can say that primary predication includes a copula be at
least in some languages like English, while secondary predication does not.
However, I propose that both primary and secondary predication include a copula be. I
suggest that be in primary predication is overt, while be in secondary predication is covert.
Observe the following examples:
(45) a. John left the hospital healthy. [AP]
b. John left the hospital a healthy man. [NP]
c. John left the hospital in good health. [PP]
to be TP and not AgrP, the explanation might be possible. The predicate phrase (AgrP) has no TP. Thus,
individual -level predicates (base- generated in [Spec TP]) cannot occur as secondary predicates, while stage -level
predicates (base-generated in [Spec, VP]) can occur within the AgrP. This possibility is left for further research.
74
d. John came running out of the house. [VP: V -ing]
(46) a. John is healthy.
b. John is a healthy man.
c. John is in good health.
d. John is running out of the house.
The secondary predicates in (45) are all forms that can appear in the post -be position in
primary predication, as in (46). I will further discuss the "covert be hypothesis" in section 4.3.
Given the hypothesis that both primary and secondary predication include a copula be, we
can compare the characteristics of be in these two types of predication.
In primary predication there are at least two kinds of be: the identificational be and the
predicational be" . Consider the examples below.
(47) [Identificational be]
a.
Mary is my mother.
a'. My mother is Mary.
b. The morning star is the evening star.
b'. The evening star is the morning star.
(48) [ Predicational be]
a.
Mary is a nurse
a' * A nurse is Mary.
b. The student was sick
The NPs in the identificational sentences must be [ +specific]. The truth value in these
sentences is not changed if the two NPs are exchanged as in (47 a -a') and (47 b -b'). By contrast,
the predicational sentences do not have such properties, as is clear from (48).
I claim that be in secondary predication should be the predicational be, not the
identificational be, while be in primary predication can be either type of be. This is the most
significant difference between primary predication and secondary predication. Consider the
following sentences.
(49) [Secondary predication -- Identificational be]
a. * Mary left the school my mother.
b. * The morning star disappeared the evening star.
(50) [Secondary predication -- Predicational be]
a. Mary left the school a nurse.
b.
The student left the party sick.
The sentences in (49) are not allowed, while the sentences in (50) are possible. The
ungrammaticality of (49) is due to the fact that be in secondary predication does not allow
identificational interpretation.
"This kind of distinction is made in Akmajian (1970) and Higgins (1973).
75
4.3. Case and the covert be
According to my consultants, there are at least several languages which do not allow NP
secondary predicates: Spanish, Dutch, and Persian. Japanese is also this type of language.
In this section, we will discuss 1) why primary predication allows NP predicates
universally while secondary predication does not, and 2) why English allows NP secondary
predicates and some other languages do not. Both of these questions will be discussed from the
perspective of Case theory, because Case theory is a logically possible licensing condition for
overt NPs.
Let us first consider the following question: Does the predicate NP (the adjunct NP) need
Case?
Chomsky (1986) states that the visibility condition does not require Case assignment to an
NP that is not theta -marked (unless this NP must transfer Case to an argument18).
(51) Visibility Condition
An element is visible for theta -marking only Wit is assigned Case.
(Chomsky 1986, 95)
In the following sentences, the Case filter requires that Case be assigned to the bracketed
NPs, but the visibility condition arguably does not.
(52) John is [a fine mathematician].
(53) [John], I consider [a fine mathematician]
(54) John did it [himself].
If we consider the visibility condition to be valid, the predicate NP does not need Case,
because it is not an argument. Then, if we assume that predicate NPs do not need Case
universally, and that the secondary predicate NP is not assigned any Case (or not Case -Checked)
uniformly, how can we explain the difference that we have seen in this section (between English type languages and Spanish -type languages)? Further, how can we explain the fact that NPs can
occur as primary predicates universally, but not as secondary predicates?
I propose that, at the very least, the predicational NP needs Case. The hypothesis is as
follows. In primary predication, the predicational NP is [- Specific] and the overt be can assign an
inherent Case universally (arguably, a partitive Case in Belletti (1988) and Lasnik (1992)). The
predicational NP in primary predication can be licensed in this way. On the one hand, in
secondary predication, the covert be does not always assign an inherent Case. Thus, in English type languages it does, but in Spanish -type languages it does not. In English -type languages the
covert be seems to have the inherent Case assigning property to license NP secondary predicates,
although such a mechanism would need further research.
One thing we have to bear in mind is that inherent Case must be related to theta
18Chomsky refers to Case Transmission in sentences such as:
(i) There is a man in the room.
76
assignment. In other words, a Case -marked NP has to be assigned a theta -role. The problem is
that a predicate NP is not selected by a head. In order to solve this problem, and to elucidate the
mechanism that licenses NP predicates, further research is required, especially on the relationship
between morphological case and abstract Case cross -linguistically.
5. Conclusion
In this paper, I argued that a PRO analysis should be adopted over a theta - marking
analysis for subject- oriented secondary predicates. I presented empirical evidence that the
predicate phrase should have a PRO subject. The occurrence of raising adjectives as secondary
predicates is one of the strongest pieces of evidence for this view. I also showed theoretical
grounds in favor of a PRO analysis. PRO analysis supports a theory of UG: VP- Internal
Hypothesis.
I claimed that the predicate phrase is AgrP because of agreement facts. In my analysis, I
considered Agr to be a null Case checker for PRO. PRO has to be licensed within AgrP, and the
head Agr meets the licensing condition for PRO under the Spec -head configuration. On the basis
of some negation data in Spanish and English, I proposed that AgrP does not dominate the Tense
node.
Finally, I dealt with the difference between primary predication and secondary predication,
and pointed out that primary predicates can either be identificational or predicational, while
secondary predicates must be predicational. I addressed the question of why NP predicates are
allowed in primary predication universally, but not in secondary predication. I also suggested the
possibility that the predicational NP as a secondary predicate might have inherent Case in
languages such as English.
References
Akmajian, A. (1970). Aspects of the Grammar of Focus in English. Ph.D. dissertation, MIT.
Baker, M., K Johnson, and I. Roberts (1989). "Passive Arguments Raised," Linguistic Inquiry
20, 219 -51.
Belletti, A. (1988). "The Case of Unaccusatives," Linguistic Inquiry 19, 1 -34.
Belletti, A. (1990). Generalized Verb Movement: Aspects of Verb Syntax. Rosenberg & Sellier,
Torino, Italy.
Carlson G. (1978). Reference in Kinds in English. Indiana University Linguistics Club,
Bloomington.
Chomsky, N. (1970). "Remarks on Nominalization," in Jacobs and Rosenbaum (eds.), 184 -221.
Chomsky, N. (1981). Lectures on Government and Binding. Foris, Dordrecht.
Chomsky, N. (1986a). Knowledge of Language: Its Nature, Origin, and Use. New York,
Praeger.
Chomsky, N. (1986b). Barriers. Cambridge, Mass.:MIT Press.
Chomsky, N. (1992). "A Minimalist Program for Linguistic Theory", MIT Occasional Papers in
Linguistics 1.
Chomsky, N. and H, Lasnik (1991). "Principles and Parameters Theory", in J. Jacobs, A. von
Stechow, W. Stemefeld, and T. Vennemann, (eds.), Syntax: A International Handbook of
Contemporary Research, Walter de Gruyter.
77
Cinque, G. (1990). Types of A'- Dependencies. Cambridge, Mass.: MIT Press.
Grimshaw, J. (1991). "Extended Projection," ms. Brandeis University.
Hasegawa, H. (1991). "Secondary Predicates, VP- internal Subjects, and Mutual C- command,"
English Linguistics 8, 1 -15.
Higgins, F. R. (1973). The Pseudo -cleft Construction in English. Ph.D. dissertation, MIT.
Hornstein, N and D, Lightfoot (1987). "Predication and PRO," Language 63, 23 -52.
Hoshi, H. (1992). "Circumstantial Predicates, PRO, and D- structure Adjunction," English
Linguistics 9, 1 -20.
Jaeggli, O. (1986). "Passive," Linguistic Inquiry 17, 587 -622.
Kitagawa, Y. (1986). Subject in Japanese and English, Ph.D. dissertation. University of
Massachusetts.
Koizumi, M. (1992). "Secondary Predicates," ms., MIT.
Koopman, H. and D. Sportiche (1988). "Subjects," Mimeographed, UCLA.
Kratzer, A. (1989). "Stage and Individual Level Predicates," In Papers on Quantification, NSF
Grant Report, University of Massachusetts, Amherst.
Kuroda, S. -Y. (1988). "Whether We Agree or Not: A Comparative Syntax of English and
Japanese," in Poser (ed.), 103 -43.
Laka, M. (1990). Negation in Syntax: On the Nature of Functional Categories and Projections.
Ph.D. dissertation. MIT.
Lasnik, H. (1992). "Case and Expletives: Notes toward a Parametric Account ", Linguistic Inquiry
23, 381 -405.
Lasnik, H. (1993). "The Minimalist Theory of Syntax: Motivations and Prospects ", ms.
University of Connecticut.
Manzini M. R. (1983). "On Control and Control Theory", Linguistic Inquiry 14, 421 -446.
McNulty, E. (1988). The Syntax of Adjunct Predicates. Ph.D. dissertation, University of
Connecticut.
May, R. (1985). Logical Form: Its Structure and Derivation, Cambridge, Mass.: MIT Press.
Nakajima, H. (1991). "Secondary Predication," The Linguistic Review 7, 275 -309.
Pollock, J. -Y. (1989). "Verb Movement, Universal Grammar, and the structure of IP," Linguistic
Inquiry 20, 365 -424.
Roberts, I. (1988). "Predicative APs," Linguistic Inquiry 19, 703 -710.
Roeper, T. (1987). "Implicit Argument and the Head -Complement relation," Linguistic Inquiry
18, 267 -310.
Rothstein, S. (1983). The Syntactic Forms of Predication. Ph.D. dissertation, MIT.
Safir, K (1987). "The Syntactic Projection of Lexical Thematic Structure," NLLT 5, 561 -601.
Stowell, T. (1981). The Origins of Phrase Structure. Ph.D. dissertation, MIT.
Stowell, T. (1983). "Subjects across Categories," The Linguistic Review 2, 285 -312.
Williams, E. (1975). "Small Clauses in English," in Kimball (ed.) Syntax and Semantics 4, 249273.
Williams, E. (1980). "Predication," Linguistic Inquiry 11, 203 -388.
Williams, E. (1987) "Implicit Arguments, the Binding Theory and Control," NLLT 5, 151 -80.
(Received January 8, 1994)
78
Evidence from Modern Greek for Refinement of the OCP*
Diane Meador
The University of Arizona ®
Department of Linguistics
1 Introduction
Modern Greek, like many languages, exhibits a phenomenon characterised by reference to the
identity of adjacent elements. It is generally agreed that the Obligatory Contour Principle (OCP)
(first proposed by Leben 1973, and first coined by Goldsmith 1976) is defined by identity and
adjacency, and that these constraints are closely associated with the notion of the tier. The nature
of this association, however, has received many various treatments.
In Modern Greek connected speech, one of two adjacent and identical vowels residing in
different words is deleted. The problems presented by the data in section 2 to previous treatments
of the OCP are threefold. First, adjacency is relevant within the moraic tier; previously, only nonprosodic tiers have been acknowledged as relevant to the OCP. Secondly, identity inheres in the
place tier; that both the moraic and place tiers are scanned by the OCP is counter to proposals in
which only a single tier is scanned. Thirdly, when words become connected, adjacent and
identical vowels are not automatically fused. Vowel hiatus resolution operates in a stage after
concatenation but before fusion in tier conflation, then, contrary to proposals that conflation is a
single -phase process.
In this article, I propose that the tiers to be scanned by the OCP are specified by adjacency and
identity functions resident in a language's grammar. The set of candidate tiers includes the moraic
tier, and more than one tier may be relevant. Furthermore, when a language incorporates rules
which resolve OCP violations in adjacency and identity, these rules preempt fusion during tier
conflation.
1.1 The OCP Defined
The definition of the OCP has evolved since Leben's (1973) examination of tonal systems: "At
the melodic level of the grammar, any two adjacent tonemes must be distinct." The principle's
interaction in autosegmental phonology is more clearly recognised in Goldsmith's (1976)
definition: "At the phonetic level, any contiguous identical (auto)segments must be collapsed into
each other." This definition foreshadowed McCarthy's (1986:208) proposal for Tier Conflation
across morpheme boundaries as one means for avoiding violations of the OCP. He defines the
OCP: "At the melodic level, adjacent identical elements are prohibited." Further refinements have
modifed reference to the "melodic level" and have constrained adjacency to within a tier:
"Adjacent identical elements on the same tier are prohibited" (Selkirk 1988, and similar definitions
in Borowsky 1987, Mester 1988, Hume 1992, and Archangeli and Pulleyblank 1995).
1.2 Tier -Adjacency
. I would like to thank Diana Archangeli for numerous discussions contributing to the arguments presented here. I
am also grateful to Colleen Fitzgerald, Chip Gerfen, Andrea Heiberg, Sung -Hoon Hong, Sue Lorenson, Chang kook Suh, and Keiichiro Suzuki for their comments and suggestions. This work was funded under NSF grant
number BNS 9023323.
79
One consequence of the inclusion of the tier -adjacency constraint is that OCP effects have
proven useful in arguments for specific feature geometries. Yip (1988 :71) notes that constraining
adjacency to a tier means that the OCP can refer only to constituents, and that "[o]nly
constituents [nodes in the feature geometry] may constitute tiers."
The general argument is that when a language resolves an OCP violation, it is because the
adjacent identical autosegments are on the same tier. It can be argued, for instance, that resolved
violations in adjacent consonants and vowels provide evidence for a common set of features and
their dominating nodes that are shared between consonants and vowels, since adjacency resides
within the same tier (for example, see Hume 1992).
The number of tiers to which identity and adjacency refer varies according to different
proposals. Hume (1992) argues that identity and adjacency are relevant within a single tier only.
Yip (1988) allows identity and adjacency within multiple tiers at or below the segmental tier.
Selkirk (1988) argues for identity within either adjacent roots or adjacent tiers below the root.
Archangeli and Pulleyblank (1995), in arguments independent of OCP effects, motivate identity
and adjacency in multiple tiers simultaneously, including root nodes and prosodic anchors. I show
in section 3 that vowel hiatus resolution in Modern Greek, as motivated by the OCP, requires
identity in one tier but adjacency in another.
If processes other than those motivated by the OCP are able to refer to adjacent constituents
above the root node, the question that arises is whether OCP- motivated processes can do so, as
well. If Yip's (1988:71) claim, that "superordinate nodes constitute tiers more often than the
subordinate nodes...", is correct, then one prediction that can be made is that prosodic tiers should
be primary candidates for scanning by the OCP. I argue in section 3 that the moraic tier must be
scanned for adjacency in Modern Greek.
1.3 Tier Conflation
Tier -adjacency also requires that some type of tier conflation is operative after morphemes are
concatenated (see, for example, McCarthy 1986 and Yip 1988). Assuming that tiers within one
morpheme are distinct from those of another morpheme, then the tiers must be conflated after
concatenation so that not only OCP effects, but also other processes like spreading may take
place. There are, however, differing interpretations of conflation. McCarthy (1986) proposes tier
conflation in a single phase, so that there are no derivational stages in which OCP violations exist.
In his model, morphemes are on separate tiers after concatenation, but before conflation. During
conflation, sequences of identical elements are automatically simplified.
Yip (1988), however, proposes a multiple -phase conflation: adjacency is created in one phase,
and another phase results in automatic fusion. In her model, languages have the option of
instantiating rules between phases. These rules, which must refer to adjacent elements in both
morphemes, include such OCP effects as dissimilation across morpheme boundaries, and
epenthesis into heteromorphemic, identical clusters. The analysis of Modern Greek vowel hiatus
resolution in section 4 supports the multiple -phase interpretation of conflation.
1.4 Instantiation of the OCP
There are also opposing views of how the OCP is instantiated within a language. McCarthy
(1986) claims that the OCP can act as either a morpheme structure constraint (MSC) or as an
output filter during a derivation, with a blocking function on rule application. McCarthy (1986)
provides examples in which syncope (in Afar) and metathesis (in Arabic) are blocked if they
80
would otherwise result in OCP violations. Yip (1988) argues that the OCP can also trigger rule
application, and as a consequence, rules no longer need to be specified with identity conditions.
By her account, any rule which involves removing identity or adjacency of a target and trigger
pair is one triggered by the OCP. She cites a number of OCP- motivated processes in addition to
syncope and metathesis: insertion (in Japanese Rendaku; Itô and Mester 1986), feature -changing
(in English Spirantization; Borowsky 1986), degemination (in Seri Glottal Degemination),
dissimilation (in Cantonese), and assimilation (in Berber). Modern Greek provides an example of
vowel deletion triggered as a means of resolving OCP violations.
A number of accounts have been proposed which balance the universality of the OCP against
language -particular effects. McCarthy (1986) states, for example, that MSCs prohibiting
language -particular featural cooccurrence are motivated by the universal principle prohibiting
adjacent identical elements. Odden (1986) argues against a universal principle in favor of
language -specific rules on the grounds that there are ordering effects. He also claims that MSCs
are not related to the OCP, but are language -particular cooccurrence constraints. Yip implicitly
provides for parameterization of a universal principle in terms of which features or groups of
features appear on a separate tier, and how OCP violations are alleviated (the latter through a
combination of language -particular rules and multiple -phase conflation). Mester (1988) obtains
language -particular effects via hierarchical tier orderings which vary between languages. In
section 5, I propose instantiation of the universal OCP in terms of language -particular conditions,
which specify what and how many tiers are relevant to identity and adjacency, and how OCP
violations are to be alleviated.
2 Vowel Hiatus Resolution in Modern Greek
The vowels in the Modern Greek inventory are: [s, i, a, o, u]. A process labelled
"Degemination" by Kaisse (1985) occurs when identical vowels in two different words become
adjacent in connected (not necessarily casual or fast) speech. The first of the identical pair of
vowels is at the end of the first word and the second begins the next word; the words in question
may occur anywhere within a sentence (Kaisse 1985). One of the vowels is absent in conned
speech where both are present when the words are spoken in isolation. Figure (1) illustrates.
Assuming that the isolated word most closely corresponds to the underlying representation, the
general process must be one of deletion. That it is deletion is clear when the alternative is
considered: insertion of a final vowel in a set of arbitrary words in isolated speech. No patterns
for the set of relevant words or for the various vowel qualities are discernible under the insertion
alternative. Therefore, hiatus of identical vowels between words in Modern Greek must be
resolved through deletion. The process is not true degemination; the vowels do not share the
same root node since they reside in different words.
Three observations can be made from this data. First, regardless of which vowel is stressed in
the isolated words, stress is preserved when a vowel is deleted in connected speech (la - le).
Either a consistent direction of rule application is preserved, and stress is transferred to the
remaining vowel where necessary, or only the stressless vowel deletes. Neither of these analyses
are favored by this data alone. Secondly, the consonant preceding a sequence of round vowels in
Data was compiled from Elefheriades (1985) and Kaisse (1985).
81
(1) Identical Vowel Elision
Isolated Words
Connected Speech
Gloss
(a) [se smsna]
[8sorítE smpistos]
(b) [tí ins]
[pósi ¡ne]
(c) [tá álla}
[sdmsna}
[ósorítÉrnpistos]
[tine]
[pósins]
'to me'
'he is considered trustworthy'
[tálla]
[tá álova]
(d) [tO anomal
[táloYa]
'the others'
'the horses'
'the name'
'the dream'
'of the sky'
'What is it ?'
'How many ?'
[twanoma]
[ta .4niro]
[t"`Oniro]
(e) [tú uranú]
[mu Aka]
(t) [aYóri itrxsts]
[msYálu slá.fyu]
[twúranú]
[mwúka]
'my figs'
[aYóri «xsts]
'boy comes'
'big deer (gen. sg.)'
[msYálu sláfyu]
the elision environment becomes rounded, as in (1d) and (le). Lastly, nonidentical vowels do not
necessarily elide, as shown in (1t) by the presence of adjacent vowels in connected speech.
Because a vowel is deleted just in case it is identical to an adjacent vowel in another word, the
process is likely motivated by the OCP. That consonants become rounded subsequent to deletion
of round vowels will be crucial in arguments for identity and adjacency, and how these notions
relate to the OCP. The processes of vowel deletion and consonant rounding are presented
schematically in (2).2
(2) Schematics
(a) Delete vowel
t
o
o noma
(b) Link [ +rd]
o noma
tw
µ
µ
# #
JFa
21
#
#
+rd
I
assume that the left-most vowel deletes; the assumption is not crucial to my arguments.
82
In the case of round vowels, [+rd] must remain available to the preceding consonant. The
indication is that the mora is deleted, rather than the vowel's featural content. If only the features
attached to the mora were deleted and the mora remained, either vowel lengthening or realization
of the unspecified [c] might be expected, but neither occurs (see Meador (1993) for arguments for
vowels specifications in Modern Greek).
By virtue of the requirements of identity and adjacency, the OCP motivates the deletion rule.
But therein lies the challenge to the OCP: the offending features are not removed. That is,
identical and adjacent features remain after the rule has applied, a clear violation of the OCP if
identity and adjacency inhere within a single tier.
3 The Single Tier vs. The Multiple Tier Hypotheses
A closer examination of the proposals for single- and multiple -tier adjacency is needed. The
following questions become germane: Can adjacency and identity refer to more than one tier? In
which tier(s) do adjacency and identity inhere? Should tiers above the root node be subject to the
OCP?
Hume (1992) maintains the strongest form of the single tier hypothesis in her argument that
identity and adjacency inhere within one tier. The argument is based on her definition for
adjacency, given in (3).
(3) Adjacency (Hume 1992:138)
The elements x and y are adjacent on tier n iff no element z on tier n
intervenes.
The OCP, according to this view, scans a tier for identical and adjacent elements. The implicit
assumption is that the tiers to be scanned are below the root node. In Modern Greek, however,
the OCP would not provide a motivation for vowel deletion, since adjacent identical elements
remain on any of these tiers. One hypothesis might be that featural identity of adjacent vowels is
the violating factor in Modern Greek. If so, then the OCP should scan the featural tier. After
mora deletion, the features remain in violation, so this hypothesis is rejected. Figure (4) illustrates
this.
(4) Identity in the Feature Tier
t
o noma
##
Identity and
Adjacency Remain
83
Assuming Hume's (1992) feature geometry, another hypothesis would require the OCP to scan
the vocoid (VOC) tier. In her model, the VOC node, dominating [LAB], would spread from the
vowel to the preceding consonant subsequent to the mora deletion posited here.' Mora deletion is
shown in (5a), and consonant rounding in (5b). This hypothesis must be rejected as well, since
identity in adjacent VOC nodes remains after mora deletion.
(5) Identity in the VOC Tier
t
w
o
#
noma
#
(a)
Root
Cons
Place
.
(b )
Identity and
Adjacency
Remain
Voc`\
Place
I
I
LAB
LAB
Yip (1988) proposes that adjacency can refer to multiple tiers. Implicit in her arguments is that
these tiers are at or below the segmental (or C -V) tier. In her analysis of English epenthesis in the
plural and past tense, for example, the OCP scans for adjacent, identical specifications for
[continuant] and [strident], each of which resides on a different tier. The OCP violation
incorporating a multiplicity of tiers is illustrated in the generic case in (6).
(6) Identity and Adjacency in Multiple Tiers
*
X
al
X
a
Hume (1992) does not discuss elements above the root node.
84
Under this proposal, identity in Modern Greek might inhere in the segment tier, while
adjacency refers to the vowel tier and all that it dominates: ROOT, PLACE, LABIAL, etc. In an
examination of assimilation in Berber, Yip (1988) argues that one of two identical matrices is
delinked, providing an empty slot as a target for spread. She recognizes the problem, similar to
that in Modern Greek, of remaining identity and adjacency after delinking:
It is possible that the spreading itself alleviates the OCP violation, but this is not obviously so,
since the offending matrix would still be present, albeit delinked. For this reason I have
assumed it is first deleted by the OCP- triggered rule, and this then leaves it empty and available
for a spreading rule.
[Yip 1988:78; fn. 13]
In Modern Greek, the matrix delinked from the deleted mora cannot itself be deleted in its entirety
since its featural content must remain available for the consonant rounding rule.4
Selkirk (1988) also argues that adjacency refers to multiple tiers, but specifies these tiers in
either roots or tiers below the root. Her definition for the tier is given in (7), followed by the
definition for adjacency in (8).
(7) Tier (Selkirk 1988:6)
Def.: Identical features define a tier iff they are dependent on
identical features (i.e. have identical heads)
(8) Adjacency (Selkirk 1988:8)
Two feature specifications are adjacent if they are either root -adjacent or tier- adjacent.
a. Del. Two feature specifications are root -adjacent if they
are dominated by adjacent root nodes.
b. Del. Two feature specifications are tier -adjacent if they are
adjacent on the same tier.
The hypothesis for identity and adjacency below the root was rejected on the basis of the
argument that identity and adjacency of the offending features remain after mora deletion. The
question of identity within adjacent roots raises two possibilities in Modern Greek: either the root
which was dominated by the deleted mora remains, or it deletes as well. Assuming Selkirk's
(1988) feature geometry and representations for primary and secondary articulations (the latter
being in a dependency relation to the former), the first possibility is illustrated in (9). Mora
deletion is shown in (9a), which should be motivated by identical features with adjacent roots in
this model. If the root node does not delete, then the OCP violation remains. Figure (9b)
41t might then be argued that [ +rd] spreads not from the delinked matrix, but from the remaining matrix of the
word -initial vowel, in which case deletion of the delinked matrix is an adequate solution. Other cases of vowel
hiatus resolution in Modem Greek (those involving nonidentical vowels) indicate that the featural content of the
first vowel must remain for consonant rounding. For example, [a!loa2 6!rxdc] ('horse comes') in isolated speech
corresponds to [a!losw c!rxdc] in connected speech (Kaisee 1985). The only source for rounding in this case is the
featural content of the deleted word -final vowel.
85
illustrates how subsequent consonant rounding, if any, alleviates the violation, since the instances
of [round] have different heads. The result of mora deletion in the environment of non -round
vowels is given in (9c), to show that when consonant rounding is not applicable, the violation
remains.
(9) Root Remains
(a)
Mora Deletion due to Root -Adjacency; Root Remains
( "RC" = consonant root; "RV" = vowel root)
t
o noma
o
RV
RC
Place
#
Root
Adjacency
Remains
# RV
Place
Place
Labial
Labial
i
I
round
round
Coronal
Identity
Remains
(b) Violation Alleviated through Labial Spread
t
w
o noma
(
RC
RV
Place
Place
Coronal
#
Root
Adjacency
Remains
# RV
Place
Labial
Labial
Nonidentical
Due to Different
Heads
round
(
round
86
(c) Violation Remains if Vowels are Nonround
t
RC
Place
Coronal
i
RV
#
#
Root
Adjacency
Remains
RV
Place
Place
Dorsal
Dorsal
l
ns
Identity
Remains
I
Alternatively, extraneous structure, including both the root and place nodes, may be deleted
after the dominating mora has been deleted. This alternative seems likely, since the mora, root,
and place nodes are not subsequently filled (neither vowel lengthening or the default vowel [e]
appear). In this case, the root -adjacency violation would be alleviated, although identity and
adjacency in the articulator tier would remain. If the head is indeed the root, this alternative
would be a satisfatory solution; the root tier would be scanned by the OCP for adjacency while
the articulator tier would be scanned for identity.
However, the model depends on specifying the root node as belonging to either a consonant or
a vowel, which is informed, presumably, by the prosody. Furthermore, while deletion of the root
to alleviate the adjacency violation would be motivated by the OCP, this account does not explain
why it is the mora that deletes. These observations indicate that the mora, as the prosodic anchor,
is the "head" instead. The implication is that the moraic tier is a candidate for scanning by the
OCP in languages like Modern Greek. If so, then adjacency in the moraic tier should be relevant.
Archangeli and Pulleyblank (1995) argue for the relevance of moraic adjacency in processes
independent of the OCP. For example, morae are often the prosodic anchors, or targets, in
processes of vowel harmony. Any phonological process is, in their model, subject to the Locality
Condition (Archangeli and Pulleyblank 1995:11), which states that "[p]honological relations
respect Adjacency and Precedence." Adjacency is defined as in (10).
(10) Adjacency (Archangeli and Pulleyblank 1995:20)
a is structurally adjacent to ß iff:
(a) at least one of the two is unassociated, both are on the same tier,
and no element intervenes between the two on that tier; or,
(b) both a and ß are associated to the same anchor tier and no anchor
intervenes on that tier between the anchors to which a and ß are
associated.
87
The argument for adjacency in the moraic tier in processes independent of the OCP can be
extended to those which are motivated by the principle. For example, the configuration in (11) is
ill- formed because "[t]wo identical autosegments are associated to adjacent prosodic anchors
(Archangeli and Pulleyblank 1995:20)." Identity is relevant in the a tier, and adjacency is relevant
in the moraic tier.
(11) Identity in one Tier, Adjacency in Another
a
a
This account differs from the single -tier hypothesis since more than one tier is scanned by the
OCP. Unlike Yip's (1988) or Selkirk's (1988) model, violations may also occur on the moraic
tier. Figure (12a) illustrates the alleviation of the OCP violation in Modern Greek through the
removal of adjacency (morae are no longer adjacent when one is deleted), despite remaining
identity in the place node and all that it dominates when [ +rd] spreads (12b).
(12) Moraic Adjacency; Identity in Place
t
o
#
#
o
noma
Adjacency
Removed
(a)
Root
Place
Identity
Remains
I
+rd
+rd
By this account, then, the OCP provides motivation for mora deletion. Adjacency and identity
may refer to more than one tier, and the moraic tier should be included into the set of those
scanned by the OCP.
4 Single- vs. Multiple -Phase Tier Conflation
A problem remains concerning how two vowels become adjacent and create the environment
for hiatus in Modern Greek, given that they reside in different words. In the definitions for
adjacency given in (3), (8), and (10), no element may intervene between the two elements in
question on a given tier. The vowels in hiatus are in different morphemes, but morphemes reside
on distinct tiers (McCarthy 1986, Yip 1988). Therefore, the vowels cannot be adjacent unless a
mechanism exists to align the tiers between morphemes.
88
McCarthy (1986) proposes a process of Tier Conflation to align the tiers. The single -phase
process he describes, however, results in adjacency only of nonidentical elements. This proposal
is problematic since vowel hiatus resolution in Modern Greek is motivated by adjacent identical
elements. Modern Greek provides support for Yip's (1988) argument for a multiple -phase
solution, in which adjacency of identical elements is possible in an intermediate stage. These
proposals are examined below.
McCarthy (1986) argues that there is no derivational stage during conflation at which an OCP
violation exists. After concatenation of two morphemes, but before conflation, the morphemic
content resides on different tiers, as in (13a). Two identical elements are automatically fused
during conflation in order to avoid an OCP violation, as in (13b).
(13) Single -Phase Tier Conflation (McCarthy 1986)
(b) During Conflation
(a) After Concatenation
t
C
+C
C
C
v
t
Because identical elements are never adjacent on the same tier, conflation does not cause OCP
violations in this model. The prediction is that there should exist no rules motivated by the OCP,
then, that are applicable when two morphemes or words become adjacent. For Modern Greek in
particular, vowel lengthening would be expected instead of vowel deletion. Because vowel
deletion is motivated by the OCP in the hiatus environment in connected speech, an alternative
solution is required.
Yip (1988) proposes an alternative in which tier conflation has two stages, one at which
adjacency is created. If a language has a specific rule to remove an OCP violation, that rule
applies at this stage. If not, then fusion takes place. Figure (14) illustrates.
(14) Multiple -Phase Tier Conflation (Yip 1988)
(a) After concatenation
t
C
+C
(b) First stage
C
C
89
(c) Second stage
In her motivation for the first stage, Yip states that:
since heteromorphemic melodic elements can only be adjacent after Tier Conflation, any rule
that needs access to both elements, and in which they are clearly distinct, is evidence against ...
automatic fusion ...
[1988:69].
She provides several examples, including dissimilation across morpheme boundaries and
epenthesis into heteromorphemic identical clusters. As another example of dissimilation, vowel
hiatus resolution in Modem Greek offers further support for multiple -phase tier conflation. It
cannot be the case that vowels are fused, as in the proscribed case in (15), since vowels do not
lengthen in connected speech.
(15) Vowel Lengthening Proscribed
t "'
noma
o:
µ\
/µ
V
Root
(
Place
+rd
Given that mora deletion in Modem Greek is motivated by the OCP, adjacency must be
introduced after words are concatenated in connected speech. Since fusion is preempted, hiatus
resolution must take place in the first stage, as shown in (16).
(16) First Stage Hiatus Resolution
(a) Concatenation
t
3
(b) First Stage
noma
o
t
o
o
+rd
Place
Root
Root
Place
µ
+
I
µ
+rd
Root
Place
90
I
+rd
nome
5 Parameterization of a Universal Principle
The most prevalent assumption implicit to OCP arguments is that it is a universal principle,
given the abundance of rules across languages which remove adjacency of identical elements.'
Yet configurations involving adjacent identical elements that are well -formed in one language may
be ill- formed in another. Languages may differ according to which tiers are relevant for identity
and adjacency, and whether or not only a single tier is scanned. For example, adjacent identical
specifications for place and manner in English are ill- formed (Borowsky 1987); in Berber,
adjacent coronals are prohibited (Yip 1988); Cantonese prohibits adjacent labials ( Yip 1988,
Hume 1992); root -adjacent labials in Berber are ill- formed (Selkirk 1988); and adjacent identical
elements on the segment tier are prohibited in Semitic languages (McCarthy 1986).
Languages may also differ according to whether they provide rules alleviating OCP violations
after concatentation of morphemes, or whether the violations are alleviated through fusion during
conflation. The mora deletion rule in Modern Greek provides an example of the former, whereas
the Semitic languages appear to employ fusion (McCarthy 1986).
Finally, the grammar of a language may include either constraints preventing OCP violations,
such as the MSC in Cantonese prohibiting two labials within a single morpheme (Yip 1988, Hume
1992), or rules which alleviate violations arising through concatenation, as in Modern Greek.
Of the languages which employ rules motivated by the OCP, some may utilise epenthesis, as in
English past tense formation ( Borowsky 1987), and others may incorporate deletion, as in
Modern Greek.
Each of these language particular effects share a commonality: identity and adjacency. I
propose, then, that the OCP is a parameterised universal. I follow Yip (1988), who implicitly
provides for parameterization in terms of which tiers are scanned, and how OCP violations are
resolved (whether through MSCs, or through language -particular rules, and therefore two -stage
conflation, or both). If the values for these parameters are specified for a language, then the
conditions for identity and adjacency need not be specified for the rules themselves (cf. Yip 1988,
who argues that identity need not be specified in rules).
Along these lines, Hong (in preparation) has proposed that (for harmony systems at least)
identity is a function taking as its arguments the trigger and target of a rule over a specific tier.
For example, the Identity Condition on Yawelmani round spread is specified as:
(17) Identity Condition (Hong in preparation)
Identical (Argument, Target) <HIGH>
That is, [round] spreads from an argument (or trigger) to a target identical in height. This
function is independent of the rule of round spread itself, since the rule does not specify height
identity. Alpha notation (representing element identity) is therefore no longer necessary in rule
formalizations.
The identity function, motivated independent of OCP effects, should also be available for
reference by the OCP if it is stated in a language's grammar. The condition governing mora
deletion in Modern Greek specifies identity in the place tier and all that it dominates. Since vowel
5See Odden (1986), however, for arguments that OCP effects are simply language -particular rules.
91
hiatus resolution in Modern Greek is an OCP effect, a value must also be set for an adjacency
parameter. I propose that another function exists in Modern Greek which allows scanning of a
tier other than that to which identity refers. The Adjacency Condition in Modern Greek,
therefore, specifies the moraic tier. The conditions proposed for Modern Greek with their
parameter values are shown in (18).6
(18) Modern Greek Identity and Adjacency Conditions
Identical (Argument, Target) <PLACE>
Adjacent (Argument, Target) <Mo &>
Consequently, the rule which specifies only the deletion of a mora in a left -to -right direction of
application (omitting alpha notation) will be applicable only if adjacent morae are specified with
identical places of articulation.
6 Conclusions
Modern Greek offers several implications for the OCP. One is that distinct tiers may be
available for adjacency and identity, which is counter to the single -tier hypothesis. Since mora
deletion adheres to conditions of identity and adjacency, the prosodic tier must be a candidate for
scanning by the OCP, in addition to tiers at the segmental tier or below. Because vowels in hiatus
are not realized as one long vowel, they are not fused, which refutes the hypothesis that conflation
is a single -phase process. Finally, if the condition of identity is specified for a language, but not
within its rules, then the condition of adjacency should also hold for the language as a whole when
more than one tier is relevant.
6The definition and role of arguments (or triggers) and targets is not entirely clear in deletion rules, and is left as a
problem beyond the scope of this article.
92
References
Archangeli, D. and D. Pulleyblank (1995). Grounded Phonology, Cambridge: MIT Press.
Borowsky, T. (1987). "Antigemination in English phonology," LI 18. 671 -678.
Eleftheriades, O. (1985). Modern Greek: A Contemporary Grammar. Palo Alto: Pacific Books.
Goldsmith, J. (1976). Autosegmental Phonology, doctoral dissertation, MIT.
Hong, S. -H. (in preparation). Issues in Rounding Harmony, doctoral dissertation, University of
Arizona.
Hume, E. (1992). Front Vowels, Coronal Consonants and their Interaction in Nonlinear
Phonology, doctoral dissertation, Cornell University.
Kaisse, E. M. (1985). Connected Speech: The Interaction of Syntax and Phonology, Academic
Press, Inc., Orlando.
Leben, W. (1973). Suprasegmental Phonology, doctoral dissertation, MIT. (Distributed by
Indiana University Linguistics Club, Bloomington).
McCarthy, J. (1986). "OCP effects: Gemination and antigemination," LI 17:2. 207 -263.
Meador, D. (1993). "Markedness, recoverability, and prosodic domain: A constraint -based
approach to vowel elision in Modern Greek," ms., University of Arizona.
Mester, R. A. (1988). Studies in Tier Structure, doctoral dissertation, UMass. New York:
Garland Publishing.
Odden, D. (1986). "The role of the OCP in phonological theory," Lg 62:2. 353 -383.
Selkirk, E. (1988). "Dependency, place, and the notion 'tier'," paper presented at the Annual
Meeting of the Linguistic Society of America.
Yip, M. (1988). "The Obligatory Contour Principle and phonological rules," LI 19:1. 65 -100.
Diane Meador
Department of Linguistics
Douglass 600 E.
University of Arizona
Tucson, Arizona 85721
email: [email protected]
http://aruba.ccit.arizona.edu/-meador
93
94
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

advertisement