Professional Documents
Culture Documents
Nonlinear Phonology: Further
Nonlinear Phonology: Further
REVIEWS Further
Annu. Rev. Anthropol. 1989. 18:203-26
Quick links to online content
Copyright © 1989 by Annual Reviews Inc. All rights reserved
NONLINEAR PHONOLOGY
Douglas Pulleyblank
Department of Linguistics, University of Ottawa, Ottawa, Ontario, Canada KiN 6N5
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
INTRODUCTION
'Morphemes are linguistic units of which words are composed-typically, but not invariably,
the smallest fonnal units that can be assigned independent meanings (see 2). Boundaries between
morphemes are indicated here by the symbol "+".
2By standard linguistic convention, ungrammatical examples are preceded by an asterisk (*).
203
0084-6570/89/1015-0203$02.00
204 PULLEYBLANK
consideration to the vowel [re] that follows [s], we see somewhat different
values for the same features: (a) unlike [s], ere] does constitute a syllable
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
peak, hence [+syllabic]; (b) like [s], the air passage through the oral tract is
continuous, and there is similarly no passage of air through the nasal cavity,
hence [+continuant, -nasal]; (c) unlike [s], the vocal cords are vibrating
during the production of ere], hence [+voiced].4 Looking at the same features
as involved in the final segment em] of Sam, we observe still different values:
(a) as with [s], em] is not a syllable peak, hence [-syllabic]; (b) unlike both
[s] and ere], air passage through the oral cavity is completely obstructed for
em], hence [-continuant]; (c) also unlike both [s] and ere], air is allowed to
pass through the nose, hence [+nasall; (d) finally, as with ere], the vocal
cords are in vibration, hence [+voiced]. The phonological representation of
Sam in a linear theory would therefore be (in part) as follows:
m
r ]r ]r ]
� ����� �ant
-nasal
-voiced
i
: ���;i��ant
-nasal
+voiced
= ����� �ant
+nasal
+ voiced
i
1.
.. . .
..
A large body of work over the last 15 or so years has demonstrated the
inadequacy of such linear representations. On the one hand, it has been
argued that the notion of "segment" does not correspond in a simple fashion
with discrete sets of binary feature specifications. In many cases, one or more
features may characterize a whole set of segments, and in others, a feature
3Here and throughout this paper, square brackets ("[]") are employed to indicate phonetic
transcriptions (of varying degrees of detail). Such transcriptions represent in a standardized
alphabet the actual sounds of an utterance. For example, the word "phlegm" would be represented
as [flem]: Orthographic "ph" represents a single sound [t]; the sound of orthographic "e" in this
word is represented phonetically as leI; orthographic "g" does not represent a phonetic segment,
hence is excluded from the transcription. Aspects of transcriptions that are crucial are commented
on in the text.
4For simple verification, place a finger on the outside of the larynx and produce first a
prolonged [a:], then a prolonged [s], comparing the presence vs absence of vibration in the two
segments.
206 PULLEYBLANK
value may change during the course of a single segment. (Examples of both
types are discussed below.) The result is that the notion of "segment" must be
separated conceptually from the notion of the features by which segments are
realized.5 In the example of Sam this necessitates the identification of three
segments (or SKELETAL units), namely eve. This move has immediate im
plications for a feature such as [±syllabic). Whereas a segment's value for
continuancy or nasality can be assigned independently, the determination of a
segment's syllabicity is in large measure a result of how the segment is
grouped together with other segments into constituents. For example, the
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
vowel [i] of a word like bee is featurally comparable to the semi-vowel [y] of
a word like yet. The crucial difference between the two segments is that the
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
former constitutes the peak of a syllable, while the latter constitutes a syllabic
margin. Hence [±syllabic] can be abandoned in favor of a direct representa
tion of syllable structure. And as a final step in the reorganization of the
features in Example I, there is considerable evidence that distinctive features
themselves are not an unorganized set of features, each with comparable
status. Instead, such features can be shown to be organized into various sets.
For example, the total set of features (the ROOT) can be divided into
LARYNGEAL and SUPRALARYNGEAL specifications, along with various sub
'��\�-
(±c = [±continuantJ) 2.
( ...... laryngeal features
+ - � - - - � _ - - -
: . : I (:!: V [:!: voiced])
-,,'/
=
-v _ _ _ _ v _ _ _ .....J
c
0--0 0-___-\ . ... supralaryngeal features
V I
+
(±n = [± nasal])
L-______-n_ +n ____
5 m
5For various proposals concerning the nature of the "segmental" level of representation, see,
for example, McCarthy (26), Clements & Keyser (9), and Levin (24). Recently, even the notion
of "segment" itself has been abandoned in certain proposals in favor of different conceptions of
coordinating prosodic units such as the mora. See, for example, McCarthy & Prince (27) and
Hayes (17).
NONLINEAR PHONOLOGY 207
Constituency
In a linear theory such as that illustrated in Example 1, the only constituents
into which phonological segments are grouped are defined morphologically or
syntactically. That is, phonological segments are grouped together to make up
morphemes, to make up words, to make up phrases, and so on. But an
examination of data from a wide array of languages makes it clear that
segments can also be grouped into constituents on purely phonological
grounds.
For example, Clements & Keyser (9: 59-60) discuss five independent
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
0'
Itk 4.
Within the foot, a "head" syllable (ultimately the stressed syllable) is located
at either the right or left periphery (depending on the language) and "nonhead"
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
W W
� ��
�a � �a �"'x�a 5.
k
11\ 11 !1
a n g a r 00 o
I 11 /1
n o m a t
11
/ oI poeI a
I
kAngaROO OnomAtoPOEIa
These structures are interpreted as follows: Main stress is assigned to the final
foot in each word (roo in the first case, poeia in the second). Within the
stressed foot, prominence falls on the leftmost syllable. In feet that do receive
main stress, a secondary stress falls on the head-again, the leftmost syllable
of the foot.
Apart from syllables and constituents established for purposes of stress,
various larger phonological constituents constitute the domain of application
6See, for example, Kahn(21), Clements & Keyser(9), Steriade (37), Levin(24), McCarthy &
Prince (27), Ito (19, 20), and Hayes (17).
7See, for example, Liberman & Prince (25), Hayes (14, 15), Hammond ( 1 3), Halle &
Vergnaud ( 1 2); for a contrary view about the role of constituency, see Prince (30).
8Keep in mind that phonological constituents are actually erected over phonological segments,
not orthographical symbols. For example, the sequence "oei" in onomatopoeia represents a single
vowel, not three separate vowels. The structures here are slightly simplified, and the reader is
referred to already cited work for details. Heads of foot ("�") and word ("W") level constituents
are indicated by a small circle. Main stress is indicated by capital letters, secondary stress by
small capitals.
NONLINEAR PHONOLOGY 209
Segment Structure
In the remainder of this article, I discuss a rather different aspect of the move
away from linear phonological representations. In addition to the grouping
together of segments into phonological constituents, much recent work in
phonological theory has demonstrated the need to establish a fairly high
degree of segment-internal structure.
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
referred to as contour segments. These are segments whose value for some
feature changes during the course of the segment. For example, affricates are
segments that begin with a complete stoppage of air (as in the stops [p] and
[t]) and that finish with friction (as in the fricatives [f] and [s]). Examples of
affricates are the initial consonants of the German words Pferd [pfe:rt] "horse"
and Zeit [tSayt] "time". A second type of contour segment involves pre
nasalization, a situation where the first portion of a segment is nasal but the
latter portion is oral, as in the segments [mb] and ["d] of Tiv: imbor "spring of
water", vende "refused". A third example is that of a contour tone. In many
languages, the pitch values on vowels serve a contrastive function. In Yoruba,
for example, the consonant vowel sequence [kJ] produces three completely
different words depending on thc pitch of thc voicc: on a high pitch, [k:S]
"build"; on a mid pitch, [kj] "write"; on a low pitCh, [k3] "refuse". Two types
of tones can be distinguished, level tones and contour tones, level tones
corresponding to a steady pitch and contour tones to a changing pitch. The
rising ( ) and falling tones ("''') of the following examples from Yoruba
"V"
9In Example 6, ":!:c" = [:!:continuantl, ":!:n" = [±nasal], "L" = low tone, "H" = high tone.
210 PULLEYBLANK
7.
LH stem: saa "seed necklace"
sami "conversation"
saabi "roofing"
mWar�gf "younger brother"
In Example 7, the five tonal possibilities are illustrated with stems contain
ing enough vowels to realize every tone of a melody individually. Land H
stems are realized by aSSigning the single tone of the melody to all vowels of
the stem:
IOpor purposes of the discussion here, stem can be interpreted as "uninflected verb or noun,"
the precise definition being unimportant. For detailed discussion, see Hyman (18). The examples
that follow abstract away from any pre-stem prefix that might occur on the surface, and give tonal
forms appropriate for a stem occurring immediately preceding another stem.
NONLINEAR PHONOLOGY 211
L stems: b a b a I a b � a
I
L
V
L L
H stems: b a b a g a b a � a
8.
I
H
V
H H
LH and HL stems are realized by distributing the two tones of the melody over
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
the available vowels of the stem. In the case of the sequence H L, the Hand L
are assigned by a left-to-right convention, the final Lsubsequently assigned to
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
HL stems; k a a k a r a k a r a g a
I I
H L H
I I
L
I V
L
9.
H
LH stems are comparable except that in examples with three vowels, the
associations are modified in such a way that the initial Ltone of the melody is
linked to the first two vowels:
LH stems: s a a s a mI g I
I I I I I 10.
L H L H H
Of particular interest, however, are the cases where there are not enough
vowels for the tones of a particular melody to be assigned to individual
vowels. It is precisely in such cases that contour tones appear�that is, tones
with a changing pitch curve. Corresponding to the HL cases of Example 9, a
falling tone is observed on a monosyllable; and corresponding to the LH cases
of Example 10, monosyllables exhibit a rising tone:
HL stems: LH stems: s a
1\
LH 11.
kii "pick" sa "weaving knot"
Similarly, monosyllables with the LHL pattcrn surface with a complex tonal
contour that first rises then falls, and disyllables with the LHL pattern surface
with a L tone on the first syllable and a falling contour on the second:
II
See Hyman (18) for a discussion of Kukuya and Pulleyblank (31) for a general discussion of
the conventions by which tones are associated.
212 PULLEYBLANK
LHL stems: bv I
;1\
LH L
k Li HAL Y 12.
All such contours are the expected result of an account with the following
properties: (a) five tonal melodies are posited for Kukuya: L, H, LH, HL,
LHL; (b) these melodies are freely combined with stems of any number of
vowels; (c) combinations of tones on a single vowel are permitted, resulting in
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
Example 1, it would not be expected that some single feature or set of features
would have a status completely independent of the others. That is, since all
features are simply components o{unordered feature matrixes, there would be
no reason to expect the existence of melodic units composed of some subset of
such unstructured features. In a nonlinear model, however, the existence of
melodic sequences (such as the tonal melodies of Kukuya) is completely
un surprising since particular feature classes can be assigned a status that is
independent of both other features and their assignment to particular seg
ments.
In addition, since sequences of features cannot be assigned to a single
segment within a segmental model, the cases in Examples 11 and 12 would
require the postulation of additional, intrinsically contoured tone features.
The immediate problem raised by such an approach concerns the distribution
of such tones. A rising-falling tone must be restricted to monosyllabic stems;
rising and falling tones must be restricted to monosyllabic or disyllabic stems,
with the additional restriction on a disyllabic stem that such contours appear
only on the second vowel. Such clearly ad hoc restrictions are unnecessary if
contours are formally analyzed as consisting of a sequence of simple L and H
tones assigned to a single vowel-the nonlinear analysis.
This analysis can also explain contextual conditions on the creation of
contours even in languages that do not exhibit any evidence for tonal melo
dies. For example in Yoruba (31), a rising contour is created when a H tone
immediately follows a L tone-a simple assimilation of the H toned vowel to
the preceding L toned vowel. Although the three lexical tones of Yoruba are
all level (high, mid, low), rising contours appear on the final syllable of words
like the following:
a. lwe "book"
h. eke "Lagos" 13.
c. likpii "scar"
For a linear theory, such cases present the same type of problem seen above.
NONLINEAR PHONOLOGY 213
a. i w e � i w e c. a kp a � a kp a 14.
I I V1 I I V1H
L H L H L H L
phonology makes the claim that all associations of features to segments are
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
Pre-pausal forms:
H stem: bii "oil palms"
baa "cheeks"
baga "show knives" 15.
biiiimii "liana"
balaga "fence"
As seen in Example 15, all H tones of a stem are affected by this rule. In an
autosegmental approach, this is straightforwardly represented by the follow
ing rule (18):
H - M I _ II 16.
a. ba ba b. baga
I I
H�M H �
V
M H � M
17.
214 PULLEYBLANK
environment):
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
18.
from biigii "show knives" with respect to the tonal lowering rule discussed,
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
because the former has two (morphologically motivated) H tones while the
latter has a single H tone. This type of representational distinction is crucial in
understanding the often rather special behavior of geminate (long) segments.
Geminates often present something of a paradox. On the one hand, such long
segments can behave as though they are a sequence of short segments; on the
other hand, they can behave as though they are a single, inherently long
segment. Within linear models of phonology, such ambiguity results in
serious problems of representation. Within nonlinear models, on the other
hand, there is no ambiguity per se: Sequential behavior results from a rule
applying to the skeletal level of representation; single segment behavior
results from a rule applying to the melodic level.
Geminate
vowels:
[1
1\
Geminate
consonants:
[1
/\
Melodic level 20.
vv C C Skeletal level
A single rule of skeletal deletion accounts for both the examples with short
vowels (Example 21) and the examples with long vowels (Example 22);12
�
a. n a g a m 0 b. b m 0 s e
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
I I I I I
C V C V
J� C V J �� v Deletion 23.
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
¢ ¢
n a g a m b m 0 s e
I I I I I I I I I \ \
c V C V C C V C V C V
Processes of this type that affect solely the skeletal level of representation are
entirely consistent with a sequential analysis of long vowels.
Single segment behavior Such behavior contrasts, however, with the results
of rules that affect the actual' feature make-up of a geminate segment. Consid
er, for example, a rule of palatalization in LuGanda. Velar consonants [k, g)
are optionally pronounced as palatal [c, j] when they precede either the vowel
[i) or the semi-vowel [y] (7). For example, both velar and palatal variants are
possible for words such as the following:
This rule, which affects not the skeletal level of representation but the actual
melodic composition of a segment, applies in a manner so as to affect an
entire geminate:
Note specifically that the geminate does not exhibit sequential properties. A
form like *oiugji is not a possible variant. In addition, note that even the
conditioning environment is melodic rather than skeletal in nature: The
121 am simplifying the process somewhat here. For details, see Piggott (29) and Halle &
Vergnaud ( 1 2).
NONLINEAR PHONOLOGY 217
Ci) nt u
Cff)
a. b. o I u
\ I /\ V I I I I
c vv c v v c v ccv
Rvr H
o I u J 1 a atarlzatlOn
PI ' 26.
I I I
c vv c v V C V CC V
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
13There is dialectal variation as to the segments affected by spirantization. See, for example,
Bagemihl (3).
218 PULLEYBLANK
important to consider both the conditioning environment for the rule and the
change the rule brings about. First, since the rule refers to a post-vocalic
consonant, reference is being made to the skeletal level of representation;
second, because the change effected by the rule involves the feature of
continuancy, the rule also involves the melodic level of representation.
k � x Melodic level
I 30.
v c Skeletal Level
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
Schein & Steriade (36) has argued that it is such structural incompatibility
with the potential results of rule application that brings about the rule's failure
to apply. A rule referring only to the skeletal level (as in the Ojibwa case
above) can affect part of a geminate; a rule referring only to the melodic level
(as in the LuGanda case above) can affect the entire geminate; but a rule
referring to both skeletal and melodic levels is blocked unless all the require
ments for rule application are met by both halves of the geminate (a condition
not met in Tigrinya spirantization).
As a final note, consider the predictions of a linear theory with respect to
cases like spirantization. A geminate consonant might either be analyzed as a
sequence of simple consonants (Example 32a), or as a single consonant
bearing an ad hoc feature such as "[+ long]" (Example 32b):
32.
a k k a kk
Under the sequential analysis of Example 32a, a linear theory would in
correctly predict the application of spirantization, producing forms such as
NONLINEAR PHONOLOGY 219
*axkat for "kind of fruit". Under the feature-based analysis of Example 32b, a
linear theory makes no predictions whatsoever: The rule could stipulate that
application is restricted to "[-long]" consonants, or the rule could apply
freely to all consonants without taking into consideration their length.
Clearly the sequential analysis is inadequate since it derives incorrect
surface forms. It can be argued, moreover, that the feature-based analysis is
equally inadequate. It predicts that a cross-linguistic survey of geminates
should produce roughly equal numbers of cases where geminates do and do
not undergo rules such as spirantization. Both Hayes (16) and Schein &
Steriade (36) argue, however, that the true cross-linguistic generalization is
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
The nasal is labial before a labial consonant (Example 35a), alveolar before an
alveolar consonant (Example 35b), palatal before a palatal consonant (Ex
ample 35c), velar before a velar consonant (Example 35d), labial-velar before
a labial-velar consonant (Example 35e).
Within a linear theory, this means that the values of a set of distinctive
features must be affected by a single rule. Assuming the relevant features for
Yoruba to be [coronal] (articulated with the tip or blade of the tongue),
[anterior] (articulated at or in front of the alveolar ridge), and [labial] (involv
ing an articulation at the lips), a linearly formulated rule could be expressed as
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
follows:
[+nasal � �acoronalJ / - �-SYllabiCJ
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
l3antenor
81abial
acoronal
l3anterior
36.
81abial
This rule is to be read, "a nasal has the same values for [coronal], [anterior],
and [labial] as a following consonant," the Greek variables cr, f3 and 8
indicating dependencies between feature values.
As pointed out by K. P. Mohanan, and developed in numerous papers since
then (e.g. I, 8, 33), there are a variety of problems with such a linear
approach. Consider two examples.
The first problem is that of feature dependency. In a linear theory, the
formal notation employed to express dependencies between feature values is
that of variables with interdependent values, as just seen. This notation,
however, allows the expression of a wide number of unattested dependencies.
For example, although it is common to have the value of a feature [F] on one
segment be dependent on a value of the same feature [F] on another segment,
it is undesirable to allow a general mechanism whereby both + and - values
of [F] depend on the values of some second feature [G]. For example, a rule
identical to Example 36 but with revised feature dependencies as in Example
37 is impossible:
[+nasall �[acoro�aIJ
l3antenor
/ - [-SYllabic]
J3coronal 37.
81abial 8anterior
alabial
Although formally just as simple as Example 36, with the Yoruba data just
cited this rule would have bizarre effects like (a) deriving a bilabial nasal [m]
before an alveolar [d] (that is, a [+anterior, -coronal, +labial] nasal before a
[+anterior, + coronal, -labial] consonant), and (b) deriving a palatal nasal
before a doubly articulated labial-velar as well as otherwise unattested con-
NONLINEAR PHONOLOGY 221
sonants before labials and palatals. Even the rule's only plausible result (the
derivation of a velar nasal before a velar consonant) must presumably be
considered accidental.
Although this problem might be solved by some form of condition on
interfeature dependencies, an additional shortcoming of linear theories of
assimilation is the problem of natural classes. Rules such as Example 36
clearly demonstrate that a single rule can affect more than one feature. But if
this is the case, then what are the feature combinations capable of functioning
as sets in phonological rules? If one adopts for the sake of concreteness a set
of 20 features [note Chomsky & Halle (6)], then there would in principle be
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
features , and so on. But allowing such free combination of features vastly
overgenerates the types of sets actually attested in phonological processes.
Problems such as these receive straightforward solutions within an appro
priate nonlinear framework. Let us begin by assuming that features do not
constitute unordered sets, but that they are intrinsically structured so as to
reflect certain articulatory properties. Following work by Clements (7) and
Sagey (33), let us assume that the overall set of features (which is labelled the
ROOT, following a proposal by K. P. Mohanan) is divided into LARYNGEAL and
tubes through which air passes to produce speech, then one can think of the
laryngeal features as determining whether, among other things, the column of
air passing through the system is vibrating or not; the supralaryngeal features,
on the other hand, determine the shape and volume of the vocal tract, thereby
determining the quality of the resulting sound . Within the set of supralaryn
geal featurcs, a further subset is the class of PLACE features, with further
subsets under place consisting of the features related to individual articula
tors-for example, LABIAL for the movements of the lower lip, CORONAL for
the tip or blade of the tongue, and DORSAL for the body of the tongue. The set
structure so far described can be represented by a tree as follows:
51\ '
I ,
I ,
_ __ 0 _ _ _ --" · ..... laryngeal node
- �\-
- - - - - - o� - : -- -- - _ .. / : · . . . .. labial node
. ;
___ � _ ___ J · . .... coronal node
instance seen in Example 35, the place features of the relevant consonants can
extend their domain so as to characterize both the supralaryngeal nodes of the
consonants and the supralaryngeal nodes of a preceding nasal (illustrated here
with the partial trees appropriate for bilabial and alveolar places of articula
tion, and with assimilation indicated by an arrow):
m - b n - d
C-4: . . . . . . skeleton
I
o
,
o· ... . . . root node
laryngeal . . . . . .
\ \ . . . . . . supralaryngeal node 39.
�------�O."l\
node
- -
a.. e kase "surround"
-
a. . a kaha "hoe" vaa "weep, cry" 40.
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
- -
:l . . 0 n :lndo "drip"
- bh:l "summon"
:l . . :l y») "make proclamation"
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
As can be seen from this table. the sequences [a . . e] and [) . . 0] are possible
only with an intervening consonant other than [h]; the sequences [a. . a] and
[ ).. )] , on the other hand, are possible with adjacent vowels, or when an
intervening consonant is [h].
Putting aside the problem of the cases involving [h], the pattern in Example
40 is straightforwardly derived by positing a rule that assimilates a vowel to
an immediately preceding vowel. Assuming that the sequences [a.. a] and
[ J.. J] do not appear underlyingly in Tiv [as can indeed be argued (32)] , vaa
and yJJ are therefore derived by the application of assimilation to prior forms
Ivael and Iy Jol. The problem is how to account for the applicability of this
rule of vocalic assimilation to sequences of vowels with an intervening [h] .
To achieve this end, let us assume that the required rule is one that spreads
the supralaryngeal set of specifications, that is, a rule along the following
lines:
v v
I I 41.
V
root node
o supralaryngeal node
This rule extends the domain of the supralaryngeal specifications of the first
vowel to include the second vowel.
An account of the applicability of assimilation over [h] is now available.
An examination of the consonant inventory of Tiv reveals that [h) is the only
'''The vowels [a, e] (both produced without lip-rounding) cannot co-occur with [=>, 0] (both
produced with lip-rounding) because of a rule causing agreement in rounding in such forms (32).
The surface tonal patterns of verbal forms depend on how a verb is inflected. Since this is
irrelevant here, tone is not indicated below (but see 31).
224 PULLEYBLANK
o 0 0 root node
I 0I I
0
I
0
I I I
0 0 0 supraJaryngeaJ node
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
Assimilation is applicable in the first two cases but blocked in the third case
by the supralaryngeal specification of the intervening consonant:
a. a a b. a h a c. a s e
V V V C V V C V
I 0I
o 0
I oI 0I I aI 0I
a root node 43.
ll'
0
V
0
bi- I I
0 0 0 supraJaryngeaJ node
CONCLUSION
further constrain the operation of phonological rules. For example, the condi
tion on rule application discussed in the context of Tigrinya spirantization
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
ACKNOWLEDGMENTS
I thank Alan Bell and Larry Hyman for discussion of certain parts of this
paper.
Literature Cited
phonology. PhD thesis. Univ. Mass. , cation, the feature hierarchy and Tiv
Amherst vowels. Phonology 5:299-326
20. Ito, J. 1989. A prosodic theory of 33. Sagey, E. 1986. The representation of
epenthesis . Nat. Lang. Linguist. Theory features and relations in non-linear
7 : In press phonology. PhD thesis. MIT
2 1 . Kahn, D. 1980. Syllable-Based Gener 34. Schane, S. A. 1984. The fundamentals
alizations in English Phonology. New of particle phonology. Phonol. Yearb.
York: Garland 1 : 1 29-55
22. Kaye, J . , Lowenstamm, J . , Vergnaud, 35. Schein, B. 1 98 1 . Spirantization in Tig
J.-R. 1985. The internal structure of rinya. MIT Work. Pap. Linguist. 3:32-
phonological elements: a theory of 42
charm and government. Phonol. Yearb. 36. Schein, B . , Steriade, D. 1986. On
2:305-28 geminates. Linguist. Inq. 1 7:69 1-744
23. Kenstowicz, M. 1982. Gemination and 37. Steriade, D. 1982. Greek prosodies and
spirantization in Tigrinya. Stud. Lin the theory of syllabification. PhD thesis.
guist. Sci. (Univ. III . , Urbana) 1 2: 1 , MIT, Cambridge, MA
103-22 38. Steriade , D. 1 987. Locality conditions
24. Levin, J. 1 985. A metrical theory of and feature geometry. NELS 1 7:595-
syllabicity. PhD thesis. MIT 617
25. Libennan, M . , Prince, A. 1977. On 39. Zwicky, A . M . , Kaisse, E. M . , eds.
stress and linguistic rhythm. Linguist. 1987. Phonol. Yearb. 4 (whole volume)
Inq. 8:249-336
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
Access provided by 188.247.75.242 on 04/11/20. For personal use only.
Annu. Rev. Anthropol. 1989.18:203-226. Downloaded from www.annualreviews.org
Access provided by 188.247.75.242 on 04/11/20. For personal use only.