Review of Postvelar Harmony
|
|
|
|
|
Review:
|
Date: Fri, 04 Jul 2003 20:43:30 -0700
From: Alexei Kochetov <akocheto@sfu.ca>
Subject: Postvelar Harmony
Shahin, Kimary N. (2002) Postvelar Harmony, John Benjamins,
Publishing Company, Current Issues in Linguistic Theory 225.
Alexei Kochetov, Simon Fraser University
SYNOPSIS
Chapter 1. Introduction
'Postvelar Harmony' is a revised version of Kimary Shahin's
dissertation defended at the University of British Columbia
in 1997. The goal of this book is to demonstrate that two
unrelated languages, Palestinian Arabic (PA) and St'at'imcets
Salish (SS), exhibit the same two types of phonological
processes. These processes, jointly referred to as 'postvelar
harmony', are 'pharyngealization' and 'uvularization'. While the
first process is triggered by 'guttural' and 'emphatic'
segments, the second process is triggered by 'emphatics' only.
The book also aims to provide a formal Optimality Theoretic
account of these processes, and to demonstrate that the
similarities and differences between PA and SS with respect to
post-velar harmony are due to different rankings of the same
general constraints.
Some clarifications of the author's terminology need to be made
from the outset. The term 'harmony' is not explicitly defined in
the book. Although Shahin states that '''harmony' refers to
assimilation and dissimilation'' (p. 40), the term is applied
throughout the book to cases of local or non-local assimilation
(mainly V-to-C and C-to-C). The 'guttural' segments include
pharyngeal and uvular approximants, as well as laryngeals (in PA
only; cf. McCarthy 1994 for Standard Arabic). The class of
'emphatics' consists of coronals and labials with secondary
uvular/pharyngeal articulation, as well as uvular stops and
fricatives. The classes of gutturals and emphatics are jointly
referred to as 'post-velars', that is segments fully or
partially articulated in the post-velar region of the vocal
tract (see Critical Evaluation).
Adopting Articulator Theory within Feature Geometry (Sagey 1986,
among others), Shahin views phonological representations as
privative distinctive features organized hierarchically and
defined in articulatory terms. The features deemed responsible
for postvelar harmony processes are the primary [Retracted
Tongue Root] ([RTR]) feature of gutturals, as well as the
secondary [RTR] and [Dorsal] features of emphatics. Shahin
adopts a standard version of Optimality Theory (Prince &
Smolensky 1993) and employs featural alignment and featural
correspondence constraints to capture harmony processes. In
addition, Shahin distinguishes between markedness constraints
that determine featural relations within a segment
('paradigmatic grounded constraints') and between segments
('syntagmatic grounded constraints'). Both types are assumed to
be ''rooted'' in the physics of speech production and the
acoustic signal. Shahin advocates a clear-cut distinction
between phonetics and phonology and lists criteria that, in her
view, distinguish phonological and phonetic phenomena. She
acknowledges, however, that some phonetic properties are
language-specific, and thus must be considered ''cognitive'' (p.
49).
This chapter also provides a detailed review of previous
articulatory X-ray studies of Arabic gutturals and emphatics; it
also makes predictions about acoustic effects of postvelar
articulations.
Chapters 2 and 3
These two chapters constitute the core of the book, presenting
investigations of postvelar harmony in PA and SS. They are
organized very similarly: for each language, Shahin introduces
segment inventories, highlights relevant phonological issues,
and presents a phonological account of the phenomena. In
addition, she provides results of acoustic analyses and, in some
cases, of perceptual experiments; these are expected to provide
additional support for the phonological analysis.
Chapter 2. Postvelar harmony in Palestinian Arabic
The focus of this chapter is the Abu Shusha dialect of PA. Most
of the data were collected by Shahin from 26 speakers during her
fieldwork in Ramallah, West Bank. The acoustic analysis is based
on recordings from two speakers.
Shahin analyzes the consonant inventory of PA as having 6
gutturals and 8 emphatics. Pharyngealized vowels are assumed to
be derived rather than part of the underlying vowel inventory.
According to Shahin, 'pharyngealization harmony' in PA involves
two processes: Short vowels have lowered and backed allophones
under adjacency to a postvelar consonant (including laryngeals).
In addition, short vowels can be pharyngealized in closed
syllables; these vowels are assumed to trigger 'non-local
harmony' within a phonological word. Shahin's OT account of this
phenomenon makes use of alignment constraints spreading [RTR] to
an adjacent nucleus and non-locally within a phonological word.
The opacity of long and stem-final vowels is captured by highly
ranked markedness constraints ensuring the incompatibility of
[RTR] with bimoraic and stem-final nuclei. The second process,
'uvularization harmony', is a long distance bi-directional
assimilation triggered by emphatic consonants (known as
'emphasis spread'). Shahin analyzes the process as being
triggered by a constraint aligning both secondary [Dorsal] and
secondary [RTR] with the left and right edges of a prosodic
word. Post-alveolar obstruents are opaque to the process, while
non-low vowels are phonologically transparent but show some
gradient phonetic effects. These opacity and transparency
effects are analyzed as consequences of the high ranking of
'grounded' constraints rooted in the articulatory
incompatibility of the fronting and raising of the tongue with
its retraction.
Chapter 3. Postvelar harmony in St'at'imcets Salish
Chapter 3 focuses on SS, an Interior Salish language spoken in
British Columbia, Canada. The word corpus was collected by the
author from 6 speakers during her fieldwork in Vancouver. The
acoustic analysis was based on the speech of 2 speakers.
The consonant inventory of SS is analyzed by Shahin as having 4
gutturals (uvular approximants) and 12 emphatics ('retracted'
coronals and uvular obstruents). Shahin's analysis differs from
some previous work (van Eijk 1997) in assuming no underlying
pharyngealized vowels and attributing unconditioned
pharyngealization of vowels in some stems to a ''floating
emphasis feature'' (p. 199). Shahin describes 'pharyngealization
harmony' in SS as a local assimilation of vowels to the
following postvelar consonant, resulting in pharyngealized
allophones. In the context of emphatics, the comparable effect
appears to hold for high vowels only. The phonological analysis
of the pharyngealization harmony involves a quite simple
interaction between the alignment constraint spreading [RTR] to
the leftward nucleus, and featural correspondence constraints.
Unlike the previous process, 'uvularization harmony' in SS
apparently involves backing of an epenthetic vowel or an
underlying low vowel before an adjacent emphatic consonant. The
process is analyzed as applying to consonants as well, however
the examples given by SS are limited to the post-alveolar
affricate in two lexical items. Shahin analyzes uvularization
harmony in SS as being driven by an alignment constraint
spreading secondary [Dorsal] and [RTR] to the leftward segment.
The neutral status of high vowels is accounted for by reference
to highly ranked 'grounded' constraints against this featural
combination.
Chapter 4. Conclusion
Chapter 4 provides a summary of all constraints used in the
analysis, together with their grounding. It also presents the
relevant fragments of OT grammars of PA and SS. In this chapter
Shahin outlines a new and exciting direction for further
research: a comparison of pharyngealization harmonies with
consonantal and vocalic sources.
Appendices and Index
The book contains a number of appendices: abbreviations used in
the book, PA and SS forms used in the acoustic analysis, Salish
language classification, and a SS word list. The book also
includes a name index, subject index, and a language index.
CRITICAL EVALUATION
General remarks
Shahin's 'Postvelar Harmony' is a solid work, impressive both in
its breadth and in its attention to detail. It is one of only a
very few recently published dissertations on theoretical
phonology based on the author's original fieldwork data.
Moreover, these data come from two unrelated languages,
uncovering fascinating phonological and phonetic similarities
between the languages and leading to some interesting
typological generalizations. Further, the author goes beyond the
traditional format of phonological analysis by seeking support
for her claims in acoustics and perception. The general
organization of the book is clear and straightforward. The
presentation of the data and analysis for the two languages
follows the same format and is easy to compare. The book
benefits from well-organized language data appendices and
multiple figures illustrating findings of previous articulatory
studies, the current acoustic results, and the summary of the
phonological analyses. The language index contains more than
fifty entries.
I have to admit, however, that as a phonologist lacking intimate
familiarity with issues of Arabic and Salish linguistics I
didn't find 'Postvelar Harmony' an easy read. At times language
data and language-specific theoretical assumptions were
presented somewhat abruptly, without sufficient introduction,
prompting me to consult additional sources on the languages. The
language-particular interpretation of the terms 'guttural' and
'emphatic' was confusing, though perhaps inevitable given
phonological and phonetic differences between the two languages.
Specifically, Shahin's class of PA gutturals includes uvular and
pharyngeal approximants and laryngeals /? h/, while gutturals in
SS include only uvular approximants. The 'dorsal emphatics' in
SS are uvular stops and fricatives, while in PA this class is
limited to the velar stop (which is still ambiguously referred
to as 'post-velar') or to a uvularized/pharyngealized glottal
stop (for one of the PA speakers).
Further, some general theoretical assumptions introduced in
Chapter 1 could have benefited from more discussion. For
instance, Shahin mentions at the beginning of the book that the
typology of harmony systems is ''essential for a clear
understanding of ... postvelar harmony'' (p. 3). Yet, this
typology is outlined in only two short paragraphs in section
1.5, leaving the reader wondering whether it is relevant for the
analysis at all. The term 'phonological visibility' is used a
number of times throughout the book, yet it is not defined or
illustrated. Similarly, discussions of the acoustic results and
of the results of the phonological analyses are often overly
brief and sketchy. The last chapter (Conclusion), intended to
provide a summary of the cross-linguistic analysis of postvelar
harmony, is only four pages long.
Acoustic and perceptual evidence
Reference to the results of the author's acoustic analysis plays
a major role in the proposed phonological account. Thus,
acoustic measurements of formant frequencies of vowel tokens in
the environment of postvelars are compared to predicted ranges
of values and taken as acoustic support for the assumed
articulation of these consonants. This is then interpreted as
support for their phonological feature specification and for the
distinction between the two types of postvelar harmony.
This approach, in my view, is potentially problematic. Formant
values at the midpoint of a given vowel (the main focus of
Shahin's acoustic analysis) indicate whether the vowel is
articulated with or without tongue body/tongue root retraction,
in other words, whether it belongs to one of the two presumably
distinct allophonic categories. However, these values do not
necessarily indicate whether an adjacent consonant is
articulated with or without retraction. It is even less clear
how the formant values for the vowel can constitute evidence for
the PHONOLOGICAL FEATURE [RTR] of the following consonant and
for the SPREADING of this feature to the vowel. This
interpretation of acoustic values appears to be forced by a
rather literal phonetic interpretation of phonological
representations, as well as by a view of harmony as a dynamic
process of concrete feature spreading.
Further, the claimed acoustic support for tongue-root retracted
articulation of PA laryngeals (p. 96) does not seem to be
conclusive, since the analysis of these segments is based on
midpoint values of adjacent pharyngealized vowels (apparently in
three lexical items). No formant measurements in phonologically
or lexically neutral environments are reported (e.g., VC/CV
transitions at word boundaries or in nonsense words; see Yeou
1997). This weakens Shahin's argument against a more abstract
feature [Pharyngeal] (McCarthy 1994). In addition, the proposed
distinction between 'pharyngealization' and 'uvularization'
harmonies in SS hinges crucially on the apparently different
realizations of vowels in two contexts. The number of tokens for
the 'uvularized' allophones, however, is rather small and these
do not seem to differ substantially from the tokens of the
'pharyngealized' allophones (see, e.g., F1-F2 plots on p. 257;
two tokens per category). No statistical results are provided to
support the distinction between the two categories, and the
evidence for two distinct postvelar harmonies in SS is thus not
entirely convincing.
Although Shahin mentions that ''gradient phonetic properties
cannot be ignored unless ruled out as speech-phonetic'' (p. 48),
the acoustic analysis is limited to the phenomena that are
somehow pre-determined to be 'phonological'. Yet, the many
gradient effects of Arabic 'emphasis spread' (Zawaydeh 1999,
Watson 2002: 268-86) are just as intriguing as its categorical
effects, and also require an explanation. Moreover, the border
between the phonetic and the phonological aspects of this
process appears to be rather fuzzy. For example, while lip
rounding can be considered as a phonetic effect accompanying
emphatic articulation in Cairene Arabic, it acts like a full-
fledged phonological feature in San'ani Arabic (Watson 2002).
The same criticism applies to the analysis of SS. The process of
lowering and diphthongization of high vowels after uvulars is
dismissed in the book as a phonetic ''effect from the adjacent
postvelar'' (p. 209-211); however, no acoustic evidence is
provided for this conclusion. In any case, the phenomenon hardly
meets Shahin's definition of 'speech-phonetic' effects as being
''purely physical''.
In Chapters 2 and 3, Shahin introduces three small-scale
perceptual experiments. The rather brief presentation leaves
some questions about the methodology used in these experiments
and the direct relevance of their results to the general
analysis. For instance, the finding that three Arabic listeners
labeled a few tokens of SS retracted consonants as emphatic is
interpreted as support for the emphatic status of these segments
(pp. 188-90). The results, however, can be more plausibly
interpreted as arising from the Arabic listeners' selection of
the closest native categories -- the view rejected by Shahin. In
the same way for example, for a Russian listener (the author of
this review), Arabic non-emphatic and emphatic consonants in
some vowel contexts sound quite similar to Russian palatalized
and plain (velarized) categories; however, this cannot be
reasonably considered evidence for a palatalized-plain
distinction in Arabic.
It should be added that these shortcomings in the phonetic
analyses of 'Postvelar Harmony' are likely inevitable
consequences of the thematic breadth of the book and of
limitations imposed by the available corpus data. This book
still remains an appealing example of how phonetic predictions
and results based on original data can be tightly incorporated
into a theoretical phonological account.
Phonological representations, constraints, and opacity
Shahin's theoretical account of the facts of postvelar harmony
combines the Feature Geometry (FG) representational assumptions
with the constraint-based architecture of Optimality Theoretic
(OT) grammar. Most of the criticism presented below applies not
so much to Shahin's analysis per se but to the general
theoretical assumptions underlying FG and to the traditional
view of lexical representations.
The consonant and vowel representations Shahin employs are well-
equipped to account for postvelar harmony in both PA and SS, and
the details of the analysis are worked out in remarkable detail.
However, as in many FG accounts, representations posited to deal
with a subset of language data cannot necessarily be extended to
a wider range of phonological phenomena. Thus, Shahin does not
address the question of feature class co-occurrence restrictions
in PA and SS roots ('OCP effects'). These effects, however,
present a serious problem for the proposed FG representations.
Formulating OCP restrictions as constraints on combinations of
the features posited for postevlars either under-predicts or
over-predicts the effects observed for Arabic (see McCarthy
1994). And, as with any set of FG categorical representations,
Shahin's representations are ill-suited to deal with gradient
OCP effects (see Frisch, Broe, & Pierrehumbert 1997). Like
Arabic, SS has been described as showing consonant co-occurrence
restrictions in roots (van Eijk 1997: 8-9). Yet, these do not
refer to the classes of gutturals and emphatics as such; in
fact, all uvulars pattern as a natural class distinct from
'retracted' coronals and all other consonants. In sum, we are
left with having to either posit a number of process-specific FG
representations or give up the idea of rigid feature
constituency in favor of violable feature classes (see Padgett
2001). In addition, it is not clear what exactly FG
representations add to the OT analysis of postvelar harmony,
since all relevant aspects of the processes seem to be
successfully captured by interacting constraints referring
directly to features.
The problem with FG representations (as they are traditionally
conceived) is not only their lack of flexibility and
insufficient 'abstractness' in dealing with higher-level
phonological phenomena, but also their arbitrariness and
insufficient 'concreteness' in dealing with the lower-level of
speech production. Despite substantial attention to the
distinction between phonology and phonetics, the book remains
silent on how symbolic FG representations are 'implemented'
phonetically. It is not clear, for instance, how otherwise
identical features [RTR] of pharyngeals and emphatics are mapped
onto two different articulatory gestures (see Ladefoged &
Maddieson 1996: 365-66). Moreover, it has been noted that the
degree of uvularization in Arabic differs between different
consonants in a given dialect as well as between different
dialects -- it is often accompanied by varying degrees of
labialization, among other effects (see Watson 2002: 279-80 and
references therein). Yet, in Shahin's FG account all this
variation corresponds to as a symbolic constituent 'secondary
[Dorsal] and secondary [RTR]'. In sum, the facts of speech
production suggest that speakers' low-level phonetic knowledge
is quite detailed and variable; this knowledge must be part of
any model of the language faculty, regardless of our assumptions
about higher-level phonological constructs. FG representations
neither encode this low-level knowledge, nor provide a feasible
mapping procedure. A possible alternative is to adopt low-level
gestural representations that encode both categorical and
gradient knowledge (Browman & Goldstein 1989).
Related to this question is the issue of the 'directionality'
and 'strength' of harmony processes in different dialects of
Arabic. While OT alignment constraints are well-suited to handle
the details of leftward and rightward local and long-distance
assimilation, they (as well as FG representations) have little
or nothing to say about why leftward spread is more common and
why some consonants are better triggers of the process than
others (Zawaydeh 1999). These facts, however, are explained, at
least in part, by the details of the gestural organization of
secondary uvular articulation, namely its substantial extent in
time and its phasing at the onset rather than at offset of the
primary gesture (see Watson 2002: 284). Interestingly, the
common gestural phasing patterns of secondary articulations
provide insight into their preferred direction of V-to-C
assimilation: while uvularization tends to be leftward,
labialization and palatalization are commonly rightward (see
Ladefoged & Maddieson 1996: 357, Watson 2002: 281-84 on
labialization; see Kochetov 2002 on palatalization).
Finally, the issue of lexical representations and opacity
deserves special attention. Shahin takes the traditional view
that the unit of lexical storage is the 'underlying
representation' stripped of all predictable information. This
creates certain problems for her analysis. For example, she
posits /tibn/ as the underlying form for the PA word {tIbIn}
'straw' based on the following analysis: (1) the second vowel of
the word is analyzed as predictably epenthetic; (2) its 'color'
is spread from the preceding vowel and (3) its pharyngealized
quality is due to 'closed-syllable pharyngealization'; (4) the
pharyngealized quality of the first vowel is due to the
spreading of [RTR] from the second vowel. This approach
obviously favors the expression of underlying-to-surface mapping
in derivational terms with reference to intermediate
representations. This mapping, however, does not fit well with
the surface-oriented nature of standard OT and its reliance on
'lexicon optimization' (not to mention learnability and
processing considerations). The constraint Align([RTR], Left,
Word, Left), which is assumed to spread the feature leftward
within a word, has to refer to the intermediate representation
{tibIn} (not indicated in Tableau 86), rather than to the non-
RTR underlying form /tibn/. Shahin's analysis, however, does not
seem to assume intermediate phonological representations.
Another potentially problematic issue for the proposed analysis
is morphologically- or lexically-conditioned blocking of
postvelar harmony, in particular, the variable behavior of /jj/
in PA (p. 153), opaque affixes in Arabic (Watson 2002: 273-82),
and variable 'retracted' suffixes in SS (van Eijk 1997: 29-30).
All these facts, however, can be straightforwardly incorporated
into the account if we assume an alternative view in which a
unit of lexical storage is a 'surface' form specified for both
unpredictable and predictable information (see Bybee 2001,
Pierrehumbert 2001 on usage-based models of the lexicon). In
this view, for instance, relations between alternants or 'root-
and-pattern morphology' forms can be expressed as a learner's
'bottom-up' generalizations over the networks of stored surface
forms. This view is also compatible with some recent work in OT
that reanalyzes cases of phonological opacity as transparent
allomorphy relations (e.g., Rubach & Booij 2001, among others).
By dispensing with the 'top-down' derivational assumptions, this
alternative also calls into question the plausibility of
conceptualizing harmony as a 'dynamic' feature-spreading
process; instead it views harmony as 'static' feature co-
occurrence and allomorphy -- emergent properties of the ever-
evolving lexicon.
In conclusion, the few inconsistencies I identified in the
analysis of 'Postvelar Harmony' are by no means flaws in the
particular book or in the author's theoretical approach per se.
Rather, they seem to reflect the general state of flux in
current phonological theory, with new ideas and concepts still
largely co-existing with decades-old misconceptions and
assumptions that have now been challenged. This work highlights
a range of crucial issues in current phonological theory; it
also provides us with exciting new language data and fresh
insights into the mechanism of postvelar harmony and the
typology of assimilation in general. I would recommend this book
to general phonologists and phoneticians, as well as to anyone
interested in Arabic and Salish linguistics.
REFERENCES
Browman, C.P. and Goldstein, L. (1989) Articulatory gestures as
phonological units. Phonology 6. 201-252.
Bybee, J.. 2001. Phonology of Language Use. Cambridge Studies in
Linguistics 94. Cambridge, UK: Cambridge University Press.
Frisch S., M. Broe, and J. Pierrehumbert. 1997. Similarity and
phonotactics in Arabic. Bloomington, IN and Evanston IL: Indiana
University and Northwestern University, ms. ROA-223, Rutgers
University Archive, http://Rutgers.edu/roa.html.
Kochetov A. (2002) Production, Perception, and emergent
Phonotactic Patterns: A case of contrastive palatalization. New
York: Routledge.
Ladefoged, P. and Maddieson, I. (1996) The sounds of the world's
languages. Cambridge, MA: Blackwell.
McCarthy, J. 1994. The phonetics and phonology of Semitic
pharyngeals. In: Papers in Laboratory Phonology III:
Phonological structure and phonetic form, P. Keating, ed.,
Cambridge University Press, Cambridge. 191-233.
Padgett, J. 2002. Feature classes in phonology. Language 78.1.
81-110.
Pierrehumbert, J. (2001) Exemplar dynamics: Word frequency,
lenition, and contrast. In: Frequency effects and Emergent
Grammar, J. Bybee and P. Hopper, eds. Amsterdam: John Benjamins,
137-157.
Prince, A., and P. Smolensky 1993. Optimality Theory: Constraint
interaction in generative grammar. Ms., Rutgers University.
Rubach J. and G. Booij. 2001. Allomorphy in Optimality Theory:
Polish iotation. Language 77.1. 26-60.
Sagey, E. 1986. The representations of features and relations in
non-linear phonology. Unpublished Ph.D. dissertation, MIT.
van Eijk, J. 1997. The Lillooet Language: Phonology, morphology,
syntax. Vancouver: University of British Columbia Press.
Watson, J. C. E. 2002. The Phonology and Morphology of Arabic.
Oxford University Press.
Yeou, M. 1997. Locus equations and the degree of coarticulation
of Arabic consonants. Phonetica, 54. 187-202.
Zawaydeh, B. A. 1999. The phonetics and phonology of gutturals
in Arabic. Unpublished Ph.D. thesis, Indiana University.
|
| |
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Alexei Kochetov is Assistant Professor of Linguistics at Simon
Fraser University. He received his Ph.D. in 2001 from the
University of Toronto. Later he was a postdoctoral fellow at
Haskins Laboratories. His research interests include
phonological theory, phonetics-phonology interactions,
markedness, and learnability. He is also interested in agent-
based modeling of emergent phonological phenomena.
|
|
|
|
|
|
