Silverman, Daniel. (1997). Phasing and recoverability. Outstanding
dissertations in linguistics series. New York: Garland Publishing. 242 pages.
Reviewed by Stefan Frisch, University of Michigan.
This book is a revised version of the author's 1995 UCLA dissertation. The
primary thesis defended by Silverman is that auditory salience plays an
important role in explaining the typology of phonological segment
inventories. The relative phasing of laryngeal and supralaryngeal
gestures is examined as a case study. Silverman demonstrates there is a
typological preference for phasing patterns in which the gestures are
optimally recoverable. Further, he proposes that sub-optimal patterns are
only found in inventories where the optimal patterns are also present.
(Note that Silverman does not crucially adopt the segment as a
phonological primitive. He is concerned primarily with gestures and their
realization in a system of syntagmatic contrasts. The terms 'segment' and
'segment inventory' are used only for expository convenience.)
Chapter 1. Introduction
This chapter introduces the primary thesis: Cross-linguistically,
laryngeal and supralaryngeal gestures are phased to make their values
maximally auditorily salient. Silverman argues that parallel production of
contrastive gestures is informationally optimal, but only if those
gestures are auditorily recoverable. In cases where parallel production
would make contrastive values unrecoverable, gestures are serially
sequenced. For example, in aspirated stops, laryngeal abduction is
sequences to follow the stop closure, resulting in broadband noise. If
laryngeal abduction and stop closure were simultaneous, the state of the
larynx would not be recoverable from the resulting acoustic signal
Silverman also introduces the 'gestural score' notation of Articulatory
Phonology (Browman & Goldstein, 1986) used in the description of the
gestural patterns. Each gestural score is accompanied by a set of
temporally aligned descriptions of acoustic cues which highlight the
importance of the recoverability of the gestures. The resulting segmental
percept is also given, to highlight whether all of the contrastive
segmental information has been effectively transmitted. Using this
notation, he exemplifies the four logically possible phasing patterns:
parallel, sequenced, expanded, or truncated. In the parallel phasing
pattern, two gestures are phased to be fully simultaneous. In the
sequenced pattern, two gestures are serially ordered. In the expanded
pattern, one gesture both begins before and ends after another. In the
truncated pattern, one gesture is phased to be simultaneous with a portion
(beginning or end) of another gesture.
Chapter 2. Previous work
In this chapter, Silverman reviews previous research on articulatory
timing, auditory response to acoustic signals, and the relevance of
auditory contrastiveness to segmental inventories. Two results are of
particular importance. First, a combination of auditory factors favor
patterns where low intensity signals are followed by high intensity
signals. Second, languages employ contrasts which are maximally auditorily
Chapter 3. Obstruents and laryngeal gestures
This chapter contains typological evidence to support Silverman's thesis.
Cross-linguistically, laryngeal gestures of abduction or constriction are
overwhelmingly phased to follow supralaryngeal constriction, which maximizes
the recoverability of both gestures. The sub-optimal pattern, where the
laryngeal gesture precedes the stop release is found only when the optimal
pattern is also present. For obstruents, which have a minimum of acoustic
energy to work with, these are the only two phasing patterns.
Chapter 4. Sonorants and laryngeal gestures
Sonorants have a greater amount of acoustic energy, and so laryngeal
gestures can overlap with supralaryngeal gestures. The most attested
pattern, for languages which do have a laryngeal contrast for sonorants,
is for the laryngeal gesture to be truncated to the beginning portion of
the supralaryngeal gesture. Parallel to the obstruent case, this phasing
pattern puts the low-energy breathy or glottalized portion of the sonorant
before the high-energy modally-voiced portion, maximizing auditory
salience. Again, the less optimal pattern truncates the laryngeal gesture
to the latter portion of the supralaryngeal gesture. Ordinarily, the
laryngeal and supralaryngeal gestures are not completely overlapped, as
contrastive supralaryngeal gestures (such as nasal place of articulation,
for example) would be rendered non- recoverable. However, an interesting
special case is found in laterals. Due to the formant structure of
laterals, languages generally do not have contrasts in their place of
articulation. Thus, in some cases (e.g. Zulu), contrastive laryngeal
gestures are realized fully parallel with the supralaryngeal gesture. A
similar pattern is found for coda nasals in Comaltapec Chinantec, where
the place of articulation is contextually determined. In contrast to onset
nasals, which have contrastive place of articulation, laryngeal abduction
is realized in parallel with the supralaryngeal gesture in coda nasals and
no contrasts are lost.
Chapter 5. Vowels and laryngeal gestures
In this chapter, vowels with contrastive laryngeal gestures are discussed.
Silverman claims that, since vowels have an abundance of acoustic energy,
laryngeal gestures can be implemented in parallel with supralaryngeal gestures
without loss of auditory contrast. However, this pattern is auditorily the
least optimal. Like the sonorant case, the optimal pattern is for the
laryngeal gesture to be truncated to the beginning portion of the vowel,
resulting in ?V or hV sequences. Less optimal is the opposite phasing pattern,
resulting in V? or Vh. Typologically hV is indeed much more prevalent than
voiceless vowels or Vh (and similarly for laryngeal constriction).
The bulk of the chapter (and the book) is devoted to 'laryngeally complex'
vowels in the Otomanguean languages. The term larygeally complex is used for
vowels which realize both contrastive phonation (breathy or creaky) and tone.
One example, Comaltapec Chinantec, has eight vowel qualities with five tonal
qualities and two voice qualities. In addition, this language has nasalization
and a length contrast which leads to 320 possible realizations of the nucleus.
Not surprisingly, words in this language are generally monosyllabic, and
nucleus quality is used to differentiate many of the lexical contrasts.
Silverman argues that simultaneous realization of breathiness/creakiness and
tone would render the tone contrasts unrecoverable. The cross-linguistically
prevalent pattern is for the laryngeal contrast to be truncated to the
beginning of the vowel, which is auditorily optimal. The tone is then
saliently realized during modal phonation in the latter portion of the vowel.
The opposite sequencing, with the laryngeal after the tonal contrast (and
modal phonation) is also attested. Silverman finds only two cases where tonal
and laryngeal contrasts are executed simultaneously (the Tibeto-Burman
languages Mpi and Tamang). In one of those cases, there are only two tones, so
tonal contrasts may not be in as much danger, as the tones would be more
distinct than in the case of Comaltapec Chinantec, with eight tones.
Overall, this is an inspirational volume demonstrating the importance of
auditory/phonetic explanation for phonological patterning. It is one of
the first of a series of UCLA dissertations on this topic, which together
address a wide range of phonological phenomena. Silverman's analysis
bridges the phonetics/phonology gap in a number of ways. For example, an
arbitrary number of phasing differences, which might be dismissed as
'phonetic implementation', are shown to be reducible to a small set that
can be phonologically contrastive. This reduction, via the more abstract
temporal relationships of simultaneity and precedence, nicely complements
work in speech perception on the cross-linguistic (and cross-species)
robustness of the categorical perception of voice onset time continua
(Kuhl & Miller, 1975; Pisoni, 1977).
To its credit, this book contains over a dozen reasonably detailed case
studies on the realization of laryngeal contrasts in different languages.
In many cases, recordings were available so the presence of the phasing
relationships were verified, and spectrograms of appropriate examples are
given. These case studies often address potential counterexamples to
Silverman's typological claims. For example, the Mon-Khmer language Chong
possesses coda stops with contrastive creakiness, but creakiness is
realized only in the non-optimal way as a pre-glottalized stop. In the
Chong case, however, other aspects of the morphophonology require the
non-optimal realization to avoid loss of contrast. In particular, coda
stops are obligatorily unreleased, and the language is non-suffixing. Due
to these additional constraints, post-glottalized phasing would not
saliently encode the larygneal contrast. This type of constraint
interaction is quite compatible with the general approach of Optimality
Theory (Prince & Smolensky, 1993), and in fact the original version of
this dissertation presented constraint tableau in such cases. The proposed
constraints are quite broad and were informally described, so the
Optimality Theoretic analysis did not add to the exposition, and the book
reads more easily without it.
Despite the lack of a formalist analysis, this book raises a number of
issues which are relevant to current formal concerns. In the Otomanguean
language Copala Trique there are different phasing relationships between
vocalic and laryngeal gestures. The laryngeal gestures (breathiness and
creakiness) can be truncated to the first portion of the vowel, the second
portion of the vowel, or can 'interrupt' the vowel, appearing in the
middle. These three locations for a laryngeal gesture support lexical
contrasts, and there is clear evidence that the interrupted vowels are
monosyllabic. Encoding these three configurations using more abstract
representations than Articulatory Phonology is no trivial task. In
addition, the correspondence theory approach to faithfulness considers
segmentally aligned and ordered input and output (McCarthy & Prince,
1995). The presence of relatively small but contrastive differences in
phasing require subsegmental correspondence relations between the input
and output. Another missing aspect of an Optimality Theoretic analysis
would be a factorial typology of constraint interactions. While
Silverman's coverage is quite extensive, I would be interested in seeing a
discussion of the pros and cons for each logical possibility in phasing
between laryngeal and supralaryngeal gestures. Such a discussion will
eventually be needed in order to determine whether the Silverman's
proposal is is truly predictive, or just informally defined to the extent
that any observed pattern could be explained.
In a few instances, Silverman makes use of the UPSID database of segmental
inventories (Maddieson, 1986) to demonstrate that the typological
predictions are satisfied. Unfortunately, quantitative differences are not
reported in most cases. While sufficient data may not be available, a
quantitative analysis is desirable in order to address a current question
in work on phonetic explanation for phonological patterns: Are
non-optimal patterns avoided to the degree that they are non-optimal? In
other words, is the hypothesized functional force of auditory
recoverability transparently reflected in the patterns within and across
languages, or is it 'phonologized' in some way by the language learner
such that there is no quantitative relationship? The answer to this
question has implications for the architecture of the phonetically
grounded grammar, as quantitative constraints or quantitative constraint
rankings would be required.
This book touches on a number of other very important issues that should
be topics of ongoing research. The central role of the recoverability of
contrasts leads inevitably to the question: What is a contrast? Assuming
the psychological reality of the segment provides a simple answer, but one
which is not entirely correct. The three contrastive phasing relationships
in Copala Trique are not amenable to a segmental analysis. In Chong, a
combination of coda unrelease and lack of suffixation is claimed to lead
to pre-glottalized stops. This case is contrasted with Korean, where there
is some suffixation, and post-glottalized stops are maintained in that
environment but neutralized elsewhere. I wonder how frequent the
neutralizing environment must be before a non-optimal pattern becomes
necessary or the contrast is lost? Why isn't the laryngeal gesture phased
before the stop closure in the neutralizing environment, but after in the
non-neutralizing environment? Perhaps some other constraint is involved
In many of the case studies given, the simple syllable structure and
monosyllabic tendencies of the languages are mentioned as reasons why
complex and non-optimal phasing relations are found at all (in most cases,
the optimal recoverability of supralaryngeal gestures is found if
contrastive laryngeal gestures are not used at all). This suggests that
there is some minimal set of contrasts required to create a sufficient
number of open class items for a language to be a useful communication
system, and that differences in syllable structure, word length, and
segmental inventory size interact in some fashion to this end. The answers
to these deeper questions appear to be closer to our understanding when
the phonological system is conceptualized as implied in this book, as a
combinatorial system of articulatory/acoustic contrasts highly constrained
by a variety of functional factors.
Browman, C. P. & Goldstein, L. (1986). Towards an articulatory phonology.
Phonology yearbook 3: 219-252.
Kuhl, P. K. & Miller, J. D. (1975). Speech perception by the chinchilla:
Voiced-voiceless distinction in alveolar plosive consonants. Science 190:
McCarthy, J. J. & Prince, A. (1995). Faithfulness and reduplicative
identity. Papers in optimality theory. University of Massachusetts
occasional papers 18. Amherst, MA: GLSA. pp. 249-384.
Pisoni, D. B. (1977). Identification and discrimination of the relative
onset time of two component tones: Implications for voicing perception in
stops. Journal of the Acoustical Society of America 61: 1352-1361.
Prince, A. & Smolensky, P. (1993). Optimality theory: Constraint
interaction in generative phonology. Rutgers University Center for
Cognitive Science technical reports 2. New Brunswick, NJ: RUCCS.
Stefan Frisch, Language Learning Visiting Research Assistant Professor,
Program in Linguistics, University of Michigan. Ph.D., Northwestern
University, 1996. Research interests include phonetics, phonology,
psycholinguistics, and computational linguistics.
Program in Linguistics
University of Michigan
1076 Frieze Building
105 S. State St.
Ann Arbor, MI 48109-1285