Hardcastle, William J., & Nigel Hewlett, eds. (1999) Coarticulation:
Theory, Data and Techniques, Cambridge University Press, 386 pages.
Marija Tabain, Institut de la Communication Parlee, Institut National
Polytechnique de Grenoble
Coarticulation is an all-pervasive and most important aspect of speech
production. Although as linguists and, more basically, as literate
speakers of a human language, we may think of speech as being made up of
discrete segments (consonants and vowels), the reality is that these
"segments" overlap in time and in space to such an extent that it is
often difficult to determine at what point one segment begins and
another segment ends. This overlap in time and space is called
"coarticulation", and has been one of the main focuses of speech
research since the 1950s, when the invention of the sound spectrograph
showed that, acoustically, there were no clear boundaries between the
various consonants and vowels of the speech signal.
An example may help to elucidate what coarticulation is, and what effect
it may have on the speech signal. The words "see" and "Sue" provide a
good example. In "see", the vowel is unrounded, whereas in "Sue" the
vowel is rounded. During the production of "s", the lips are rounded for
"Sue" but unrounded for "see". The lips are thus anticipating the
following vowel, since part of the spatial definition for the vowel
(i.e. lip-rounding) has infiltrated the temporal domain of the
consonant. Moreover, the effects of this lip-rounding during the
fricative can be heard by the listener. During the "s" in "see", the
frequency of the main resonance is quite high. During the "s" in "Sue",
however, the frequency of the main resonance is lowered, because the
presence of lip-rounding has the effect of lengthening the resonating
cavity. Indeed, perceptual tests show that listeners are capable of
predicting the following vowel based on the initial "s" alone.
Coarticulation, therefore, can be advantageous both for the speaker (in
that it allows him or her to produce [aspects of] more than one speech
sound at a time, thereby speeding up the speech process), as well as for
the listener, who is given more time to "decode" the signal (in the
example above, the listener has both the vowel itself and the fricative
noise to help determine the vowel identity, while aspects of the "s"
also persevere into the vowel). Other manifestations of coarticulation
are, of course, less obvious both to the speaker and the hearer, and can
only be discovered using instrumental techniques of investigation.
As pointed out on the first page of the introduction to "Coarticulation:
data, theory and techniques" (edited by William Hardcastle and Nigel
Hewlett), as well as on the first page of the first chapter (by Kuhnert
and Noland), a study of coarticulation is extremely useful in
determining which aspects of speech are universal, and which are
language-specific. For instance, the so-called "allophonic" variation in
velar consonant production (where realizations in front and back
vowel-contexts are described as separate allophones) can best be
described as coarticulation. Simply put, the tongue body/back is a very
slow articulator, involved in both vowel and velar consonant production.
When a vowel and a velar consonant need to be produced in sequence, they
overlap maximally in time and space: the constriction for the velar
consonant is located in the region where the vowel is to be produced.
This has the same advantages in production for the speaker as the above
"see Sue" example does, and similar acoustic consequences for the
hearer. To date, all languages investigated have shown this sort of
behaviour for velar consonants. It would therefore be misleading to say
that this is an allophonic phenomenon which needs to be specified in the
phonology of a language. For this reason, a basic understanding of
coarticulation is crucial to any phonologist who believes that phonology
should be able to account for real speech data. When one further takes
into account that coarticulation is affected by such linguistic and
sociolinguistic variables as stress, the presence of various prosodic
boundaries, rate (fast, slow etc.) and speaking style (clear, casual
etc.), the intimate relationship between coarticulation, phonetics,
phonology and sociolinguistics becomes all the clearer.
THE BOOK
"Coarticulation: data, theory and techniques", edited by William
Hardcastle and Nigel Hewlett, presents a concise outline of the current
state-of-the-art in speech research on coarticulation. As such, it
should be of interest to phoneticians, phonologists, psychologists,
computer scientists and speech engineers. The book contains 16 chapters
plus an introduction by the editors. Each chapter is written by a
leading researcher in the field, and provides a good summary of research
into one particular aspect of coarticulation. For those involved in
speech research, it provides a handy reference, while for others with an
interest in speech but not necessarily in coarticulation, it should
provide enough information for a good introduction to the subject.
In some ways, this book is a logical progression from two previous books
co-edited by Hardcastle. The first is "Speech Production and Speech
Modelling" (co-edited with Alain Marchal, and published in 1990
following a NATO-sponsored workshop in Bonas, France), in which two of
the four sections were dedicated to aspects of coarticulation (Section 2
"Coarticulation and other connected speech processes"; and Section 4
"Theories and models of articulatory organization and timing"). Many new
results on coarticulation were presented here, and certain chapters,
such as those by Farnetani and by Lindblom, are still referred to today.
The second book is "The Handbook of the Phonetic Sciences (co-edited
with John Laver, and part of the Blackwell "Handbooks in Linguistics"
series), in which only one chapter (that by Farnetani) is ostensibly
dedicated to coarticulation, but in which many of the chapters deal with
coarticulation as an inescapable phenomenon of speech production. The
current book is, therefore, a sort of "handbook" of coarticulation,
where coarticulation is given the space it deserves to be treated in its
own right. Indeed, several authors in the current book were contributors
to the previous book. Farnetani is a contributor to all three books, and
Stone, Ni Chasaide, Gobl and Nolan are contributors to the 1997 and
current books. This latest book arose out of the ACCOR project
(Articulatory-acoustic correlations in coarticulatory processes: a
cross-language investigation") funded by the European Union under the
ESPRIT framework, and many of the authors were involved in this project.
However, not all of the authors are currently working in Europe (Manuel
and Beckman are in the US, and Fletcher and Harrington are in
Australia). Although the more theoretical aspects typical of American
research into coarticulation are not emphasized in this book (with the
exception of chapter 2 which provides an excellent summary of the
various theories that have been advanced to account for coarticulation -
see below for more detail), the chapters do provide a very good overview
of the empirical results from research into the articulatory strategies
involved in producing sequences of sounds.
The book is organized into 4 sections. The first section deals with
theories and models of coarticulation, while the third section deals
with broader linguistic theories and how coarticulation can be made to
fit in to linguistic models in general. The chapter by Farnetani and
Recasens (chapter 2) is perhaps my favourite chapter in the book. It
outlines clearly and concisely the main theories used to describe and
explain coarticulation, with an expose and diagrams followed by a
critical evaluation. For researchers who are confused about the
differences and similarities between theories of coarticulation, this
chapter provides an excellent reference. It is clearly set-out and
extremely well-referenced (a comment which can be made about many
chapters in the book). The theories include Lindblom's theory of Hyper-&
Hypo-articulation (Lindblom 1990), and related theories of adaptive
variability; Ohman's vowel-to-vowel model (a fascinating model presented
in 1966, which suggests that speech is a sequence of continuous vowel
gestures, and that consonants are merely superimposed on this stream of
vowels); models of coarticulation within generative and featural
phonology (to which many pages of criticism are dedicated); Keating's
Window model (Keating 1988); and the Task Dynamic model of speech
production (Saltzman and Munhall 1989) which is related to the
Articulatory Phonology approach of Browman and Goldstein (1992). There
is also a welcome summary of the debate regarding look-ahead,
coproduction and hybrid models of coarticulation. This debate concerns
the timing of the onsets of speech gestures, although it is not clear
that the debate (which is still not clearly resolved) still motivates
research into coarticulation today.
The chapter by Kuhnert and Nolan (chapter 1) provides an interesting
(pre-)history of the study of coarticulation in the late 19th and early
20th centuries. The authors also provide a summary of what is known
about coarticulation as a part of speech acquisition. To date, it
appears that results are contradictory, and one can perhaps assume that
not much more progress will be made until less invasive techniques are
devised for studying coarticulation (most techniques which investigate
speech production cause some discomfort to the speaker, and may even
require medical supervision). When these results are available, one
would hope that a full chapter (in a future book) can be dedicated to
coarticulation in child speech.
The chapter by Manuel ("Cross-language studies") is potentially very
important. As mentioned above, a study of coarticulation can provide
useful information on what are universal and what are language-specific
aspects of speech production. Manuel separates the chapter into two
sections: 1) the role of contrast, and 2) prosodic structure and vowel
harmony. The first section deals with the belief that the greater the
number of vowel or consonant phonemes in a language, the more
constrained the speaker is to produce a more "precise" and "canonical"
token of a given phoneme, in order to avoid perceptual confusion and to
facilitate comprehension on the part of the listener. Unfortunately,
most of these studies are a) based on acoustic results only, and b) not
readily available (I was not familiar with many of the works cited in
this chapter). Whilst Manuel's own work seems to support the above
hypothesis, at least in acoustic terms for vowels, it is not clear that
this hypothesis is a good one. Firstly, the cues to vowel identity
(mainly F1 and F2, with F3 and perhaps nasal resonances for more
complicated systems) are much fewer than those to consonant identity,
and hence the variation could be thought to be more limited in larger
vowel phoneme systems. Secondly, it is well-known that a stable acoustic
target can be produced using many different articulatory strategies and
constriction locations (Stevens 1989), so that acoustic stability does
not necessarily imply articulatory or coarticulatory stability.
Nevertheless, this is a very important chapter, in that it provides a
summary of what is known regarding coarticulation across different
languages. As Farnetani points out in her chapter on labial
coarticulation (p. 162), more *explicitly* cross-linguistic studies are
needed of coarticulation, and to this I would emphasize both acoustic
and articulatory.
The second part of Manuel's chapter deals with vowel harmony and with
prosodic factors such as stress. As noted above, Ohman's (1966) results
showing that vowels are coarticulated across consonants suggest that
speech is a sequence of vowel gestures, with consonants superimposed.
Given this view, it can be seen that vowel harmony is an extreme form of
coarticulation, with the vowels maximally coarticulated within a given
word. It would seem that this is a language-specific form of
coarticulation, since not all languages exhibit this extreme sort of
behaviour (although, it appears, all languages coarticulate vowels
across consonants to a certain extent). The section on prosodic factors
is not convincing, I believe, perhaps because it is couched in terms of
"stress-timed" vs. "syllable-timed" languages, an impressionistic
description of languages for which, to date, there is no empirical basis
(for instance, English is supposed to be a stress-timed language, in
which temporal distances between stresses are isochronous, whereas
French is supposed to be a syllable-timed language, in which temporal
distances between syllables are supposed to be isochronous).
The chapter by Beckman, although not as clearly delineated as some of
the other chapters, provides a stimulating discussion of what is the
phonology and what is the phonetics of coarticulation. She concludes
that better acoustic models of the effects of coarticulation are needed,
in order to better understand the role of perceptual input to the child
learner of language (and I would add, to the adult speaker of language).
The five chapters contained in section II, "Research results", are
mostly very clear and very well-referenced. The chapter on lingual
coarticulation by Recasens in particular is a model of scholarship and
conciseness, all the more so since lingual coarticulation (in particular
the coarticulation of tongue-tip and tongue-body coarticulations) is one
of the more well-studied areas of coarticulation. The chapter on labial
coarticulation (another well-studied area) by Farnetani contains a
welcome, brief description of the muscles involved in lip-rounding, with
diagrams. Such a description (with diagrams) may have been of interest
to readers of other chapters, such as that on lingual articulation,
although it should be admitted that this may at times have been beyond
the scope of a given chapter, or have made it too bulky. Whereas
Farnetani's chapter focuses more on the horizontal aspect of lip
coarticulation (i.e. lip-rounding), the chapter on lip and jaw
coarticulation by Fletcher and Harrington focuses on the vertical aspect
of lip and jaw coarticulation.
The chapter in which I was most disappointed is chapter 5 on laryngeal
coarticulation. Even after a brief re-reading of the chapter, I was
confused as to what the reader was meant to learn from it. I believe
this confusion is mostly due to the artificial separation of the chapter
into two parts, with two different sets of authors (Hoole for the first
part, and Gobl and Ni Chasaide for the second part). I believe that much
confusion would have been avoided and that the chapter would have been
much more coherent if it had been given to only one set of authors. For
instance, there is overlap in some of the topics treated. On p. 141, for
example, Gobl and Ni Chasaide discuss the early abduction of the vocal
folds in voiceless fricatives, which was precisely the topic of Hoole's
section (see especially, pp. 108-109).
The other problem I have with this chapter relates to the fact that it
deals with aspects of speech which I believe are at best peripheral to
the concerns of coarticulation, and could by some definitions be
considered not to be coarticulation at all. Consider Manuel's definition
of coarticulation on p. 179:
"patterns of co-ordination, between the articulatory gestures of
neighbouring segments, which result in the vocal tract responding at any
one time to commands for more than one segment"
By this definition, the study of the timing and amplitude of the
laryngeal abduction gesture in voiceless stops versus voiceless
fricatives does not count as coarticulation. Although the laryngeal and
oral gestures are co-ordinated, it is for the production of a single
segment, not for neighbouring segments. Of course, when a stop and
fricative are neighbouring segments, the laryngeal gestures for the two
must be co-ordinated, and this often leads to overlap (in fact,
depending on the strength of the prosodic boundary between /s/ and /t/,
the two separate laryngeal gestures may merge into one,
greater-amplitude gesture), and the chapter does indeed deal with this
case. It is possible that I had greater trouble following this chapter
because my own research has not touched upon laryngeal control; however,
I have also not done any research into velopharyngeal control, and I had
much less trouble following the chapter on velopharyngeal coarticulation
(by Chafcouloff and Marchal) than I did that on laryngeal
coarticulation.
The final section, entitled "Instrumental techniques", contains seven
chapters which provide a good, clear introduction to each given
technique. They provide a useful overview of the various techniques used
in coarticulatory studies, with the advantages and disadvantages of each
technique for the most part clearly outlined (however, some chapters
overlap to a degree with chapters in Hardcastle & Laver's 1997 book).
Most of the chapters are, as usual, very well-referenced, apart from
Stone's chapter on imaging techniques, which provides few references
apart from her own work. Although Stone is clearly a leader in this
field, I would have appreciated references to basic papers on the
techniques described, which are mostly found in biophysical and
bio-medical engineering journals. However, Stone does provide good
reasoning for why imaging techniques (such as x-ray, ultrasound and MRI
which provide 3D views of the vocal tract) should be used for
coarticulation studies, given that most of these techniques have thus
far been restricted to the study of static speech sounds.
Other chapters in this section include descriptions of palatography (by
Gibbon & Nikolaidis), used to study contact between the tongue and the
palate; electromagnetic articulography (by Hoole and Nguyen), used to
study movement of the tongue, lips and jaw; electromyography (by
Hardcastle), used to measure action potential in muscles; transducers
for investigating velopharyngeal function (by Chafcouloff); and various
techniques for investigating laryngeal articulation (by Hoole, Gobl and
Ni Chasaide). This chapter on laryngeal techniques, like chapter 5, is
divided into two sections, the first written by Hoole and the second by
Ni Chasaide and Gobl. I found that there were fewer problems with this
division of labour than there was for chapter 5, since each section
deals clearly with a separate technique, with no overlap. Hoole deals
specifically with the articulatory techniques, whereas Ni Chasaide and
Gobl deal specifically with inverse filtering (an acoustic technique for
isolating the glottal source based on the entire acoustic output of the
vocal tract). The final chapter, by Recasens, deals with acoustic
analyses of coarticulation. At first I was surprised at the inclusion of
this chapter, since I normally think of coarticulation as an
articulatory phenomenon (despite my protestations to the contrary
above!). However, this chapter simply provides a list of the various
acoustic cues (and there are many) that have been used in an attempt to
determine the extent and the nature of coarticulation in the
articulatory domain. Although the chapter may initially seem like a
"downer" after so many mainstream articulatory and even theoretical
chapters, it is in fact an excellent reminder that the study of
coarticulation is ultimately meaningless unless it can be related to an
acoustic output, and hence to the perception of the speech signal.
In sum, I strongly recommend this book to anyone with an interest in
speech production, speech acoustics, speech perception, phonetics or
phonology.
BIBLIOGRAPHY
Browman, C. & Goldstein, L. (1992) "Articulatory phonology: an overview"
Phonetica (49) pp. 155-180
Hardcastle, W. & Laver, H. (eds) "The Handbook of Phonetic Sciences"
Oxford: Blackwell (1997)
Hardcastle, W. & Marchal, A. (eds) "Speech Production and Speech
Modelling" Dordrecht: Kluwer (1990)
Keating (1988) "The window model of coarticulation: articulatory
evidence" UCLA Working Papers in Phonetics (69) pp. 3-29.
Lindblom, B. (1990) "Explaining phonetic variation: a sketch of the H&H
theory" in Hardcastle & Marchal (eds) "Speech Production and Speech
Modelling" pp. 403-439.
Ohman, S. (1966) "Coarticulation in VCV utterances: spectrographic
measurements" Journal of the Acoustical Society of America (39) pp,
151-168.
Saltzman, E. & Munhall, K. (1989) "A dynamic approach to gestural
patterning in speech production" Ecological Psychology (1) pp. 333-382.
Stevens (1989) "On the quantal nature of speech" Journal of Phonetics
(17) pp. 3-45.
BIOGRAPHY
Marija Tabain received her Ph.D. from Macquarie University, Sydney, in
1999, for a thesis entitled "Articulatory and acoustic aspects of
coarticulation in CV syllables". She has published in "Journal of
Phonetics", "Phonetica" and "Language and Speech". Her research
interests include acoustic and articulatory phonetics, cross-linguistic
phonetics and phonology, coarticulation and Australian languages. She is
currently a postdoctoral research fellow at the Institut de la
Communication Parlee in Grenoble, France.
|
