Bernhardt, Barbara, John Gilbert and David Ingram (eds.) (1996).
Proceedings of the UBC International Conference on Phonological
Acquisition. Cascadilla Press, Somerville, MA.
Reviewed by Charles Reiss <reiss@alcor.concordia.ca>
The papers in this volume were presented at a conference held in
June, 1995 at the University of British Columbia. The conference was
hosted by that institutionUs School of Audiology and Speech Sciences and
Department of Linguistics. The proceedings are arranged into the following
four sections which reflect the four sessions at the conference.
SECTION I: Feature Acquisition
Keren Rice, 'Aspects of variability in child language acquisition'
Isao Ueda, Segmental acquisition and feature specification in Japanese'
Mieke Beers, 'Acquisition of Dutch phonological contrasts within the
framework of feature geometry theory'
Judith A. Gierut, 'Featural categories in English phonemic acquisition'
David Ingram, 'Some observation on feature assignment'
SECTION II: Prosodic Structure Acquisition
A. Syllable structure
Joseph Paul Stemberger, 'Syllable structure in English, with emphasis on
codas'
M. Joao Freitas, 'Onsets in early productions'
E. Jane Fee, 'Syllable structure and minimal words'
B. Stress and word structure
John Archibald, 'The acquisition of Yucatec Maya prosody'
Katherine Demuth, 'Alignment, stress and parsing in early phonological
words
Kerstin Naucler and Eva Magnusson, 'Prosodic structure acquisition:
evidence from children's awareness'
SECTION III: Interaction of Prosodic and segmental tiers in
acquisition
Daniel A. Dinnsen, 'Context effects in the acquistion of fricatives'
Mary Louise Edwards, 'Word position effects in the production of
fricatives'
Marlys Macken, 'Prosodic constraints on features'
Shelley L. Velleman, 'Metathesis highlights feature-by-position
constraints'
Heather Goad, 'Consonant harmony in child language: evidence against
coronal underspecification
Carol Stoel-Gammon, 'On the acquisition of velars in English'
Conxita Lleo, 'To spread or not to spread: different styles in the
acquisition of Spanish phonology'
Clara C. Levelt, 'Consonant-vowel interactions in child languageU
SECTION IV: Instrumentation in phonological analysis
A. Instrumental analysis
Martin J. Ball, 'An examination of the nature of the minimal phonological
unit in language acquisition'
Eugene H. Buder, 'Experimental phonology with acoustic phonetic methods:
formant measures
from child speech
B. Computer modeling
William Turkel, 'Biological metaphors in models of language acquisition'
Steven Gillis and Gert Durieux, 'Data-driven approaches to phonological
acquisition: an empirical test
Comments
In an oft-quoted passage Chomsky (1965) characterizes the goals of
linguistic theory as follows:
Linguistic theory is concerned primarily with an ideal speaker listener,
in a completely homogeneous speech community, who knows its language
perfectly and is unaffected by such grammatically irrelevant conditions as
*MEMORY LIMITATIONS, DISTRACTIONS, SHIFTS OF ATTENTION AND INTEREST, AND
ERRORS (RANDOM OR CHARACTERISTIC)* in applying his knowledge of the
language in actual performance. This seems to me to have been the position
of the founders of modern general linguistics, and no cogent reason for
modifying it has been offered. To study actual linguistic performance, we
must consider the interaction of a variety of factors, of which the
underlying competence of the speaker-hearer is only one. In this respect,
study of language is no different from empirical investigation of other
complex phenomena. [EMPHASIS MINE-CR]
The emphasized part of this quote characterizes no one if it does not
characterize children. This fact has been recognized by some researchers.
For example, in her chapter on Phonological Acquisition in the Handbook
of Phonological Theory, Marlys Macken (1994) discusses the difficulty of
distinguishing competence from performance in children's output: "we must
attempt the difficult, perhaps impossible task of separating the grammar
from the processor...".
One approach to facing this challenge is to deny its existence. As
implausible as this approach is, it has been adopted, e.g., by Smolensky
1996 who rejects the notion that there is a `dramatically greater
performance/competence gap for children' (p.1).
Unfortunately, the papers included in the volume under review do not
consistently rise to the difficult challenge which Macken mentions.
Instead, the authors tend to accept children's output rather uncritically
as the basis for generalizations about the state of their grammars. This
forces me to be somewhat uncertain about the validity of the results and
hypotheses presented in this collection of papers.
Section I
The papers in Section I are all concerned with the 'order of acquisition'
of phonological features and the contrasts these features allow the
learner to exploit. All of the papers assume some form of
underspecification in early
grammars. It is interesting that this approach is in direct conflict with
some contemporary work on learnability of phonology. For example, Yip
(1996) comes to the following conclusion:
"This paper finds inconclusive evidence for abstract underlying
representations, and concludes that the balance of the evidence suggests
that learners acquire something rather close to what they hear, unless
information from alternations or paradigms forces them to do otherwise.S
Inkelas (1994) holds a similar opinion:
"underlying representation is determined solely by optimization with
respect to the grammar, not by imposing any type of constraints directly
on underlying representation ... [this] results in the use of
underspecification only when there are alternant surface forms..." [p1].
"[t]he only motivation for underspecification is to capture alternations
in the optimal way" [p.2]
Hale and Reiss (1995, 1996ab, 1997, to appear) argue that access to the
universal feature inventory at the initial state of the grammar follows
from the correct application of the subset principle to phonology. The
assumption of 'full access' is also consistent with the evidence for young
infants' superior performance on phonetic discrimination tasks than older
children and adults (e.g. Streeter 1976, Goodman and Nusbaum 1994).
While this contrast in viewpoints does not prove that either side is right
or wrong, it does highlight an unacceptable schism between researchers
working on learnability and those working with actual child data. Given
the fact that it is in the nature of science that theories partly
determine the data with which they are concerned, we must try to make a
coherent and consistent theory of the phonological learning path which
provides principled guidelines for the interpretation of the data.
For the sake of concreteness, consider Judith Gierut's paper which
describes a clever experiment meant to show that children who do not
produce theta (voiceless [th]) in their output also do not recognize this
sound as distinct from say [s]. Unfortunately, we are not given enough
detail or statistical analysis to decide if the following statement is
actually warranted: "Both groups of children were exposed to /th/ in the
ambient [English-cr] phonology, but only the [children who produced [th]]
differentiated this category". This surprising result contrasts with what
others have found:
"by the time infants are starting productive use of language they can
already discriminate almost all of the phonological contrasts of their
native language. While they cannot yet produce adult-like forms,
theyappear, in many respects, to have adult-like representations, which
are reflected, among other things, in their vociferous rejections of adult
imitations of their phonologically impoverished productions" [Faber and
Best 1994: 266-7]
Given such widely reported cases, we must be hesitant to accept Gierut's
claims that children's behavior shows us the state of their
representations. Gierut's experiment required not only attention to a
given stimulus, but also performance of a further task (putting chips into
a trashcan upon recognition of certain sounds). An instructive parallel
for accessing children's knowledge states is described by Karmiloff Smith
(1995). This author discusses work by Spelke and by Baillargeon to show
that earlier conclusions concerning children's lack of knowledge of
object permanence were probably misguided. Earlier experiments concluded
that the failure to manually seek objects that were shown to the child,
then hidden from view, pointed to the child not having the concept of
object permanence. However, when tested in a fashion that did not require
them to actively seek hidden objects, children do, in fact, show awareness
of object permanence. I mention this example, just to point out how
difficult it is to ascertain that we are in fact accessing children's
grammars in any particular case. It would be interesting to know if the
children in Gierut's study who do not produce [th] and are claimed not to
represent this sound as different from [s] would fail at a task requiring
them to distinguish lexical items like 'thick' and 'sick' spoken by an
adult.
Section II
These six papers contain a lot of interesting data and theorizing. Due to
a lack of space, I will restrict my comments to Archibald and Stemberger's
papers. Archibald's paper discusses Yucatecan Maya, which has been
variously described as a pitch accent, a tonal and a stress language. The
paper attempts to shed light on the question of how to describe the
language, as well as on how it is acquired. The use of pitch trackings and
spectrograms is not only useful to his discussion, but is also worthy of
emulation by ALL researchers of child phonology. The availability of
relatively inexpensive acoustic analysis software makes the use of
transcription by ear unacceptable for the description of children's speech
and the analysis of their phonological systems. The dangers of making
generalizations based on transcriptions of children's speech has been
reported but not widely heeded. For example, studies such as Kornfeld and
Goehl (1974) and current research at UCLA (Donca Steriade p.c.) indicate
that transcriptions of child speech are rife with inaccuracy. Acoustic
analysis reveals subtle distinctions, for example, between supposedly
merged adult /r/ and /w/, which transcribers tend not to observe. As Faber
and Best (1994:264) state, `[The] child may, despite the apparent lack of
contrast, have acoustic differences between 'red' and 'wed' such that the
initial consonants are perceived by adults as representing the same
phonemic category.' Maslon and Ross (1996) report that mistiming of
voicing by infants may lead to acoustically voiceless vowels and sonorants
which adult listeners will tend to ignore in transcription. Ignoring these
segments may even lead to unjustified reports of 'lost syllables'. So,
Archibald's use of phonetic data is to be applauded.
Stemberger's paper adopts a theoretical stance which appears to be
contrary to the goals of linguistic theory as commonly understood and as
laid out in the quote from Chomsky above. I believe that Stemberger's
views are unlikely to lead to any interesting results. However, it is
necessary to recognize the service he performs by laying out his
theoretical stance so explicitly. He appears to reject the notion of a
language faculty that is not just a combination of general cognitive and
physiological factors: "I assume that [linguistic] constraints are
universal because normal human beings have essentially the same vocal
tracts and cognitive systems...the constraints are not innate in a way
that is different from the innateness of the cognitive system and vocal
tract physiology" [p 62]. Stemberger suggests that the
performance/competence distinction is not a valid one, and adopts a
version of OT which is even more functionalist than many of the other
versions currently around: "The constraint ranking defines what is
relatively "easy" *for a given speaker* vs. what is relatively difficult."
Since he also assumes that "what is relatively easy (a) is different for
every person, and (b) can be changed via learning" [pp.62-3, emphasis in
original] it is difficult to understand what purpose the functionalist
approach serves.
Section III
In my opinion, most of the papers in this section are particulary
susceptible to criticism for not clearly distinguishing the aspects of
children's output which are due to their grammars (competence) and that
which is due to performance effects involving motor control as well as
cognitive factors like motor planning and shifts of attention. Some of the
papers do, thankfully, provide enough information to allow the reader to
make her/his own judgements. For example, in discussing substitution of
fricatives by other sounds, Mary Louise Edwards reports that "Rather than
an orderly progression in the types of substitutes that
occurred...different types of substitutes usually occurred in each
session" (150-1). Are we willing to draw conclusions about child PHONOLOGY
based on such TmessyU data?
Edwards paper also provides an interesting piece of data which is relevant
to many discussions in the child phonology literature: a 4;8 subject,
Jason, who is said to have a "severe" phonological disorder, produces
[pfw] "to represent nearly all word-initial liquid clusters, as well as
initial labial fricatives" (153). As far as I am aware, [pfw] is not a
very common initial cluster crosslinguistically. By any of the normal
measures such a cluster should be 'highly marked'. Since Jason is
classified as having a "severe" phonological disorder according to the
Goldman-Fristoe Test of Articulation, are we to assume that phonological
disorders can lead to more marked, viz., more complex phonology? This
example, along with others such as the cases of voiceless sonorants
mentioned above, lead me to be suspicious of several related claims in the
literature. First, is Jason's problem a phonological one? Second, is
markedness a coherent notion at all? And finally, is the claim of
"emergence of the unmarked" in children's speech, as discussed in the OT
literature (e.g. Smolensky 1996, Gnanedesikan 1995) at all coherent, given
that the literature is full of reference to kids (normal and 'disordered')
who speak like Jason?
The discussion of "velar fronting" by Carol Stoel-Gammon also raises
questions about the reliability of transcriptions. "Velar fronting" is a
phenomenon in which children pronounce velars as alveolars, e.g. 'go' is
pronounced [do]. Recent research involving palatograms (Jim Scobbie,
p.c.) has shown that some children who are apparently making [t] for [k]
are actually making both an alveolar and a velar closure, but the relative
timing of the release of the two closures is responsible for the
perception of a [t]. It would be interesting to know how many of the
reported "neutralizations" in the literature are actually similar to such
cases of Tdouble articulationU.
Section IV
Eugene Buder's discussion of the use of acoustic analysis in the study of
phonological acquisition is my favorite paper in this volume. While
portions of the paper may be too technical for many readers, Buder still
manages to provide some helpful hints for obtaining dependable readings
from spectrograms of children's speech, as well as some discussion of why
certain difficulties do arise.
Conclusions
The papers in this collection represent a range of theoretical viewpoints
and present data from a variety of languages, so they are definitely
worth reading. In response to some of the criticisms I have made here
concerning the methodology used in many acquisition studies, I have been
told that there would be nothing left for acquisitionists to do if we
decide that the data is, in general, misleading. I think that this is an
overly pessimistic view. There remain open several paths to a better
understanding of children's phonological systems, if we are willing to
devise ingenious experiments that force them to be revealed.
Bibliography
Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, M.I.T.
Press.
Faber, Alice and Catherine T. Best. 1994. The perceptual infrastructure of
early phonological development. In Reality of Linguistic Rules. Susan D.
Lima, Roberta L. Corrigan, Gregory K. Iverson, eds. Amsterdam ;
Philadelphia : J. Benjamins.
Gnanadesikan, Amalia E. 1995. Markedness and faithfulness constraints in
child phonology. ms., Rutgers Optimality Archive.
Goodman, Judith and Howard Nusbaum. 1994. The development of speech
perception. Cambridge: MIT Press.
Hale, Mark and Charles Reiss. To appear. Formal and empirical arguments
concerning phonological acquisition. Linguistic Inquiry. Early draft
available on Rutgers Optimality Archive.
Hale, M. and C. Reiss. 1997. Evidence in Phonological Acquisition:
Implications for the Initial Ranking of Faithfulness Constraints.
Proceedings of CLRF 28. CSLI Publications, Stanford CA.
Hale, M. and C. Reiss. 1996a. How to Parse (and How Not to) in OT
Phonology. To appear in Proceedings of NELS 27, McGill University. Ms.
Concordia University.
Hale, M. and C. Reiss. 1996b. The Comprehension/Production Dilemma in
Child Language: A Response to Smolensky. ROA.
Hale, Mark and Charles Reiss. 1995. On the Initial Ranking of OT
Faithfulness Constraints in Universal Grammar. Also available in
Concordia University Working Papers in Language and Linguistics and
Rutgers Optimality Archive.)
Inkelas, Sharon. 1994. The Consequences of Optimization for
Underspecification. Ms. UC Berkeley.
Karmiloff-Smith, Annette. 1992. Beyond modularity : a developmental
perspective on cognitive science. Cambridge, Mass. : MIT Press.
Kornfeld, J. R. and H. Goehl. 1974. A new twist to an old observation:
kids know more than they say. In Papers from the Parasession on Natural
Phonology, A. Bruck, R. Fox and M. LaGaly (eds.). Chicago:CLS.
Macken, Marlys A. 1995. Phonological acquisition. In John Goldsmith (ed.),
Handbook of phonological theory. Oxford: Blackwell.
Masilon, Todd and Kie Ross. 1996. Weak syllable deletion: an Articulatory
Phonological Account. ms. UCLA Linguistics Department.
Smolensky, Paul. 1996a. On the comprehension/production dilemma in child
language. LI 27.
Streeter, L. A. 1976. Kikuyu labial and apical stop discrimination.
Journal of phonetics 4:43-9.
Yip, Moira. 1996. Lexicon Optimization in Languages without Alternations.
Ms. University of California, Irvine. ROA.
The author of this review is Charles Reiss, Assistant Professor of
Linguistics at Concordia University in Montreal. Reiss received a PhD from
Harvard in 1995 and is interested in Phonological Theory, Acquisition and
Learnability Theory and Historical Linguistics.
|
