Date: Sat, 12 Feb 2005 00:25:03 +0100 From: Piotr Glowacki Subject: Constraints in Phonological Acquisition
EDITORS: Kager, René; Pater, Joe; Zonneveld, Wim TITLE: Constraints in Phonological Acquisition PUBLISHER: Cambridge University Press YEAR: 2004
Piotr Glowacki, Wroclaw University, Poland
This book is a collection of ten papers presenting a broad overview of current issues in phonological acquisition, especially aiming at considering them within the paradigms of Optimality Theory (OT). The volume was inspired by the papers presented during the Third Biannual Phonology Workshop organized by Rene Kager and Wim Zonneveld in June 1998 at the Research Department for Language and Speech of Utrecht University (Chapters 5, 6 and 9). They are accompanied by other papers, written on request. The volume is dedicated to the memory of Peter W. Jusczyk, who was one of the contributors.
Chapter 1: 'Introduction: Constraints in phonological acquisition' by René Kager, Joe Pater and Wim Zonneveld This chapter falls into four main parts and a conclusion. The first part presents an overview on previous research concerning the issues dealt with in this book. It begins with presenting the theoretical background - Jakobson's 'laws of irreversible solidarity' and Chomsky's concept of Universal Grammar (UG). As an application, N. V. Smith's case study of his son's language is presented (Smith 1973). Two issues are emphasized there, as they are relevant in the context of current phonological theory development: opaque rule interactions (resulting in counterfeeding rule ordering) and the nature of the underlying representations in child grammar. Exploration of the former topic leads to the explanation of Chomsky and Halle's theory of markedness. Then Stampe's criticism of the theory of markedness in the Natural Generative Phonology paradigm is presented, supported by the comparison of Danish and Tamil coronals behaviour. Yet another alternative view is presented, by Kiparsky and Menn, who treat Jakobson's and Stampe's theories as 'rather deterministic' and propose treating adult speech as a target at which a child aims in the process of language acquisition. They also stress the usefulness of output constraints. The focus now returns to Smith and his claim that child grammar's underlying representation are mainly built of adult surface forms, which is contrasted with Braine's 'partial perception hypothesis' and Menn's 'two-lexicon model'. Finally, Chomsky's Principles and Parameters Theory is presented on the example of Kaye's parametric theory of syllable structure.
The second part serves as a brief tutorial on OT, introducing its main concepts and notions, such as constraint ranking, Richness of the Base, Freedom of Analysis, division of constraints into markedness constraints and faithfulness constraints, Lexicon Optimization or Correspondence Theory. There is also very brief but condensed comparison of OT with classical Generative Phonology and Natural Phonology included, which I assume is very relevant. Then two learnability theories are briefly presented, Constraint Demotion Algorithm (CDA) by Tesar and Smolensky and Gradual Learning Algorithm (GLA) by Boersma and Hayes.
The third and the fourth parts include summaries of the following chapters grouped thematically.
Chapter 2: 'Saving the baby: Making sure that old data survive new theories' by Lise Menn The paper begins with the author's subjective opinion on what a successful theory should be and how it should profit from the previous theories managing (or not managing) to deal with some issues (part 1). In part 2 Menn analyses the ways in which output constraints were treated in phonology throughout the last fifty years. We can encounter here the theories of Stockwell, Jakobson (again), Jones (whose formalism Menn calls 'an intuition-killer), Ingram, Smith's puzzle-puddle (again), Menn's 'two-lexicon model' (again), Kisseberthian rule conspiracies (again), Stampe's Natural Phonology (again) and Menn's own views on output constraints throughout the years. She gives an exhaustive guide to the non- OT bibliography on child phonology afterwards. Part 3 is a short introduction to the role of OT in phonological acquisition, after which (Part 4) Menn presents a 'Historical Annotated Inventory of Things We Know About Child Phonology', where she attempts to state if fourteen problems (which were highlighted before whilst presenting the history) can be dealt with by means of OT, and if not, what is the cause. Finally Menn advises that OT should not be made 'a Theory of Everything' and postulates having several good partial models of language.
Chapter 3: 'Markedness and faithfulness constraints in child phonology' by Amalia Gnanadesikan This chapter is dealing with the analysis of a single child's grammar (the author's daughter Gitanjali), focusing mainly on the examples of the Emergence of the Unmarked. Gnanadesikan claims constraints are universal and innate and at the beginning of acquisition all markedness constraints outrank all faithfulness constraints and adopts Smith's (1973) view that a child's inputs are somewhat similar to adult outputs. She then presents a very brief outline of OT (again) and Correspondence Theory, and proceeds to the analysis afterwards. First she focuses on Gitanjali's syllable onsets, which cannot be longer than one consonant, and discusses how Gitanjali narrows multiple-consonant input onsets to her one-consonant output onsets. Gnanadesikan employs Universal Sonority Hierarchy here and compares Gitanjali's case to Sanskrit's reduction of onset clusters in the course of reduplication. Then she concentrates on Gitanjali's dummy syllable 'fi-' and the effect it has on the onset of the following syllable. The next part is concerned with the occurrence of coalescence within the two phenomena presented earlier. Next comes the analysis of the interaction of OCP (Obligatory Contour Principle - a constraint which prohibits adjacent identical elements (Archangeli 1997:122)) with other constraints regarding Gitanjali's syllable onsets. The paper finishes with a very brief comparison of the results with standard rule-based theory and the conclusion, which includes some remarks about the theories of Smith and Stampe.
Chapter 4: 'Input elaboration, head faithfulness, and evidence for representation in the acquisition of left-edge clusters in West Germanic' by Heather Goad and Yvan Rose The problem considered here is connected with the previous chapter; unlike there, here it is not an analysis of one-child grammar. The authors claim that the adult inputs are fully prosodified, and, whether children's inputs are similar to adult ones or not, they reduce their onset clusters following one of two patterns which the authors call 'sonority pattern' and 'head pattern'. The data presented comes from a number of sources concerning English, Dutch and German. Goad and Rose first analyse the construction of onset clusters in those three languages (indicate the difference between a branching onset and appendix + onset construction) and then focus on different inputs in sonority pattern and head pattern. The next step is presenting constraints which are going to be used in the tableaux coming next. Then the alternative way of explaining this phenomenon is discussed (constraint re-ranking instead of different inputs) and rejected. A discussion concerning the need for having constituent heads follows.
Chapter 5: 'Phonological acquisition in Optimality Theory: the early stages' by Bruce Hayes Hayes begins with a very short introduction describing his aim - to show that current phonological theory can work very closely with the experimental line of acquisition research. He then defines his view of phonological acquisition as including 'the child's internalised conception of the adult language' (p. 159). Presenting an overview of experiment results (Eimas's 'feature detectors', Kuhl's 'perceptual magnet', Werker's losses in phonetic abilities at the age of 10 months) serves as an introduction to discussing the three kinds of phonological knowledge, as described by OT (contrast, legal structures (phonotactics) and alternations). Because the experiments show that the children first acquire phonotactics and then learn alternations, Hayes addresses the former problem first. The aim is to develop an algorithm which 'given input data and constraint inventories, can locate appropriate constraint rankings' and which leaves legal input forms unaltered, but alters illegal input forms in the output. He starts with the explanation of Tesar and Smolensky's Constraint Demotion Algorithm (CD) (Tesar and Smolensky 1993) and its application to the data from Pseudo-Korean (a language similar to Korean, but containing only vowels and stops contrasting for aspiration). CD generates legal output from legal input, but it fails when it is fed with illegal input. Hayes therefore presents his altered version of CD, called Low Faithfulness CD, which performs much better on the same data. Yet Low Faithfulness CD is not perfect, and its limitations are mentioned. Then Hayes moves on to the learning of alternations. Basing his argument on the strongly emphasized belief that phonology is conspirational, he points out the importance of output-to -output correspondence and a particular stage of children's acquisition when they are vulnerable to dialect contamination.
Chapter 6: 'Syllable types in cross-linguistic and developmental grammars' by Clara C. Levelt and Ruben van de Vijver The authors begin with explaining their starting point and the theoretical assumptions. Then they present Blevins's (1995) 'Syllable Type Inventories' and four markedness constraints, which, along with one unspecified faithfulness constraint, are used to build a factorial typology of Blevins's syllables. This is compared with the learning path which Dutch language learners take to acquire the most complex type of Dutch syllable. Levelt and van de Vijver guide us then through the learning paths and attempt to explain why these ways, and not the other, were taken by the learners. They take into account, apart from typology, also the frequency with which children hear particular syllables.
Chapter 7: 'Bridging the gap between receptive and productive development with minimally violable constraints' by Joe Pater The author's aim, exposed in the introductory section, is to employ OT to shed some light on the differences between language perception and production in children's grammar. He claims that a gap arises 'when the perceptual representations are more marked than the representations evinced in production' (p. 220). His solution is to employ a 'perception specific faithfulness constraint'. The next section is devoted to presenting and explaining briefly two experiments: habituation/dishabituation procedure by Werker et al. (1998, 2000) and Headturn Preference Procedure by Jusczyk et al. (1999). Then Pater takes a part of the latter experiment and makes it example to illustrate his theory. He labels the standard OT notions of 'input' and 'output' as 'S' if it is a perceived surface form and 'L' if it is a stored lexical form, and, depending on if it is perception or production, each 'S' and 'L' can appear as either input or output. He also proposes Max(SL) and Max(LS) constraints, which make use of the notions presented before. Section 4 compares Pater's model with other approaches - Smolensky's (1996), Pater's 'mixed model' and 'dual lexicon model' by Menn and Mathei. Section 5 presents the conclusions.
Chapter 8: 'Learning phonotactic distributions' by Alan Prince and Bruce Tesar This chapter is, similarly to Chapter 5, an attempt to improve Tesar and Smolensky's Constraint Demotion Algorithm (CD). The focus of the introductory section is set on the Subset Problem (Angluin 1980) - that a learner using only positive evidence can make overgeneralizations apart from errors - and the attempts of solving it. Then the authors discuss the properties of the target grammar. Section 3 presents Recursive Constraint Demotion algorithm (RCD) - Tesar's (1995) modification of Tesar and Smolensky's (1993) CD. In the next section the authors propose modifying RCD in such a way that it would prefer markedness to faithfulness constraints, to make it more compatible with their position that the Markedness >> Faithfulness relationship helps coping with the Subset Problem. Their version is called Biased Constraint Demotion (BCD). Sections 5 and 6 focus on further issues connected with BCD, namely faithfulness gangs, markedness cascades, and subset relations between constraints.
Chapter 9: 'Emergence of Universal Grammar in foreign word adaptations' by Shigeko Shinohara Shinohara's aim is to show that Universal Grammar's unmarked patterns arise in foreign word adaptations even if they contradict the native constraint system. The author uses data obtained from Tokyo Japanese speakers proficient in French. He analyses three main cases: assibilation of the alveolar plosives, accentuation of loanwords from French and English and problems with syllabification of loanwords into Japanese.
Chapter 10 'The initial and final states: theoretical implications and experimental explorations of Richness of the Base' by Lisa Davidson, Peter Jusczyk and Paul Smolensky The paper includes two experiments: concerning the initial state of the grammar (infant speech) and the final state (adult speech). The former experiment tests the presence of the initial Markedness >> Faithfulness ranking in infant grammar. The technique used is the Headturn Preference Procedure and the data used is nasal assimilation in English. The latter experiment examines the ability of English speakers to pronounce word- initial consonants clusters which are not tolerated by English phonotactics. The authors propose employing 'floating constraints' (e.g. Reynolds 1994) and a new, almost-one-page-long, definition of Extended Richness of the Base.
Chapter 11: 'Child word stress competence: an experimental approach' by Wim Zonneveld and Dominique Nouveau The authors examine the word stress competence of 3 and 4 year old children by giving them existing and nonsense words to pronounce. These tests are accompanied with basically the same tests performed on adults - as a comparison. They divide the stress patterns into four types: regular cases (A), exceptional cases (B), very irregular (C) and prohibited (P). Then a VERY detailed analysis of the experiment results is presented using OT framework, including even the NO-CLASH constraint, which has its roots in Metrical Phonology (Liberman and Prince 1977).
Chapter 1 presents a strong, detailed historical description of the studies on phonological acquisition. It is quite lucid, written in a formal language (except for 'Smith's original hunch' p. 12), clearly stating what each presented theory is concerned with and, in most cases, supporting the description with some examples. The explanation of OT begins with a note, which I assume is very helpful, indicating the books where a reader can find general, but less brief, introductions to OT.
Chapter 2 is written in a less formal way than Chapter 1, using much less formal language (e.g. "[W]hat kind of creature is the child such that 'guck' may be easier than 'duck'" p.57). Menn presents her own views, her own thoughts of other theories (e.g. calling Jones's formalism 'an intuition-killer'), or even her own theory ('The term 'two lexicons' is somewhat misleading' p. 57). Yet in my opinion it is the editors' task to avoid repeating the same information in two different chapters. Menn says at the beginning of her paper: '[W]hat I have been requested to do is to provide a historical perspective on OT as a theory of phonological development' (p. 55). Therefore I am wondering why the editors did the same job again writing the first chapter. The only reason I can think of is that Menn presented history in a subjective way and Kager et al. did it rather objectively and in greater detail. However, I suppose that repeating the same citation (Ch. 1, p.1 and Ch. 2, p. 59) should have been dealt with by the editors.
Chapter 3 is a well-planned investigatory paper with much language data and many tableaux. Gnanadesikan presents both the data and the conclusions very clearly. Yet I find the definitions of constraints a little vague. There is a short, concise introduction stating directly the author's views and beliefs, including a brief 'outline of OT' (p. 18 and 75), with a similar example tableau (p.19 and 75). I think that part 7, where Gnanadesikan compares her results with standard rule-based theory could have been discussed in greater detail.
Chapter 4 is written in a very scientific language. Once a reader gets used to it, the understanding ceases to be a problem. Numerous tables and charts (especially those dealing with the internal structure of onset clusters) are very helpful. Constraints are defined very clearly and in great detail.
Chapter 5 is in my opinion the most reader-friendly chapter in this book. Hayes divides it into many parts and sub-parts, which make the paper very lucid. The language is rather formal, but in my opinion the author has some kind of gift which keeps the reader really awake during reading (and it is not the first work by Hayes I have read). Hayes carefully explains the reader what he is doing when and why. And thank him for note 6 ('For reasons of space, I cannot provide a summary of OT' - p. 196)! Another case of space limitations is carefully solved by directing the reader to Hayes' own Web page, where one can find full descriptions of the simulations which results are discussed here, software helping to conduct such simulations, and, among others, this very paper (!).
Chapter 6 is my favourite. Concise, lucid and to the point from the very beginning to the very end. Ezra Pound would be proud of it - almost every word is necessary there! The authors do not worry to remind the reader that they do not present The Only Truth ('One explanation [i]s that we do not have enough acquisition data yet. Awaiting a larger study, an alternative explanation is explored below' p. 212). In my opinion, this paper is a masterpiece.
Chapter 7 is, in my opinion, not very clear. Pater seems to be talking about several different topics at once, not linking them together well. It may be so because the chapter is an altered version of ROA-296, 'From phonological typology to the development of receptive and productive phonological competence: Applications of minimal violation'. The main difference between the papers is that the ROA version follows the Parse- Fill theory, and the 2004 version employs Correspondence Theory. The text has been changed very much, even by deleting some chapters and writing new ones, yet some portions have been left unaltered. That makes the discourse less comprehensible - for me the ROA version was much more informative and easier to read.
Chapter 8 is very well-planned. The authors guide the reader through the intricacies of learning algorithms. I especially liked presenting algorithms in a 'generic pseudocode', which looks like a kind of programming language (and has comments!). It helps to understand the algorithm very much. The constraints used are explained very clearly and three appendices show additional problems and further research areas.
Chapter 9 presents a fair argument. Shinohara has much data and shows many examples to the reader. He delivers a strong theoretical basis for his research as well. In my opinion the problem he touches is extremely interesting, especially when one has ever tried to talk with a Japanese person in English.
Chapter 10 touches the very core of OT and sheds some new light on the understanding of the theory. With proposing the new definition of Richness of the Base, the authors claim all work done under standard Richness of the base 'to be the study of base grammars, which are determined by the inventory of native forms' (p. 341). Yet I have some doubts if using Headturn Preference Procedure is really a good measuring device for infants. In note 4 the authors inform us that "[a]dditional infants [...] were tested but not included for reasons of excessive fussiness or crying, FAILING TO ORIENT PROPERLY TO THE TEST APPARATUS, experimenter error, or parental interference" (my emphasis). It means that for experiment 5 only 16 from 30 results were included! However, the question we should ask is WHY the children failed to orient properly. There are many other factors that can distract the children and alter the results significantly - even a child's own hand. The results obtained using this method do not persuade me. Having analysed the results of the experiments I conclude that they provide too weak a basis for any conclusions. Mean looking times do not differ much; moreover, if we look at Standard Deviation, we can see that the results overlap in the worst case! The number of children tested (16) was too small in my opinion, and the experiment should be redone and the results compared.
Chapter 11 utilises the experimental results to the maximum. The analysis is very detailed, yet all the time clear. The experiment itself was very simple, easy to perform and did not need elaborate machinery, hence the results should be quite reliable. The authors present their reasoning in a lucid, 'user-friendly' way. It is a wonderful dessert at the end of the meal.
The book is very well produced, there are only a few typos (I found four) and after days of intensive reading it is still in one piece. The topics of chapters partially overlap, which makes the book more coherent; however, it causes repetition, most clearly visible in the chapters 1 and 2. The book was in preparation for a very long time - that is why most of the papers were written in the last century (but fortunately revised recently). Not all papers benefited from revision - Chapter 7 was much more comprehensible in the earlier version, which I found on the Web - as well as all but two chapters (Ch. 2 and Ch. 11) - very easily and for free. Most of Web versions differ from their book variants only very slightly (except Ch. 7). 'Constraints in Phonological Acquisition' follow the trend in the current phonological literature of publishing in print older papers which most phonologists have read and know, but which have never been published in print before.
Angluin, D. (1980). Inductive inference of formal languages from positive data. Information and Control 45. 117-135.
Archangeli, D. and D. T. Langendoen (eds.) (1997). Optimality Theory: an Overview. Oxford: Blackwell.
Blevins, J. (1995). The syllable in phonological theory. In: Goldsmith, J. (ed.) The Handbook of Phonological Theory. Oxford: Blackwell.
Jusczyk, P., D. Houston, and M. Newsome (1999). The beginnings of word segmentation in English-learning infants. Cognitive Psychology 39. 159-207.
Reynolds, W. (1994). Variation and Phonological Theory. Ph.D. dissertation, University of Pennsylvania.
Smith, N. V. (1973). The Acquisition of Phonology: a Case Study. Cambridge: Cambridge University Press.
Smolensky, P. (1996). On the comprehension/production dilemma in child language. [ROA-118, http://roa.rutgers.edu].
Tesar, B. (1995). Computational Optimality Theory. Ph.D. dissertation, University of Colorado, Boulder. [ROA-90, http://roa.rutgers.edu].
Tesar, B. and P. Smolensky (1993). The learnability in Optimality Theory: An algorithm and some basic complexity results. [ROA-2, http://roa.rutgers.edu]
Werker, J., L. Cohen, V. Lloyd, M. Casasola and C. Stager (1998). Acquisition of word-object associations by 14-month old infants. Developmental Psychology 34. 1289-1309.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Piotr Glowacki is an MA student at English Philology Institute, Wroclaw University. His research interests include phonetics and phonology, especially acoustic phonetics and Optimality Theory. He is currently writing his MA thesis focusing on Optimality Theory.