This book presents an overview of the field of phonological acquisition within current linguistic theory. It is the first volume that brings together a number of prominent acquisition researchers working within the Optimality Framework (OT) (Prince & Smolensky, Reference Prince and Smolensky1993). A range of phonological data from typically developing children is explored from a number of perspectives: descriptive, experimental and computational. Also included is a chapter on loanword adaptation. This volume is useful for anyone who is working on or interested in phonological acquisition within a linguistics framework. There are, however, some shortcomings. Many of these stem from the fact that there was considerable delay between the conception of the book and the time that it was finally published. As a result, there are a few papers that first appeared, in various forms, in the mid-1990s and these papers are, on occasion, not reflective of the current advances in the field. Nonetheless, this volume has a number of strengths and demonstrates the advances made within OT in understanding children's acquisition of phonology.
The eleven chapters vary in readability, clarity and their approach to phonological acquisition. This is understandable and perhaps expected in a collection of papers. On occasion, there is unnecessary overlap in content. For example, the Introduction provides a comprehensive overview of OT, how it works and how it relates to acquisition, and Chapter 3 also includes a basic OT tutorial (likely because the original 1995 manuscript included, and required, one). There are a number of errors in the references and in the index, which is unfortunate. Each chapter has its own references, rather than consolidating all the references at the end of the book. As a result, there are discrepancies (e.g. Antilla's paper is cited as 1997 in one chapter and 1998 in another). These, however, are minor shortcomings, and the overall content of the book is impressive and comprehensive.
While the volume is not organized according to themes, a number of themes emerge. Chapters 1 and 2 provide an overview of phonological acquisition from both traditional and current approaches. Chapters 3 and 4 explore the acquisition of clusters. Learnability and modeling phonological patterns are the focus of Chapters 5 and 8. Experimental approaches are utilized in Chapters 7, 10 and 11. Acquisition of units above the segment is the focus of Chapters 6 and 11. Finally, loanword adaptation is discussed in Chapter 9. One quite obvious theme that is missing is phonological acquisition in disordered populations. A great deal of research within the OT framework has examined this population (for examples see Gierut, Reference Gierut1999; Barlow, Reference Barlow2001), and such research touches on all of the issues raised in this book: input, representations, nature of constraints, acquisition of features, acquisition of clusters, etc. The addition of this research area would strengthen future volumes.
One other central theme is the question of a possible initial state of the universal set of constraints, for which there has been much discussion in the acquisition literature. Chapter 3 by Gnanadesikan is the first in the book to explicitly address this issue (and also one of the first within OT acquisition). She assumes that markedness constraints initially outrank faithfulness ones, and the authors of the remaining chapters in this volume adopt this assumption. This position is experimentally tested and supported in Chapter 10 by Davidson, Jusczyk & Smolensky. Some researchers, however, argue that not all markedness constraints outrank faithfulness (at least at the onset of production, which is often taken to be the starting point in phonological acquisition research) (Velleman & Vihman, Reference Velleman and Vihman2002), and others argue that faithfulness initially outranks markedness constraints (Hale & Reiss, Reference Hale and Reiss1998).
The particular chapters are as follows. The introductory chapter provides an overview of acquisition research prior to the formulation of OT. Kager, Pater & Zonneveld discuss, among other works, Jakobson's implicational universals, Chomsky's generative approach and Universal Grammar, and Chomsky & Halle (Reference Chomsky and Halle1968). They provide a comprehensive discussion of Smith (Reference Smith1973), whose son Amahl has been the source of data for a number of acquisition studies. Following the review, the authors provide a solid tutorial on OT and identify three priorities of the framework: (1) accounting for universal patterns in phonology, including a substantive theory of markedness; (2) developing a formal theory of phonology to characterize the child's emerging competence; and (3) accounting for learnability. They then give a concise summary of the contributed chapters.
The history of phonological acquisition is again presented in Chapter 2, ‘Saving the baby: making sure old data survive new theories’, written by Menn. This review is complementary to the one in Chapter 1. Menn attempts to characterize which acquisition phenomena might be captured or explained in an OT approach and which might cause problems for the framework. She suggests that OT may not need to account for everything since other theories have not been able to either. This is a bit unsatisfying, especially since Menn does not attempt to identify which acquisition challenges should be accounted for and what the metric is for deciding whether or not a pattern needs to be explained in a theory of phonology. In particular, she suggests that perhaps several partial theories/accounts are required. However, how the acquisition patterns are divided up and how the different theories are to be integrated remains to be explained or discussed.
The contribution from Gnanadesikan, ‘Markedness and faithfulness constraints in child phonology’, is a slightly revised version of her 1995 manuscript. As noted above, this seminal paper was one of the first to explore acquisition data within the OT framework. Gnanadesikan argues that constraints are universal and therefore innate. Current research suggests that perhaps some or all constraints are not innate, but learned (Boersma, Reference Boersma1998; Curtin, Reference Curtin2002; Hayes, Reference Hayes, Darnell, Moravscik, Noonan, Newmeyer and Wheatly1999). However, given this innateness assumption, Gnanadesikan proposes that adult constraints should be sufficient for explaining child phonology. Thus, there is no need to have the child endowed with more representational levels or more rules/constraints than adults. With this assumption in mind, and support from data from Gitanjali, Gnanadesikan provides an analysis of onset clusters. Her proposal argues that sonority plays a role in determining which member of an onset cluster is retained in production.
This chapter also addresses the question and nature of constraint re-ranking. Gnanadesikan assumes that a re-ranking of constraints occurs with the promotion of faithfulness constraints. This assumption is challenged in a number of learnability accounts that are included in this book, such as those proposed in Chapter 5, by Hayes, and in Chapter 8, by Prince & Tesar. Additionally, the Gradual Learning Algorithm (GLA), proposed by Boersma (Reference Boersma1998) after this chapter first appeared, allows for gradual increases and decreases in constraint ranking, and thus eliminates the need for constraint re-ranking. In the notes section of this chapter, the editors comment on this more recent development and a number of other issues that have been addressed by more recent work. However, it is unclear why the editorial comments have not been integrated into the text.
The acquisition of clusters is further explored in Goad & Rose's Chapter 4, ‘Input elaboration, head faithfulness, and evidence for representation in the acquisition of left-edge clusters in West Germanic’. This chapter is well written and clearly lays out the authors' assumptions, arguments and posited constraints to explain some of the cluster patterns observed in child language. Goad & Rose argue that only some children follow a sonority-driven pattern, whereas other children follow a head-driven pattern. Assignment to a particular group relies solely on the children's productions of S+sonorant clusters, as other cluster patterns can be accounted for with either analysis. It would be nice, however, to have an independent measure for group assignment. Their main argument assumes full prosodic specification of inputs. This is crucial because the analysis completely depends on the input rather than on the constraints or ranking. Children who are ‘head-driven’ will preserve the head of the cluster in their productions. In order for the child to select the head, she must have an understanding of headedness (which, according to the authors, comes for free) and realize that there is no branching within an onset that is universally left-headed. The authors propose that the mechanism available to the child to select heads relies on relative prominence, which will initially assign head status by default to the least sonorous element in a cluster. While the authors refer to a mechanism, it is not clear what this mechanism is or how it works. Moreover, the child has to come with knowledge that head selection depends on position: most sonorous for nucleus, least sonorous for onsets. Overall, this is an interesting analysis that will need to be considered by researchers working on cluster acquisition.
There are two chapters that explore learnability. The first, Chapter 5, is Hayes' ‘Phonological acquisition in Optimality Theory: the early stages’, which is the most accessible chapter in the book. The second is Chapter 8, by Prince & Tesar, ‘Learning phonotactic distributions’. Both chapters focus on learning phonotactics and use the Constraint Demotion Algorithm (CDA) (Tesar & Smolensky, Reference Tesar and Smolensky1998) as their starting point. The result is two very similar chapters that end up making similar changes to the CDA. Prince & Tesar propose a Biased Constraint Demotion (BCD), and Hayes proposes his own variation of the CDA: Low Faithfulness Constraint Demotion. The conclusions that result from these chapters are that initially markedness and output-to-output correspondence constraints must outrank faithfulness constraints (as was argued in previous chapters); that throughout the acquisition process faithfulness constraints must be ranked as low as possible; and that general constraints which have more specific variation, such as positionally restricted faithfulness constraints, are ranked below the specific ones. It would be interesting to see how the BCD and the Low-faith CDA fare against the Gradual Learning Algorithm (Boersma, Reference Boersma1998), as the GLA (briefly outlined in Chapter 1) is the main learning algorithm competing with the CDA.
Chapter 6, ‘Syllable types in cross-linguistic and developmental grammars’, by Levelt & van de Vijver, is the first chapter in the book to argue that frequency of forms in the target language influences the child's developmental pattern. Usually, factors such as frequency are not considered in generative phonology or phonological acquisition (but see, for example, recent work by Zamuner, Gerken & Hammond, Reference Zamuner, Gerken and Hammond2004). Levelt & van de Vijver explore whether developmental grammars are comparable to fully developed grammars in terms of their syllable inventories. Their results demonstrate that some developmental grammars are not consistent with fully developed grammars. Moreover, not all possible language grammars are observed in the developmental progression from early productions to the final grammars. To account for some of the patterns, the authors appeal to constraint conjunction for markedness constraints. Whether or not one agrees with conjoining constraints and the implications that result (are all markedness constraints freely allowed to conjoin? is this part of universal grammar? etc.), the overall approach is interesting and novel, and provides an important and much-needed component for the understanding of developmental patterns. Based on their findings it will be necessary to address issues of frequency in future explanations of phonological development.
The seventh chapter, by Pater, ‘Bridging the gap between receptive and productive development with minimally violable constraints’, explores the reason for differences in children's ability to comprehend and produce language. Pater appeals to experimental research from a number of domains and draws on the work of Stager & Werker (Reference Stager and Werker1997) to argue that children have lexical representations that are reduced in segmental complexity. Stager & Werker found that in a word–object associative learning task, 14-month-old infants failed to look longer to a switch between the word–object pairing when the words minimally differed from each other (e.g. [bΙ] versus [dΙ]). Pater interprets this to mean that children's representations are underspecified, although it is important to note that this interpretation of the results is not the same as Stager & Werker's. The latter argue that the phonetic detail is not accessed due to computational demands. This does not mean that detail is not encoded (see Werker & Curtin, Reference Werker and Curtin2005, for further accounts of the speech perception and word learning data). In his overview of prosodic comprehension, Pater concludes that there is a genuine comprehension/production gap, when children truncate polysyllabic words, and he provides an analysis of the truncated productions within OT. He argues that we need perception-specific faithfulness constraints and that they are universally fixed in their ranking over production-specific faithfulness constraints. In order to accept Pater's proposal, one needs to accept his interpretation of the speech perception data, and, as noted, the correct interpretation is up for debate.
Sinohara's ‘Universal grammar in foreign word adaptation’, Chapter 9, is the only chapter that explores loanword adaptation. The main argument is that Universal Grammar is responsible for Japanese speakers' adaptations of French words. Sinohara's analysis for gemination of French consonants appeals to ‘Sympathy’ (McCarthy, Reference McCarthy1999). Unfortunately, how the Sympathy candidate is identified (and how Sympathy functions in OT) is not explained. Nor are implications of Sympathy in loanword adaptation and, more generally, in acquisition, ever discussed. Recent work by Dinnsen, McGarrity, O'Connor & Swanson (2000) addresses some of these issues.
The tenth chapter, by Davidson, Jusczyk & Smolensky, ‘The initial and final states: theoretical implications and experimental explorations of Richness of the Base’, is one of the most exciting collaborative works in the volume. The authors' goal is to experimentally explore the psychological reality of Richness of the Base (RoB), which states that no constraints hold at the level of underlying forms. In other words, the grammar must consider as possible output productions a set of inputs (base) that contains all universally possible inputs (richness). The authors discuss learnability and the initial state and provide experimental support for ranking markedness constraints above faithfulness constraints. The final state is explored by experimentally testing adults' ability to produce non-native clusters. The prediction for RoB is that regardless of the input to the adult grammar, the output will be consistent with the general pattern of the grammar. If RoB is correct, then the grammar will repair the inputs and produce legal sequences. The adult patterns result in repairs and are explained by probabilistically floating constraints. The authors conclude that RoB plays a role in linguistic performance and thus may be psychologically plausible.
The final chapter is also experimental. Zonneveld & Nouveau's ‘Child word stress competence: an experimental approach’ examines Dutch children's acquisition of stress. The results demonstrate that children aged three to four years have difficulty imitating irregular or illicit stress patterns. Moreover, children tend to regularize these patterns and preserve regular stress. The authors demonstrate that a No-Clash constraint, which does not appear to be responsible for patterns in the adult system, is nonetheless active in the grammars of children. They suggest that this further supports the claim that all markedness constraints precede all faithfulness constraints.
In short, this collection brings together foundational articles exploring phonological acquisition within the Optimality Theoretic framework, and combines them with articles dealing with current issues in OT using experimental approaches. While there are a number of shortcomings, this is still a highly important book for people engaged in phonological acquisition research and/or working in OT. Moreover, this volume would be a wonderful textbook to use in an upper level undergraduate or graduate course in phonological acquisition if used in conjunction with more recent articles.
