Consider the following three cases of phonological development.

First, there is a stage in the acquisition of English at which polysyllabic words are imperfectly realised as ‘truncated’, depending on the target's stress pattern: kangaˈroo as roo, poˈtato as ˈtato, ˈalligator as ˈagator (Kehoe Reference Kehoe1999). Next, these words are produced more target-like, resulting in syllabic accuracy for potato and alligator, but an intermediate stage of kanˈroo for kangaˈroo. In Optimality Theory (OT), using ranked constraints (Prince & Smolensky Reference Prince and Smolensky1993, Kager Reference Kager1999, McCarthy Reference McCarthy2002), such cases can be modelled as in the tableau in (1) for kangaroo, adapted from Kehoe (Reference Kehoe1999: 53–58) (capitalised syllables are stressed; accent indicates main stress, i.e. the word's prosodic ‘head’).

1. (1) ROO to KAN-ˈROO

   

At the earliest stage, monosyllabic ROO is selected as the optimal output from among a number of candidates (supplied by the Generator of Universal Grammar) by an Optimality evaluation process. The tableau in (1) lists constraints highest-ranked to lowest-ranked from left to right: Align-Head-Left enforces main stress on the leftmost syllable, whereas Align-Head-Right enforces main stress on the rightmost syllable; Right-Anchor requires right-edge syllables in the target to materialise (at the right edge) in the output; Stress requires syllables to be stressed; and Max Input-Output requires that no syllables be deleted. Constraint violations are indicated by an asterisk, and exclamation marks signal fatal violations. Following Optimality practice, the development from ROO to KAN-ˈROO (which comes to be the optimal output, ☺) can be visualised as a formal property of the grammar: based on the positive evidence of the systematically encountered correct pronunciation kangaˈroo, Align-Head-L is demoted to a bottom-ranked position of this partial grammar and out of the three surviving ROO-final candidates bisyllabic KAN-ˈROO is selected. The emergence of faithful kangaˈroo (which finally comes to be ☺) can be formalised as the subsequent demotion of the Stress constraint to the same bottom position in the final-stage grammar.

Second, ‘chain shift’ is an acquisition phenomenon that is attracting increasing interest and is more and more frequently observed. As the first true acquisitionist working in Chomsky & Halle's (Reference Chomsky and Halle1968) standard framework, Smith (Reference Smith1973: 149–150) discussed such a case – under the heading of ‘puzzles’ – in his meticulous study of the speech of his son Amahl (between age 2 ; 2 and 3 ; 7), and the (less detailedly analysed) output of his ‘nephew R’ (at age 3 ; 2). Amahl simultaneously exhibited puzzle becoming pu[d]le and puddle becoming pu[g]le, and nephew R realised sick as [θ]ik and thick as [f]ik. In chapter 6 of the book under review, ‘Innovations in the treatment of chain shifts’, Michele L. Morrisette & Judith A. Gierut report on a shift of exactly the latter type for a child with a delayed phonological competence, providing an OT analysis (211–217) of the post-shift longitudinal data in stages, as represented in the tableau in (2).

1. (2) Chain Shift: s>θ>f

   

In this chain shift pattern, two constraint families interact. Two markedness constraints ban the fricatives /s/ and /θ/, and two identity constraints, for stridency/noise and for place of articulation, demand ‘faithfulness’ between candidates and targets. Local Conjunction of Identity (LC-Id) is a – not uncontroversial – constraint type, which only indirectly refers to candidates' phonological characteristics. In this case, LC-Id is violated when the two identity constraints are simultaneously violated. The constraint ranking selects [θ]ok for sock and [f]umb for thumb, thereby capturing the shift. Morrisette & Gierut describe the results of attacking the shift in a phonologically informed clinical treatment (at age 4 ; 2–4 ; 6). Testing immediately after treatment revealed the emergence of [s]ock, whereas [f]umb was not affected; this can be modelled as the demotion of *s below *θ. Two months later, initial [θ] started emerging, as in [θ]ief; this can be modelled, based on the target grammar, as the onset of the demotion of *θ below Id-Place.

Third, in chapter 8 of the book under review, ‘Recalcitrant error patterns’, Daniel A. Dinnsen discusses two cases of delay observed in two children at age 4 ; 2, and 4 ; 10, respectively. The shared pattern is velar fronting, which replaces velars with dentals, e.g. gum is realised as [d]um, and big as bi[d]. By Optimality logic, the grammars of these children share a high-ranked markedness constraint that bans velars: *Velar » Id-Velar. One would expect clinical treatment, providing a sufficiently large number of input velars, to enforce demotion of the markedness constraint, resulting in across-the-board improvement. But this is not how things actually developed. Dinnsen observes that

Thus, the postvocalic context seems to be the easier one for the acquisition of velars. This suggests two possible treatment options: a ‘conventional’ one that targets the easier, postvocalic context first, and another that targets the more difficult, ‘marked’ initial context first, hoping to benefit from the observed implicational relation between the initial and final contexts. One child in Dinnsen's study underwent the first kind of treatment, resulting in postvocalic target-appropriate velars, and only those. Another child underwent the second kind of treatment, resulting in the suppression of the error pattern both word-initially and postvocalically. The observed acquisitional asymmetry, which was thus clinically confirmed, can be modelled in OT as a fixed, inviolable ranking of two partly overlapping constraints in a special-case–general-case relationship: *#Velar » *Velar, i.e. a ban on word-initial velars outranks a ban on velars as the marked place of articulation. The interaction between the two constraints can be represented as in tableau (3).

1. (3) Velar fronting in acquisition

   

A piece of clinical input data with a postvocalic velar violates *Velar, but not *#Velar, and demotion of just the former constraint accounts for bi[g] and incorrect [d]um. Conversely, initial velar input will violate both constraints, but it is especially the directly targeted violation of *#Velar that accounts for the across-the-board change observed in the second child: the fixed ranking implies demotion of both the higher and the lower constraint, modelling exactly the clinically obtained result.

The above examples, especially the second and third case, are the stuff that Optimality Theory, phonological acquisition and disorders (henceforth OTPAD) is made of. It presents modern theoretical, rather than just theoretically informed, clinical phonology. In the words of the editors, the approach ‘consists of a two-step process with theoretical and applied missions’ (38–39). Thus, they not only have the legitimate (albeit limited) aim of showing how phonological theory, in an OT model, can be applied to clinical data, but also intend to make a contribution to theory development by using clinical data to fill gaps in the empirical underpinnings of OT and, where appropriate, weighing up different theoretical options (e.g. local conjunction vs. ‘comparative markedness’, controversial ‘faithfulness to the marked’, etc.) on the basis of those data.

OTPAD's chapter 1, by Daniel A. Dinnsen, is an introduction to the ‘Fundamentals of Optimality Theory’. It is followed by two chapters (discussed below) that describe the book's underlying database and explain methodological issues. The database derives from the Developmental Phonology Archive created by the Learnability Project at Indiana University, Bloomington (www.indiana.edu/~sndlrng), where the editors are principal investigators. The main body of the book contains eleven chapters, subdivided into three parts, which provide clinical-phonological ‘research reports’, all of them theoretically couched in OT. Part II (chapters 4–6) deals with ‘Opacity effects’; part III (chapters 7–9) is concerned with ‘Developmental shifts and learning’; and part IV (chapters 10–14) addresses the ‘Acquisition of consonant clusters’. Finally, chapter 15, ‘On the convergence of theory and application’, by the editors, summarises and concludes the book.

Together with the ‘Preface’ by the editors, chapter 2, Judith A. Gierut's ‘Phonological disorders and the Developmental Phonology Archive’, informs the reader of the book's aims (cf. the quote above) and about the Indiana University Learnability Project, which started in the late 1970s, ‘focus[ing] primarily on the phonologies and learning patterns of young children between the ages of 3 and 7’ (ix) ‘with functional phonological disorders who are acquiring English as their first (and only) language’ (41). A phonological disorder is defined (39) as ‘a breakdown in the production, perception, or processing of sounds in the ambient language in the absence of any other overt symptoms’, where ‘children with phonological disorders typically present with a severely limited repertoire of sounds in the inventory’. Breakdowns and severely limited repertoires can also refer to (and in fact here often do refer to) cases involving ‘sounds that are considered “late acquired” based on normative scales’ (39). To date, the project has enrolled 279 children, and the database contains more than 700,000 utterances, ‘providing a wealth of descriptive data for linguistic comparisons and analyses’ (38). The study of the behaviour of atypical populations is claimed to afford two distinct research advantages: one is that the children are able to complete more complex linguistic tasks than their younger peers, i.e. subjects are typically cognitively more competent; a second is that the study of phonological disorders, when it includes clinical remediation, can serve as ‘an experimental testing ground, [where] the phonological changes that take place over the duration of treatment provide snapshots of learning, thereby allowing us to trace the emergence of successive grammars for insight into the longitudinal course of language development … [T]he validity of phonological claims and theoretical proposals can be assessed by evaluating longitudinal data from a given child, in conjunction with the cross-sectional results obtained across children, to form a typological convergence of evidence’ (ix–x).

Chapter 3, ‘Fundamentals of experimental design and treatment’, also by Judith A. Gierut, is an extremely instructive chapter for any investigator grappling with practical and methodological issues of longitudinal research: baseline competence, developmental group heterogeneity, maturation, day-to-day learning, etc. The central methodological notion espoused here is that of ‘generalisation’, i.e. the essential result that the effects of treatment be visible in accurate output beyond the original test items.

Part II of the book considers opacity, i.e. interacting phonological processes of which at least one is strikingly not surface-true – a phenomenon which has been of considerable linguistic interest since Kiparsky (Reference Kiparsky and Koutsoudas1976) formulated the claim that opacity is ‘marked’, in the sense of tending to be avoided in grammars, and probably hard to learn because it is not fully motivated by the primary linguistic data. While opacity tends to be much more common than originally thought, ‘few cases have thus far been identified from developing phonologies’ (145). Not only is OT a theory for which ‘(not) surface-true’ is not a significant concept, given that constraints are routinely violated even in optimal candidates, opacity has also been shown to be a hard nut to crack. Existing approaches, such as local conjunction (Smolensky Reference Smolensky1995, Łubowicz Reference Łubowicz2002), sympathy (McCarthy Reference McCarthy1999), Output-Output correspondence (Benua Reference Benua1997), and comparative markedness (McCarthy Reference McCarthy2003), extend the basic constraint set in some way or other, either by adding constraints referring to tableau characteristics (as in local conjunction and sympathy analyses), or by dividing up constraint families (faithfulness and markedness, respectively) so as to enable the creation of empirically more adequate hierarchies (as in Output-Output correspondence and comparative markedness analyses).

Chapter 4, Daniel A. Dinnsen's ‘A typology of opacity effects in acquisition’, presents the following typical development: a transparent stage 1 with deaffrication in data such as chew [tu], velar assimilation in tiger [kaɤgοʊ], and both processes in cheek [kik]; in stage 2, following targeted treatment of velar assimilation without deaffrication, velar assimilation was lost from tiger but not from cheek, making the process opaque; at stage 3 all output was target-faithful. Dinnsen argues that both local conjunction and comparative markedness appear to be able to deal with the opaque stage, allowing constraint demotion as the mechanism of grammar change, although comparative markedness ‘probably enjoys a slight advantage’ over conjunction when a certain amount of variation in velar assimilation at stage 2 is taken into account (174).

Other cases considered include blocking in derived environments (e.g. correct har[dl]y ‘hardly’ vs. velarised pu[g]le ‘puddle’, as described in Smith Reference Smith1973), surface overapplication (e.g. vowel lengthening before both present and omitted voiced obstruents in stem-final position, as in [dɔːgi] ‘doggy’ vs. [dɔː] ‘dog’), an ‘unusual error pattern’, as described in Leonard & Brown (Reference Leonard and Brown1984), where a child produced all words with final /s/, and the chain shift given above. In chapter 5, ‘An unusual error pattern reconsidered’, Daniel A. Dinnsen & Ashley W. Farris-Trimble suggest that ‘as developments in phonological theory have been brought to bear on each of these cases, many of these error patterns have been rendered more tractable and found to be consistent with general principles of phonology and acquisition’ (203).

Where part II deals fairly detailedly with stages of grammatical development, part III considers larger, more global changes. Chapter 7, ‘Developmental shifts in children's correspondence judgments’, by Judith A. Gierut & Daniel A. Dinnsen, is concerned with the role of Output-Output faithfulness in acquisition, i.e. similarity of a form not to the input/target, but to another output form that is present in the child's speech. In the study reported, thirty children, who either had or lacked accurate production of English non-strident fricatives, participated in an experiment that examined the children's similarity judgements for voiceless obstruents distinguished for place, fricatives distinguished from plosives, and strident fricatives distinguished from non-strident ones in CVC stimuli. Both groups of children could meaningfully and consistently detect similarities and differences between output items, which suggests that Output-Output faithfulness occurs in child speech, independently of whether a child's grammar is fully target-faithful for the sound class in question.

Chapter 8, by Daniel A. Dinnsen, discusses ‘Recalcitrant error patterns’. Treatment of chain shifts and ‘implicationally related errors’ often turns out to be only partially successful, and this chapter attempts to understand why. The provisional answer lies in the observation that, typically, more than a single markedness constraint requires demotion to arrive at accuracy, a fact which the treatment will have to recognise. Recalcitrant patterns can be taken to indicate that more than one markedness constraint is present in the grammar, leading to the assumption of two markedness constraints (a general and a specific one) in the velar fronting example discussed above.

Chapter 9, ‘The prominence paradox’, by Daniel A. Dinnsen & Ashley W. Farris-Trimble, considers counterexamples to the Optimality-Theoretical claim (cf. Beckman Reference Beckman1998, Lombardi Reference Lombardi1999) that domain-initial positions, as opposed to domain-final ones, are ‘strong’, in the sense that this is where contrasts are preserved and neutralisation avoided. Final devoicing in languages such as German, Russian, and Dutch is a well-known example of this phenomenon. The authors find that often the reverse seems to be the case in (early) child speech, e.g. velar fronting may be limited to initial position, maintaining a final contrast. On the basis of this example and others (the chapter contains a large appendix that lists ‘Children's word-initial mergers’ as documented in the literature), the authors propose a ‘prominence shift’ from final to initial in the course of development, formalised as the demotion of a FinalProm constraint to below InitialProm in the course of acquisition. While this may capture the broad facts, it is not so easy to see what would be the child's trigger for this shift. The authors provide some psycholinguistic support for an early prominence of rhymes (i.e. syllable endings), motivating an early default ranking of the two prominence constraints in which Final » Initial; and they speculate that the shift occurs to allow lexicon growth.

Part IV opens with chapter 10, ‘Syllable onsets in developmental perception and production’, by Judith A. Gierut, Holly L. Storkel & Michele L. Morrisette, which provides experimental evidence for children showing awareness of the Sonority Sequencing Principle, according to which inside-out sequencing of consonants in syllable margins follows the sonority hierarchy, and, to some extent, also awareness of the branching, but not the number, of constituents, as the organising principles of onset structure.

Chapters 11 (‘Experimental instantiations of implicational universals in phonological acquisition’, by Judith A. Gierut) and 13 (‘A typological evaluation of the split-margin approach to syllable structure in phonological acquisition’, by Jessica A. Barlow & Judith A. Gierut) aim to experimentally verify two proposed implicational universals in English children with phonological delays. Based on studies of the acquisition of German and Spanish, it has been proposed that the occurrence of consonant clusters implies that of affricates (faithful sweep implies cheese), hence affricates appear before clusters, branching at the segmental level inducing branching at the syllabic level (Lleó & Prinz Reference Lleó, Prinz, Hannahs and Young-Scholten1997). This finding is confirmed and the authors account for it by ‘fixedly ranked’ constraints (*ClusterOnset » *Complex Segment) reranked separately (the latter reranked first) relative to Faithfulness. The prediction based on data from second-language learning that consonant-liquid clusters imply a contrast between /l/ and /r/ is found not to be confirmed. The apparent implication that languages that permit onset clusters also allow codas, but not the other way around, and the way this implication pans out in empirical distributional facts, have recently been used to argue that the old-school OT constraints of *Complex (no clusters) and *Coda (no final consonants) should be abandoned in favour of so-called ‘split-margin’ constraints (Baertsch & Davis Reference Baertsch and Davis2003) which relate mirror-image, inside-out versions of the Sonority Hierarchy to linear, left-to-right cluster structure. However, the delayed phonology data considered in chapter 13 is inconclusive as regards establishing the validity of such an approach.

Chapter 12, ‘Gapped s-cluster inventories and faithfulness to the marked’, by Ashley W. Farris-Trimble & Judith A. Gierut, deals with the acquisition of English word-initial s-clusters. Contrary to expectation, the acquisition of these clusters does not necessarily conform to the sonority hierarchy, displaying an order in which s+nasal and s+glide is acquired before s+liquid. Putting it in the broader context of ‘gapped’ or ‘disharmonic’ inventories, the authors discuss the phenomenon against the background of de Lacy's (Reference Lacy2006) ‘stringent markedness’ approach.

Chapter 14, ‘Constraints on consonant clusters in children with cochlear implants’, by Steven B. Chin, examines variation in the realisation of initial consonant clusters by profoundly deaf children with cochlear implants. For the author, such cases ‘provid[e] a unique opportunity to address the question of how robust language learning is in the presence of radically degraded and impoverished auditory input’ (430). The study investigates the speech of two children at two different stages of development. There are no surprising cases of reduction or consonant and cluster realisation compared with what is found in typically developing phonologies, and the variation can be accounted for by ‘indeterminate ranking’ of constraints (Anttila & Cho Reference Anttila and Cho1998). There are also few unexpected longitudinal results; where they do occur, the inconclusive conclusion is that these patterns raise the as-yet-open question whether they are ‘in some way related to [the child's] deafness and use of a cochlear implant’ (445).

OTPAD constitutes a success: it is sound, interesting and entertaining. It is enviably well written and well produced, attempting to explain issues both broadly (how events should be viewed) and in detail (how events should be described). The authors are well-versed in and familiar with OT, both at the descriptive and explanatory level, and one can imagine OTPAD serving a very useful purpose in an OT introductory course. It carefully and elaborately explains the underlying methodologies of its data gathering.

The book is interesting – possibly more interesting than expected – in its combined theoretical and applied orientation. Not only do the authors apply OT to atypical data, but they also aim to be OT theorists who contribute to and advance the theory. They forcefully argue that acquisition data and, especially, clinical data can serve this purpose.

Interestingly, Gierut suggests that ‘research on the Learnability Project has converged upon a single finding’ with respect to complexity in phonological change (113). A long-held assumption is that sounds that are acquired earlier are ‘easier’ than later acquired sounds. This view is a hold-over from Piagetian theory and the assumptions of maturational stage models. When applied to phonology, a stage model assumes that development follows a linear trend, such that sounds are acquired in a fixed sequence in accord with maturational development, and that early acquired sounds are easier to learn since they develop first. However, as argued in OTPAD, it appears instead that ‘complexity governs phonological change, such that more complex structure promotes the greatest gains in children's sound systems’ (113), where complexity can be associated with, for example, typological markedness. The prediction is that ‘treatment of marked structure implies unmarked structure, but not vice versa’ (114), as confirmed by the third case study discussed at the beginning of this review. It is not easy to be convinced, though, by the authors' expectation that clinical therapists will be able to successfully extract from this publication and other similar ones ‘recommendations … for direct application’ (457).

The book is entertaining in its discussion of child speech patterns, drawn from the Learnability Project's database (where, as they point out, each child can be considered to speak ‘a language’), and of problematic, controversial and/or enlightening OT issues.

Nevertheless, there are, to this reviewer's mind, also a number of weak points. Chapter 1, although a useful introduction, is not wholly felicitously geared towards the later parts. While it is sufficient for the book's purposes to present phonological acquisition (whether or not for an atypical population, and whether or not through treatment) as ‘constraint demotion’ in a grammar, the space-consuming explanation and illustration of the steps involved (which includes the use of W(inner)–L(oser)-style tableaus that do not recur anywhere else in the book) seem superfluous. The same can be said of the section on ‘Lexical diffusion’ (5.3), which contains a discussion of the ‘Generic word frequency faithfulness constraint’, whose multisyllabic name fails to disguise the fact that it does not at all explain why velars emerge first in high-frequency words but rather simply helps restate that they do.

Moreover, OTPAD's individual chapters, most of which are at least loosely based on previous publications by the members of the Learnability Project, are not very well integrated. Chapter 9, for instance, presents delayed complete velar fronting (i.e. velars are replaced by alveolar stops in all contexts) for a particular child. Chapter 1 does exactly the same for another child, but lacks any cross-referencing, and gives partly different references. Discussions of the same chain shift (θ>s>f) occur in more than one chapter, using constraint hierarchies that cannot be straightforwardly reconciled, again with only partial cross-referencing. While in chapter 4 comparative markedness is argued to be the better device for capturing these shifts, later chapters simply present local conjunction accounts. Temporal references such as ‘to date’ should have been edited out, and reading the book from cover to cover (as this reviewer did as per job description) can be frustrating at times. Virtually all cases presented in OTPAD deal with segmental phonology; one will search in vain for prosodic cases such as the example at the very beginning of this review. This is all the more a pity because of the pronounced OT interest in truncation, as also manifested in reduplication (cf. McCarthy & Prince Reference McCarthy and Prince1994, Reference McCarthy and Prince1995). Also in contrast to the example at the beginning of this review, OTPAD presents hardly any cases of typical development, not even for purposes of comparison with its own numerous atypical cases. Finally, although the authors excel in their command of OT theorising, they are (in this reviewer's view) excessively ambitious in being unable to resist involving the latest, and hence underassessed, theoretical proposals (e.g. McCarthy Reference McCarthy2007), thus threatening to leave readers less familiar with the breakneck speed of development in OT with the impression that their efforts at teasing out the often intricate details of dozens of tableaus (and counting) were quite possibly in vain, as these might already be out of date. In the end, however, these weaknesses pale into insignificance compared to OTPAD's strengths. Taking note of this publication is strongly recommended to phonologists and non-phonologists, to acquisitionists and non-acquisitionists, and to OT-practitioners and non-OT-practitioners alike.