Research on child bilingual development has experienced a renewed impetus in recent years, exploring aspects of early bilingualism and second-language learning that have called into question widely held assumptions. For a while it seemed that work on transfer/interference, unequal development and dominant language-induced attrition had come to be disfavored, undermining an ideal standard of a completely balanced bilingual with equivalent competence in each language. In their book, Yip & Matthews' extensive report of their work addresses difficult issues that practitioners and specialists have tended to shy away from.

The book takes up what is one of the central objectives of current research today on early bilingual development: the construction of a model that can account for both simultaneous balanced bilingual development (of two first languages: 2L1 development) and imbalanced bilingual development, in its different expressions and ultimate attainment outcomes, for example, delayed development of one language, first language competence plus an emerging second-language-type competence or attrition of a non-dominant former-L1. The first two chapters chart out the theoretical issues. Chapter 3 provides an overview of the methodological problems of naturalistic child language study, Chapters 4 to 7 report the authors' findings from their study of a group of young bilingual children and Chapters 8 and 9 conclude with some interesting speculation about the how a Language Faculty internal process in children might help explain external sociolinguistic phenomena in language contact-induced change. Overall, the review of the literature and discussion of findings contribute a good measure of clarity to a number of questions related to Cross-Linguistic Influence (CLI) in child bilingualism, which are still not well understood to this day.

The Introduction and Theoretical framework present a comprehensive research agenda that identifies some of the most important problems still pending in the field. The actual findings directly bear on only part of it, but readers get a clear view from the first two chapters of the bigger picture into which this study of asymmetry in bilingual development is inscribed. The overarching conceptual backdrop to the study is the Poverty of Stimulus (PoS) problem, specifically in how it presents itself in the acquisition of two languages. Special test cases of PoS potentially provide the strongest evidence for how young children upgrade the language input available to them, for example, creole emergence from Nicaraguan sign pidgin (Senghas, Sotaro & Ozyürek, Reference Senghas, Sotaro and Ozyürek2004). Early bilingualism, while not ‘special’ because it is so widespread, and even typical, is another example where children overcome insufficient and conflicting evidence in the available input. Yip & Matthews refer to this as the Logical Problem of Bilingual Acquisition and the poverty of dual stimulus. The arguments turn out to be quite convincing (pp. 5–6, 30–3); the authors discuss how Primary Linguistic Data made available from two grammatical systems, instead of just one, presents children with a more formidable learnability problem, one that they nevertheless surmount without recourse to additional or special learning resources.

Following a review of the research on early separation of the linguistic systems, an apparent contradiction is posed, one that has seemingly troubled some researchers in the field. If the grammars of the young preschool bilingual show a precocious differentiation into autonomous linguistic systems, how can we account for prolific cross-linguistic influence, interference/transfer and asymmetry? As it turns out, the facts of dominant/non-dominant imbalance, even massive transfer in one direction, do not call into question the findings of early separation and autonomy of each developing linguistic system. In fact, a close examination, in Chapters 4 to 7, of transfer and CLI between Cantonese and English showed them to be selective and systematic, and largely predictable from independent assessments of dominance. This finding provides additional evidence that disconfirms the theory of a unitary and holistic (i.e. non-modular) mental distribution of bilingual children's languages. Non-random and systematic transfer indicates that the knowledge of language in bilinguals is organized into systems.

The details of the evidence in Chapters 4 to 7 are well worth plowing through, presented, as they are, with clear examples and non-technical easy-to-follow explanations. We get the beginnings of a picture here of the complexity of the actual mechanisms of cross-language interaction. Three aspects of the early English grammar of the bilingual children studied that are affected by CLI are presented for analysis: the emergence of wh-in-situ constructions in questions, null objects and prenominal relative clauses. What is of interest in these chapters is how the accounts of transfer from Cantonese hold up a mirror to problems of development in all child bilingualism.

For example, while English interrogatives typically involve movement of the wh-word, in Cantonese it remains wherever the expression replaced by the wh-word occurred, as in English ‘echo-questions’ (‘You spent HOW MUCH at the mall this morning?’). The bilingual children evidently appropriated this feature far more extensively than the monolinguals, and maintained it longer. For now, we can put aside the questions of to what extent the non-target wh-in-situ development is transfer-induced or developmental, and whether it reflects competence or performance. Rather, subject Timmy's transitional construction (at 3;02) highlights a more difficult and more interesting problem for researchers: how bilingual children process input ambiguity (1):

1. (1) What are you cutting? You want to cut the what? (p. 100)

Evidence from the dominant language (Cantonese) prompts the bilingual to apply wh-in-situ across the board, and evidence from English is divided. The relevant distinctions are subtle, belonging to the domain of interface between syntactic structure and semantics/pragmatics. As the authors point out, the learnability problem is acute. How do young bilinguals overcome the poverty of dual stimulus problem, potentially compounded by input from competing input models that don't even present a clear-cut contrast? In one sense, the problem of how children master these distinctions overshadows the discussion about how early separation of the language systems is achieved. There may actually be a ‘delay’ in effecting complete separation (e.g. in cases of dominant/non-dominant bilingualism). But its consummation by age four or five poses the same unanswered research questions as evidence of earlier separation does, given the continued unavailability of higher-order meta-level learning strategies.

A similar instance of the Logical Problem of Bilingual Acquisition is presented in Chapter 5: the persistence of null objects. This is another syntax–semantics/pragmatics interface domain issue; the input ambiguity problem is of the same order. Cantonese permits object omission widely; in English the question of which transitive verbs require the object to be expressed, and in which contexts, is highly complex, as in (2) (from Goldberg, Reference Goldberg2001):

1. (2) Pat gave and gave, but Chris just took and took.

In actual language use, unexpected null object constructions in English could be said to perhaps violate expectations of coherent discourse, being grammatical in other respects.

At this point, after studying the examples of transfer and how the bilingual children began to recover from its influence, we might want to make the following proposal regarding the bilingual input ambiguity learning problem: part of the solution must include the early separation of grammars into autonomous representations for each language. With basically Primary Linguistic Data to work with, preschool-age bilinguals overcome, sooner or later, acutely ambiguous evidence spontaneously in the same measure of completeness as monolingual children. Such an automatic and untutored resolution of such intra- and interlinguistic distinctions would be hard to explain unless one's working model accepts one form or another of early differentiation between language systems. This suggestion by the authors prompts us to consider an important research question: Would evidence of even sharper imbalances and more prolific cross-linguistic interaction between dominant and non-dominant languages be compatible with early separation?

When considering the transfer of prenominal relative clauses, in Chapter 6, we might assume that the issue of input ambiguity doesn't apply. Patterns like (3) would be rare among English-speaking children, monolingual and bilingual:

1. (3) Where's [_NP[_Syou buy] that one] (Timmy 2;07: p. 164)

The authors, however, hold out, even here, the possibility that the contrast with Cantonese may not be as straightforward as first appears, arguing that you buy is a relative clause following the Canonese prenominal pattern (p. 164): Where's that one [that] you buy>Where's you buy that one. In English, adjectives and other modifiers also precede nouns, as do the relative clauses in the English of these Cantonese-dominant children.

Another example involves an episode in which uncle Pet-Pet (aka Patrick) has bought a videotape for Timothy:

1. (4) I want to watch videotape. Butterfly. [Patrick buy that one]

2. (5) I want [Pet-Pet buy that one videotape]. (2;11: p. 164)

According to the authors, Patrick buy that one in (4) should not be analyzed as a main clause, but rather as a relative clause: the one that Patrick bought. Along the same lines, (5) should not be taken as I want Pet-Pet to buy that videotape because Timothy knows that it has already been bought. In any case, an interesting research question would be to compare bilingual children's recovery from different sets of incorrect hypotheses; would they be more difficult for null objects and wh-in-situ?

We can now follow the logic of Yip & Matthews' claims and apply them to scenarios in which dominance and imbalance are permanent and progressive, as in L1 attrition (an unfortunate misnomer) and early stabilization of a bilingual's ‘weaker’ language; see Paradis (Reference Paradis2004) for one explanation for how this is possible. It appears that the subjects in this study, three of whom are the authors' own children, did not fall into the latter category of early stabilization or attrition of English. This raises the interesting possibility of whether the ‘temporary’ imbalances at issue could be attributed to either: (1) one, more rapidly, developing system inhibiting the other (mechanisms of performance/processing); or (2) an actual uneven development of competence. Interested readers should consult Yukawa (Reference Yukawa1997) for a discussion of some pertinent considerations, and Meisel (Reference Meisel2007) on the question of distinguishing between aspects of performance and competence in these situations.

Yip & Matthews argue that the theoretically more interesting aspect of imbalanced bilingual development is underlying competence; how the actual knowledge structures of each language system interact (scenario #2 above). Saying that, presumably, does not mean that factors strictly limited to performance and processing cannot also account for emerging imbalances between Language-A and Language-B in young children (or even that a ‘processing inhibition’ might appear as a precursor to loss of competence). In any case, the most important part of their argument seems to be that it is a mistake to limit explanations of dominance and asymmetry to aspects of performance.

Chapter 8 considers Thomason's (2001) idea of how early bilingualism plays a role in what is called ‘language contact-induced change’. Table 1.2, in the Introduction, summarizes the relationship: how Language Faculty internal developing imbalances can help explain sociolinguistic changes in a community of speakers. Again, but with a different twist, the Logical Problem of Bilingual Acquisition is part of the explanation, in this instance, for rapid language shift. Just as the (undelayed) development of grammatical knowledge in two languages is underdetermined by the input available to the child, the delayed development of a non-dominant language may be attributed, in part, to asymmetries in the external conditions of language use, but its course is not determined exhaustively by these conditions. That is, the delayed development of a child's ‘weaker’ language may be related to restricted input or diminished exposure to that language, but such experiential factors do not account for or explain all cases of normal delay and attrition. In such manner, first-language ‘attrition’ (now in quotes) can be studied developmentally. An immediate practical application of this approach would be to help us understand why some accounts of contact-induced change cannot be correct: for example that language shift is caused by a diminished or incomplete form of a socially disfavored language being ‘passed on’ from one non-balanced bilingual generation to the next, or that code-switching and non-reciprocal borrowing from the socially preferred/prestigious language directly contributes to degrading the subordinate language. The language acquisition mechanisms, fully accessible during early childhood, have been shown to be able to upgrade the most rudimentary pidgin system (Goldin-Meadow, Reference Goldin-Meadow2005). Therefore, an important part of the explanation for dominant/non-dominant divergence, early stabilization and attrition must be sought in the internal dynamics of cross-language interaction. This study makes an important contribution to understanding them better.

I would like to offer one last observation. Most readers will notice that the authors frame their study with the basic assumptions of Universal Grammar in mind. But at the same time, they recognize the contributions made by usage-based and functionalist approaches to our understanding of bilingualism, and incorporate them into their analyses throughout the book. The authors don't explicitly call attention to the potential dialogue. They drop a hint, though, in the concluding chapter as to why it is that in the study of bilingualism such a discussion has a better chance of sprouting legs. The reason for this possibility is conjunctural, but worth pursing nonetheless: that research on bilingualism must attend to a greater diversity of Language Faculty internal and general cognitive domains and mechanisms than is the case in monolingual development, thus allowing for a broader space of potential common ground. This initiative on their part, from within the field of bilingualism, happens to coincide with, or rather be a part of, a wider initiative within cognitive science to open new lines of discussion between generativists and researchers working on problems in language development from other perspectives.