Researchers and students of applied linguistics and anyone interested in bilingualism and second language learning will find this nontechnical and accessible state-of-the-art survey useful. Regardless of which neurolinguistic theory of bilingualism ends up holding sway, it is in the central nervous system that the core mental operations that underlie language use are materially computed. So, as we do our work, it is good to keep an eye on what's going on here.

Chap. 1, “Components of verbal communication,” starts off by setting out the general approach to the study to bilingualism in this book: the assumption that knowledge of language and our ability to use it can be studied analytically (not to deny that holistic approaches can also provide for useful points of view). Chap. 1 sets the stage for a version of modularity informed by findings from the author's branch of cognitive science. The first of a number of component/subsystem breakdowns that we are asked to consider is the distinction between two kinds of knowledge of language: (i) the internalized implicit system, acquired incidentally and not available to conscious awareness, and (ii) explicit knowledge and metalinguistic awareness. The first is primary and domain-specific, and consists of computational procedures unavailable to introspection – for example, the mental grammar structures of morphology and syntax. The second, more integrative and “flexible,” would correspond, for example, to the declarative component of vocabulary, “secondary” in the sense that this is the kind of knowledge that we learn.

Chap. 2, “Implicit and explicit language processes,” is where applied linguists, especially educators, will find the most interesting material to reflect upon. Picking up on the discussion from the previous chapter, Paradis offers another way to understand an ongoing debate in the area of second language (L2) learning. How learners might deliberately analyze a grammatical pattern is different from the way in which it will be “internalized” – that is, how it becomes an integral part of implicit linguistic competence. The underlying structure of this knowledge of language is “nowhere to be noticed” (p. 37). The language learner might be able to gain awareness of what is produced (in the input and output of performance), but not how. And being a matter of two different kinds of mental representation, explicit, declarative metaknowledge cannot be converted into procedural knowledge. Acquisition, then, is not a matter of “proceduralizing” explicit rules that were arrived at by means of analysis, inductive learning, memorization, or reflection; rather, what happens is a gradual shift from relying on metalinguistic knowledge to using implicit knowledge. Readers familiar with the various models of L2 learning will recognize this as coinciding, more or less, with the “non-interface” hypothesis, which draws a sharp line between “learning” and “acquisition.” In its strongest version, it discards the usefulness of direct instruction, focus on grammatical form, metalinguistic strategies, corrective feedback, active monitoring and controlled practice. Paradis, however, leaves the door open to a number of different ways of conceptualizing how things might actually turn out in the modular “non-interface” (not a good metaphor) relationship between learned declarative knowledge and underlying implicit competence. A close reading of the two final sections devoted to this topic (45–53) should make both opponents and proponents of instructed L2 acquisition pause before assigning the author's modular theory to one camp or the other. For example, to say that formal instruction and explicit learning make an indirect or secondary contribution does not forcibly imply that these general learning factors are not necessary or strongly facilitative for second language development. First language (L1) acquisition, though, would be a different matter altogether.

Chap. 3, “Bilingual aphasia,” makes a strong case for another dimension of neurofunctional modularity, based on a comprehensive review of the literature on impairment of language ability. Bilingualism provides for a unique and privileged vantage point when we consider recovery patterns. Recovery is not always parallel (same pattern of proficiency and dominance before and after trauma); the various kinds of differential recovery (some of them at first seem strange and implausible) require a working hypothesis that assumes separate representations, of some kind and on some level, for each language subsystem. But since any explanation has to take into account both parallel recovery patterns and the entire variety of differential recovery, the idea of a localization of each language at a gross anatomical level would have to be set aside. (By analogy, different combinations of scratches or perforations to the hard drive of a computer would result in all sorts of compromises to a given network of information depending on how it is fragmented, distributed, and interconnected.) The leading hypothesis, then, is that first language and second language are subserved by separate and autonomous neural circuits interwoven into the same larger language areas, and that physiological inhibition is the common mechanism that can account for all of the findings.

Chap. 4, “Cerebral lateralization and localization,” follows up on chap. 3 and gets into details that are well worthwhile plowing through. Figure 4.1, “Schematic representation of various models of the organization of languages,” lays out the logical possibilities, as well as Paradis's Subsystems Hypothesis.

Chap. 5, “Neurofunctional modularity,” brings together the two major dimensions of how “knowledge of language” (quotation marks are the reviewer's) is componential: (i) the two different kinds of knowledge (explicit/declarative and implicit/procedural), and (ii) the separation of the L1 and L2 subsystems. Along another dimension, we need to reckon with the internal diversity of language, in the bilingual mind now differentiated into (instantiated by) separate networks/circuits. For example, “The morphosyntax module contains as many subsystems as the person speaks languages” (130). In the end, however the different vertical and horizontal dimensions, layers, and successive subsystems may come to be parceled out, the general idea is that knowledge components are isolable and richly interconnected. To what degree and how they are will be settled empirically, and not anytime soon. In the meanwhile, there is plenty of room still for a number of competing and potentially compatible proposals for further research.

Chap. 6, “Neuroimaging studies of the bilingual brain,” is an overview of central methodological questions and an assessment of where things stand in this brave new science, which is still largely at the “poking stage” (186).

Chap. 7, “An integrated neurolinguistic perspective on bilingualism,” features the graphic presentation of the Three-Store Hypothesis (Figure 7.2). Two related models of the architecture of bilingual proficiency to which interested readers should refer as they study this concluding series of proposals by the author are Paivio's (1991) Bilingual Dual Coding Model and Cummins's (1991) Linguistic Interdependence Model. Despite important differences, all three try to capture key features of the two dimensions of modularity discussed so far. The objective here is to account for the fact that some aspects of language proficiency in bilinguals appear to be distributed between separate linguistic domains (autonomous subsystems comprised of two lexicons and separate subsystems for the respective morphosyntactic and phonological structures), while other aspects of proficiency appear to be “shared” (there being access to a common nonlinguistic conceptual domain). This chapter covers so much ground that it deserves two readings; it touches on a range of big questions, from a new way of thinking about the Sapir-Whorf Hypothesis to explaining in a psycholinguistically coherent way the phenomenon of borrowing and interference in L2 learning.

What captures one's attention as an overall theme, perhaps unintended by the author, is the possibility of a more open cross-disciplinary discussion, even on some proposals that appear at first to be rather controversial. In a recent Language in Society review by Burling (2003) of Jackendoff's Foundations of language, there is a similar probe: an invitation to consider new developments in cognitive science that are redrawing the old lines of debate. Foundations of language makes a strong case for this possibility and should be studied together with A neurological theory of bilingualism for a different perspective on some concepts that are often dismissed as incompatible with, or not relevant to, sociolinguistic or anthropological approaches to problems of language use in society.