For all linguists who do not view language as a Platonic object, it should be uncontroversial that there is a “fit” between language and the brain (as expressed in Chomsky [Reference Chomsky1986], where Chomsky introduces his notion of I-language). The common goal is to find out how they fit, and to what extent our cognitive system is dedicated to language.

Christiansen & Chater (C&C) argue that no part is dedicated to language. In brief, that there is no Universal Grammar (UG). This is a strong position that deserves extensive research, taking into account the properties we know language possesses (the derivation of condition B, in Reuland Reference Reuland and Müller2005b, illustrates what this involves). How well, then, do the authors succeed?

C&C focus on what they call “abstract” properties (see sect. 2) and discuss two such properties to establish initial plausibility. One is subjacency (Chomsky Reference Chomsky1981). C&C argue that subjacency is not specific to language, but derivable from general processing constraints (sect. 6.3) or, possibly, “cognitive limitations on sequential learning” (sect. 7.2), referring to Berwick and Weinberg (Reference Berwick and Weinberg1984) and Ellefson and Christiansen (Reference Ellefson, Christiansen, Gleitman and Joshi2000) in this regard. Subjacency effects are real. But we know by now that restrictions on extraction result from a variety of factors (see Szabolcsi [Reference Szabolcsi, Everaert and Riemsdijk2005] for a state-of-the-art overview; and see, e.g., Khomitsevich [Reference Khomitsevich2008] for cross-linguistic variation in island effects). Subjacency is not a primitive. The nature of the resources for language processing is still open, and more may be involved than just general working memory (see Reuland [forthcoming], elaborating on Ullman [Reference Ullman2004] and Baddeley [Reference Baddeley2007]). Even granted C&C's conception, the experiment in Ellefson and Christiansen (Reference Ellefson, Christiansen, Gleitman and Joshi2000) does not bear on subjacency: nothing forces a forbidden extraction. The other issue C&C bring up is binding (discussed further on).

The main line of the article takes a different tack: The assumption of an innate UG leads to a logical problem of language evolution. C&C argue that the emergence of a dedicated language system/UG can neither be explained by adaptationist (Pinker & Bloom Reference Pinker and Bloom1990), nor by non-adaptationist mechanisms. Hence, they conclude, there must be a third way: no UG and “language is shaped by the brain” (emphasis added).

Their discussion gives the paradoxical impression that C&C, though cognitive scientists, espouse a Platonistic view of language: an object external to the mind, which evolves and adapts to the latter; a view I characterize as neo-adaptationism. Note that C&C just assume that “functional” principles are domain-general, whereas only a moment's reflection shows that there is nothing “natural” in the way salience, topicality, and so on, are encoded in natural language (Slioussar Reference Slioussar2007). Isn't there a great rhetorical tradition in which what is most important comes last?

C&C reject the non-adaptationist position, positing that non-adaptationists postulate overly rich innate structures. But, in view of the Minimalist Program (Chomsky Reference Chomsky1995) and the approach to evolution in Hauser et al. (Reference Hauser, Chomsky and Fitch2002), this is incorrect. No more is required than a finite set of vocabulary items (lexical and functional), a combinatory operation of (external and internal) Merge plus associated interpretive operations, a checking operation (comparing the feature contents of elements), and locality conditions in operations. These show up in the form of restrictions on “attractors” for movement (“phase heads” in current terminology) and in the form of restrictions on the domain where attractors can find their targets. All this is close to what C&C would have to assume anyway (also construction grammar does not allow dispensing with combinatory principles, unless to the penalty of incoherence).

What is problematic in C&C's presentation is that (1) they do not make clear what language is, and – not surprisingly, given (1) – that (2) they don't discuss how language is encoded; and (3) that, although they claim that “linguistic adaptation allows for the evolution of increasingly expressive languages” (sect. 5, para. 1), C&C say nothing about what it means for one language-stage to be more expressive than another, and what discontinuities must be involved.

For reasons of space, I focus on signs here and argue that even their evolution involves more “UG” than meets the eye. I start with C&C's characterization of the initial state, proto-language, as a “cultural product likely to be highly variable over both time and geographical locations” (sect. 2, para. 7). But even so, language must be a product of the mind, not an external object that emerged by miracle.

Such a proto-language minimally stands for a collection of Saussurean signs, conventional pairings <f, i>, where f is a form in a medium (sound, gesture) and i its interpretation as a concept (see Saussure Reference de Saussure, Bally and Sechehaye1916). The term “cultural” glosses over a crucial issue. The pairings themselves may be culturally determined, but the fact they are possible cannot be. Even a neo-adaptationist must assume an evolutionary discontinuity here. The same applies to arbitrariness of pairings. Arbitrariness is a yes-no property. It can only emerge by a discontinuity. By its very nature, it must be dedicated to language, hence reflect UG as commonly understood. So, even on their own terms, C&C are committed to a minimal UG.

But more is at issue here. C&C leave open whether their proto-language allows free combination of signs. However, free combinability – recursion – must develop at some point. Recursion minimally requires concatenation, combining two objects into an object of the same type, which is in turn available for concatenation. This entails that linguistic signs must be triples of the general form in (1) shown below, rather than pairs.

1. 1 <f, g, i> with f and i as explained above and g a formal instruction driving the computation.

The addition of g, as a formal instruction representing combinability, leads us beyond the Saussurean sign. It is a minimal change, but qualitative in nature. Adding this slot to the sign frees the way for (varying) grammatical operations. Free combinability requires a concomitant interpretive operation, determining the interpretation of a complex element in terms of the interpretation of its parts.

A different change is needed for signs as abstract mental objects, as in Example (2) shown below:

1. 2. <c _f, g, c _i> where c _f and c _i are “formal” mental encodings of instructions for the language-external systems of realization and interpretation.

This feature is needed for planning, embedded recursion, and handling dependencies (see Reuland Reference Reuland2005a; in press). Such items are not simple signs anymore, but instructions to form signs. Forms and concepts may reflect culture, but the ability to form such pairings does not. This ability involves, again, a qualitative change, incompatible with neo-adaptationism.

Human language has many further characteristic properties. Take Example (3), shown below, where he cannot be valued as no one, since no one cannot denote an individual.

1. 3. No one has a gun. *Will he shoot? (the asterisk here indicating ill-formedness)

But he can depend for its interpretation on no one in Example (4):

1. 4. No one was convinced that he would be welcome.

Example (4) exhibits binding, that is, interpreting an expression in terms of another expression. Again, this is a yes-no property, not subject to gradation. Binding is a core feature of language as we know it (see Reuland, under contract). Binding transcends culture and pragmatics. Its emergence involves a discontinuity that has to be encoded in the system itself, and hence, reflects UG. Such discontinuities are a challenge for any neo-adaptationist approach, but fascinating to investigate nonetheless.