The chapters in this volume represent a very important set of contributions to the recent debates regarding the nature of recursion in natural language and the typological issues surrounding its (non-)manifestation in the syntax of various languages. In many of the chapters, important new data from the indigenous languages of Brazil are also presented.Footnote ¹

Although the issues at stake here are long-standing ones in linguistic theory, the immediate stimulus to the conference comes from two papers published in the early 2000s: Hauser, Chomsky and Fitch (Reference Hauser, Chomsky and Fitch2002) and Everett (Reference Everett2005).

Hauser, Chomsky and Fitch (Reference Hauser, Chomsky and Fitch2002) made an important conceptual distinction between the Faculty of Language in the broad sense (FLB) and the Faculty of Language in the narrow sense (FLN). The former includes both the narrow-syntactic component that generates structural descriptions of sentences along with the two principal interfaces (Articulatory-Perceptual, governing the transduction of syntactic representations ultimately into a perceptual modality, speech or sign, and Conceptual-Intentional, converting syntactic representations into objects of thought, belief or judgement). It was suggested that aspects of FLB might not be specific to humans. FLN, on the other hand, was argued to consist purely of the computational system of the syntax, whose central property is discrete infinity: the ability to generate an infinite set of structural descriptions from the iterated application of the single structure-building operation Merge. Merge, as a property of finite human minds, must itself be finitely specifiable. Formulating Merge as a recursive function, able to apply to its own output in iterative fashion with no limit in principle, makes possible the generation of infinite sets from finite means. In slightly more formal terms, Merge can be seen as the intensional definition of a set of structural descriptions, whose extension is infinite (see Watumull et al. Reference Watumull, Hauser, Roberts and Horstein2014 for more detailed discussion of this aspect of Merge). Hauser, Chomsky and Fitch thus argued that the central property of FLN was a recursive operation. They further suggested that FLN, and hence Merge, is a uniquely human trait, and one which is likely to have evolved very recently in human phylogeny, perhaps through exaptation from some other aspect of cognitive or motor function.

The obvious inference to make from Hauser, Chomsky and Fitch’s conclusions is that recursion is a property of all, and only, humans. Therefore, evidence of recursive structures of one kind or another should be available in all human languages; in fact, recursion, as part of FLN, forms part of the definition of a possible human language. This view was directly challenged by Everett (Reference Everett2005), who argued that Pirahã, an indigenous language isolate spoken in Amazonas, Brazil, lacks evidence for what is often seen as the clearest form of syntactic recursion, namely sentential embedding. If this conclusion is correct, then the view of FLN espoused by Hauser, Chomsky and Fitch may be challenged (although there are reasons to question this conclusion, as we will see below). More generally, since Hauser, Chomsky and Fitch were articulating a particular version of the general programme of generative grammar as formulated originally by Chomsky (1955/Reference Chomsky1975, Reference Chomsky1957), Everett’s conclusions may be seen as a challenge to the entire enterprise of generative grammar as it has been conceived since the 1950s; this is certainly how Everett himself sees them.Footnote ²

Unsurprisingly, Everett’s conclusions have been controversial; for extensive discussion of the nature of the syntactic evidence for and against recursion, and how this may or should be interpreted, see Nevins, Pesetsky and Rodrigues (Reference Nevins, Pesetsky and Rodrigues2009a, Reference Nevins, Pesetsky and Rodrigues2009b), Everett (Reference Everett2009) and Sauerland (Reference Sauerland2010b). Everett’s views have also received considerable media attention, in the form of two popular books by Everett himself (Everett Reference Everett2008, Reference Everett2012), one film, articles in The New Yorker and elsewhere, and regular appearances by Everett in the media.Footnote ³

The central question which the chapters collected here address is then: what is the evidence from the indigenous languages of Brazil and elsewhere for and against recursive structures in natural language? On the simplest interpretation of what is at stake here, one could think that if such evidence is not directly forthcoming, then it is right to conclude, as Everett and others (notably for example, Evans and Levinson Reference Evans and Levinson2009) have, that the Chomskyan programme for linguistic theory is so fundamentally flawed that it must be abandoned. If, on the other hand, irrefutable evidence for such structures is available, then Everett’s challenge can be deemed to have failed, and Chomsky’s programme is thereby supported. The chapters in this volume certainly tend to favour the latter conclusion. Moreover, such a conclusion would in any case be naive. One could argue, as Chomsky himself has done (Chomsky Reference Chomsky2012:30), that the absence of recursion in Pirahã proves nothing about the overall nature of the human language faculty, any more than the discovery of a group of humans who do not walk upright would disprove our innate capacity for bipedalism. Alternatively, one could argue that any linguistic structure containing more than two elements must feature binary Merge and is hence recursive, and so Everett’s observations regarding Pirahã syntax are beside the point.

What is at stake in these debates is more than either the question of the correct analysis of various syntactic structures in a range of languages (from Brazil or otherwise, Indo-European or otherwise, “exotic” or otherwise), or the correctness of an influential theory of language. These debates go deeper: they directly address the question of what it is to be human. The capacity for the acquisition of complex language under naturalistic conditions without explicit instruction is universal to, and unique to, human children. The human capacity for language underpins human culture, civilisation and technology. Therefore, our view of the essential nature of language profoundly informs our view of human nature, the human mind and human culture.

For Chomsky, as we saw from the brief summary of Hauser, Chomsky and Fitch (Reference Hauser, Chomsky and Fitch2002) above, the central property of language is the fact that sound and meaning (the two interfaces implicated in the FLB) can be related over an unbounded domain; Berwick and Chomsky (Reference Berwick and Chomsky2015:1) refer to this as the Basic Property of human language. This is possible because the two interfaces are mediated by the syntactic component whose central formal property is Merge. Recursion lies at the very heart of the definition of the language faculty (broad or narrow). It is the cognitive capacity to manipulate symbols in a recursive fashion that is central to human nature. To the extent that this ability is not shared with other species, it must be somehow instantiated in the human genome, such that the genetic blueprint for building a human brain contains an “instruction” to create the neural substrate for such representations (in our current state of ignorance, we have no more idea as to how these representations are neutrally instantiated than we do of how any “higher” cognitive functions are). The ability to manipulate such structures emerges spontaneously in human development, as long as (and perhaps as soon as) a child is exposed to language. Somehow, at some stage since the human lineage diverged from that of the most closely related primates, this cognitive capacity must have evolved (see Berwick and Chomsky 2015 for a recent discussion of language evolution in the light of the Basic Property).

But there is more, at least arguably. The cognitive capacity to manipulate symbols using a recursive schema such as Merge may underlie other human abilities: our ability to manipulate numbers (Merge is formally very close to the successor function, the recursive function S such that S(n) = n+1 for every natural number n), our musical capacities (Lehrdahl and Jackendoff Reference Lehrdahl and Jackendoff1983), our moral sense (Hauser Reference Hauser2006) and our capacity to recognise and ascribe content to other minds (i.e. Theory of Mind, deVilliers Reference Villiers2007). Thus much of what many would agree makes us human may be traceable to a simple formal property of human mental computation. There is, moreover, a still deeper point at stake here, one with its origins in Cartesian philosophy: the recursive nature of syntax is a necessary component of what Chomsky has called the “creative aspect of language use”, i.e. the fact that humans are able to produce and understand utterances that have never been produced before. This formal property quite literally allows us to give expression to our freedom of will. So the postulation of the ability to produce recursive cognitive representations elegantly captures profound aspects of human uniqueness.

Everett’s view, on the other hand, takes culture as the central concept in understanding human language and human nature. Culture, rather than computation, is the key to the understanding of human nature (although Everett does not deny that humans are capable of recursive cognition; he merely asserts that it is not central to natural-language syntax). In order to understand the nature of human cognition, we must understand human culture and cultural evolution. Language is, as the title of his 2012 book implies, a cultural tool: something that, like other tools, humans have invented; something that may vary greatly from culture to culture, and that has developed through cultural, rather than biological, evolution. The fundamental nature of language is determined by society, rather than by any property of the individual. Hence, since Pirahã culture differs profoundly from “Western” culture, it is no surprise that the Pirahã language should also differ profoundly from (Indo)-European languages, structurally and in many other ways too (see in particular Everett Reference Everett2005 for details on several strikingly “exotic” aspects of Pirahã).

These two views arguably reflect two rather different historical currents in linguistics (they are obviously also connected in a very general way with rationalist as opposed to empiricist views of epistemology). We can, in a rather superficial but nonetheless useful way, discern two distinct traditions in the linguistics of the past centuries. On the one hand, there is the “comparative/historical” approach, and on the other the “formal/universalist” approach.

The comparative/historical approach dominated nineteenth-century linguistics, leading to the establishment of many of the major language families of the world through painstaking empirical work and the development of the informal methodology of comparative reconstruction. It has its roots in the orientalism of Sir William Jones, Friedrich Schlegel (see Schlegel Reference Schlegel1808), as well as in medievalism (see the work by the Grimm brothers on aspects of medieval German language and folklore: Grimm and Grimm Reference Grimm and Grimm1812–1815), and, more generally, in German Romanticism. This approach to language is particularist, in that details of individual languages are focussed on without regard to issues concerning language universals. The emphasis is empirical and historical, and questions such as the nature of the relation between language and logic play little or no role. Instead, the emphasis is by and large on cultural and historical explanation. Language is, if anything, the mirror of society and history, rather than the mirror of the mind. In addition to Jones, Schlegel and Jakob Grimm, this approach to linguistics is epitomised by Bopp (Reference Bopp1816), and was in many respects inherited by Saussure (Reference Saussure1916) and Bloomfield (Reference Bloomfield1933) (although both of the latter authors made interesting comments on universals; see Roberts (Reference Roberts and Roberts2007); moreover, starting with Osthoff and Brugmann (Reference Osthoff and Brugmann1878), a more formal approach to comparative reconstruction based on the regularity of sound change emerged). It is also to a large extent the approach to linguistics that characterises modern language typology from Greenberg (Reference Greenberg and Greenberg1963) onwards (especially more recent typological work which has moved away from the postulation of implicational universals of the Greenbergian kind in favour of diachronic and areal accounts for similarities across languages; see for example Bickel Reference Bickel2007).

On the other hand, the “formalist/universalist” tradition has its modern origins in seventeenth-century rationalist philosophy (its ancient origins lie in Stoic and Platonic philosophy), in particular in the work of the Cartesian Port-Royal grammarians (see Arnauld and Lancelot Reference Arnauld and Lancelot1660). It is primarily concerned with the search for universal features of language. There is a related concern for the relation of grammar to logic, and, more generally, for the connection between the laws of language and the laws of thought. There is a concomitant emphasis on formalisation for precision and clarity. The first attempts in the modern era to formalise thought and language originate with Dalgarno (see Cram and Maat Reference Cram and Maat2001), John Wilkins (Wilkins Reference Wilkins1668) and, most intriguingly, Leibniz (see Leibniz Reference Leibniz1666; Watumull and Roberts Reference Watumull and Roberts2014). In the mid-nineteenth century, Boole attempted to formalise thought (Boole Reference Boole1854), with the great breakthroughs in modern logic coming half a century later (Frege Reference Frege1892; Russell Reference Russell1905). Through a well-known historical path (see Tomalin Reference Tomalin2003 for details), Russell and Whitehead’s (Reference Russell and Whitehead1910–1912) attempt to formalise arithmetic on logicist principles eventually led to the development of recursive-function theory in the 1930s, and this led directly to Chomsky’s formalisation of natural-language syntax (Chomsky Reference Chomsky1955/1975).

It would seem obvious, even on the basis of this superficial presentation, that Chomsky epitomises the “formalist/universalist” approach, while Everett may represent (perhaps a rather extreme version of) the “comparative/historical” approach. But this is not really my point here, and again I leave the readers to their own conclusions on this (and, indeed, on the validity of the distinction I have tried to make). What I would like to suggest, instead, is that the chapters in this volume, with their great attention to the empirical detail of certain constructions in the indigenous languages of Brazil, their concentration on psycholinguistic experimentation, and/or the emphasis on the detailed study of aspects of child language, in fact represent a consilience of these two strands in linguistics. What we see in the chapters that follow is a set of highly empirical studies of linguistic phenomena in the service of attempting to resolve a profound question about the nature of human language and thought. The two historical strands converge. Most strikingly in the discussions of the indigenous languages of Brazil, we see very clearly how close attention to fine empirical detail and a clear, precise, formal sense of the overarching theoretical questions fundamentally inform one another. This book is testimony to what, when the right questions are posed and the answers are carefully and intelligently sought, modern linguistic theory can achieve. At its best, as here, modern linguistic theory is the true heir to both traditions.

Ian Roberts