Introduction to Part IV: Launching language: the development of a linguistic species
Abstract and Keywords
This article deals with the different views of researchers on the central properties to be accounted for in language evolution. Stephen Anderson presents certain structural regularities become established in the world's languages, including universal properties such as structure dependence. Anderson argues that there is no need to assume a dichotomy between a genetically determined language faculty and a language faculty shaped by external factors, such as functional pressures and the effects of grammaticalization. The language faculty supports the learning of specific kinds of linguistic systems, and it would not be at all surprising if natural selection favored those who were able to acquire language most efficiently. Grammars that are most easily learned will be the ones that are acquired, because generations of better learners also shaped grammars to be more learnable. James Hurford, Michael Corballis, Stevan Harnad, Terrence Deacon, and Robbins Burling investigate what cognitive capacities must have evolved before the evolution of any kind of language. These capacities include the development of meaning (semantics and pragmatics), the origins of grounded symbols, and the ability to learn and store words. Hurford argues that other animals possess at least proto-conceptual categories, which form the basis for conceptual meaning. Animals exhibit planning abilities, have mental representations of territory, and can make calculations based on their knowledge, such as transitive inference.
In Part IV, we present the most immediately ‘linguistic’ chapters, dealing with the central properties to be accounted for in language evolution. The first chapter (39), by Stephen Anderson, asks why certain structural regularities become established in the world's languages, including universal properties such as structure dependence. Some have argued that languages are the way they are solely because the same functional pressures have shaped them (see Chater and Christiansen, Chapter 65). Other theorists see the evidence as necessitating a rich, species‐specific cognitive capacity for language, though the precise contents of this are less certain. If the (p. 354) language faculty is genetically fixed, its contents must presumably be (largely) adaptive. But this is highly implausible, as both Bickerton (Chapter 49) and Lightfoot (Chapter 31) also argue: how could very specific aspects of syntax, for example, enhance an individual's fitness as the language faculty was evolving?
Anderson argues that there is no need to assume a dichotomy between a genetically determined language faculty and a language faculty shaped by external factors, such as functional pressures and the effects of grammaticalization. These are not alternative explanations for observed patterns, but rather, can be seen as part of the same process, once we take into account the Baldwin effect. This refers to a situation—an ordinary part of natural selection—in which a particular phenotype previously produced only in response to specific environmental influences subsequently becomes the norm in a population. (The trait selected for may subsequently go to genetic fixation, so that no phenotypic plasticity remains, a development known as genetic assimilation; Waddington 1953). The language faculty supports the learning of specific kinds of linguistic systems, and it would not be at all surprising if natural selection favoured those who were able to acquire language most efficiently. Grammars that are most easily learned will be the ones that are acquired, because generations of better learners also shaped grammars to be more learnable (a type of niche construction; Odling‐Smee et al. 2003). As Anderson notes, this makes it harder for linguists to figure out what linguistic properties have to be attributed to the language faculty, rather than having external explanations. But an account along these lines also allows us to propose a language faculty that was shaped (at least in part) by natural selection, like other complex biological systems.
The next five chapters (40–44), by James Hurford, Michael Corballis, Stevan Harnad, Terrence Deacon, and Robbins Burling, investigate what cognitive capacities must have evolved before any kind of language could get off the ground. These capacities include the development of meaning (semantics and pragmatics), the origins of grounded symbols, and the ability to learn and store words. Hurford argues that other animals possess at least proto‐conceptual categories, which form the basis for conceptual meaning. Many animals can think about things that are not in the here and now, as both laboratory experiments and observations in the wild show. Animals exhibit planning abilities, have mental representations of territory, and can make calculations based on their knowledge, such as transitive inference (e.g. X is higher ranking than Y and Y is higher ranking than Z, so X must be higher than Z). Moreover, there is evidence that our closest primate relatives are capable to some extent of knowing what conspecifics know, thus exhibiting at least some of the properties that constitute theory of mind. Why, then, do only humans have language? Hurford argues that one critical factor distinguishing humans from other primates is shared intentionality, but the theme of ‘why only us?’ recurs often in Part IV, and receives different answers from various quarters. Also, work cited by (p. 355) Gibson (Chapter 3) indicates that apes do sometimes share intentions, so shared intentionality is not completely unique to humans.
Two chapters, by Harnad (42) and by Deacon (43), focus on the emergence of symbolic meaning. Like Hurford, Harnad argues that the first step is the ability to categorize—to distinguish between members and non‐members of categories such as ‘edible item’ or ‘predator’, an ability that is clearly widespread in animals. However, whereas other animals can only learn categories by sensorimotor induction, humans also acquire categories directly from other people, perhaps initially in evolution via prelinguistic gestural communication (Harnad suggests). Category names—words—are a later development. Of course, not every meaning can be transmitted by what Harnad terms ‘symbolic instruction’: this is the ‘symbol grounding problem’. Some core set of meanings can only be acquired by sensorimotor induction, and Harnad's research suggests that if about 10% of categories are acquired the hard way, by direct experience, and then given a label, this ‘grounding kernel’ of meanings can then be combined and recombined to define all remaining meanings in a lexicon.
Deacon's chapter (43) looks in depth at symbolic reference in language, arguing that the symbolic units of our mental lexicons construct a complex, interrelated network of indexical relationships, in which words always ‘point’ to each other: each morpheme is interpreted with respect to other entries in the network. Word‐learning is a very special capacity in humans. It is not merely a matter of learning arbitrary associations, which a number of animal species can certainly manage; nor is word reference analogous to the referential alarm calls found in various animals. Deacon emphasizes, though, that there is no evidence of linguistically specialized neural apparatus (structures, processes, or regions) having evolved in humans. Rather, existing cognitive capacities form the foundation for language, using existing neural systems, and language systems are widely distributed in the brain, showing much neural plasticity; as other authors in Part II emphasize, there is no ‘language module’ in the brain.
The idea that the gestural modality came before the vocal modality, mentioned above in connection with Harnad's chapter, is treated in more detail by Corballis (Chapter 41); see also MacNeilage (Chapter 46). Corballis argues (in agreement with Deacon) that primate calls do not make a good model for language. Vocal learning is very limited in other primates, whereas we know from ‘ape language’ research that all the other great apes are capable of learning the rudiments of a protolanguage presented as a visual system of manual gestures or symbols on a keyboard. Thus, Corballis argues, the earliest forms of language were likely to be gestural. Corballis suggests that as the use of the hands for other purposes (such as tool‐making) increased, so the use of facial gestures and eventually phonation would need to compensate, with the vocal modality finally taking over from the gestural, possibly quite recently. The idea that language had gestural origins is also supported by some of the authors in Parts I and II (see especially de Waal and (p. 356) Pollick, Chapter 6, and Arbib, Chapter 20). However, it remains both controversial and a minority view amongst theorists in language evolution.
Burling's chapter (44) revolves around the lexicon, and the cognitive capacities needed to acquire it. Like Hurford, Burling starts with the importance of a rich conceptual system, and of shared intentionality—specifically, in the sphere of word‐learning, the human ability to achieve ‘joint attention’ (where a child and an adult focus together on a third entity) without instruction; see also Hurford on the ‘naming insight’ in young children. Burling also emphasizes other central abilities which need to develop beyond the homologous capacities in apes—the capacity for imitation, without which no words will be learned, and the ability to perceive patterns (thus, tying in with Harnad's suggestions concerning categorization, which must also rely on pattern‐finding). As for the lexicon, a uniquely human and uniquely linguistic trait, Burling proposes an explanation for why it would be advantageous to have a system that requires so much learning: it is highly flexible, since form‐meaning pairs can be added throughout life; and, of course, flexibility is a highly adaptive trait in our species, since it enables us to adapt to (and indeed, to create) new environments and niches.
A frequent question concerns the apparent chicken‐and‐egg problem of how the first person with an advantageous linguistic mutation of some kind would benefit if no one understood them (see also de Boer, Chapter 63). Burling suggests an answer: it's not production capacities that are originally selected for, but comprehension. Even if there is initially no intention on the part of the signaller to communicate information, an animal that interprets a signal correctly will always benefit, whether the meaning is ‘I'm about to attack you’ or ‘I want to mate with you’. It seems from ape language research (see Gibson, Chapter 3) that comprehension abilities are rather advanced in our closest relatives; Kanzi is the most impressive example. (See also Seyfarth and Cheney's discussion of comprehension in baboons, Chapter 4.) Yet these abilities have clearly not been followed by the production of protolanguage in the wild state. Also, production requires fine motor skills that comprehension does not, and these skills, both manual and vocal, are lacking in most animals. Apes, for instance, seem to lack vocal motor skills. Complex motor skills not only require a lot of brain power, they also necessitate specific changes in peripheral anatomy. Burling may thus be correct that comprehension came first, but production certainly must have been selected for.
The following three chapters (45–47), by Michael Studdert‐Kennedy, Peter MacNeilage, and Andrew Carstairs‐McCarthy, look at the origins of phonetic distinctions, phonological structure, and morphology. The critical event in the evolution of phonology is the formation of the ‘particulate’ system, whereby a small number of sounds are concatenated to form an infinitely large set of meanings. As Studdert‐Kennedy points out, non‐specialists often assume, largely on the basis of alphabetic writing systems, that consonants and vowels can easily be isolated in the acoustic and/or articulatory stream of speech—but they cannot. (p. 357) What, then, are the atomic elements of sound systems? Consonants and vowels (known by linguists as segments) form the basis of all words in all languages, but they are in fact abstract, linguistic (i.e. cognitive) entities, rather than measurable physical entities.
MacNeilage (Chapter 46) starts off by outlining the standard view, that each speech segment consists of a bundle of distinctive features, such as ‘voice’ (e.g. distinguishing [p] from [b]) or ‘nasality’, though he subsequently suggests that phonological features are not innate entities, and may not even truly exist, but instead may merely be convenient categories adopted in linguistic description. Studdert‐Kennedy (Chapter 45) adopts a different view; for him, the basic particle of speech is the phonological gesture, a constriction formed by one of six organs making up the vocal tract. He defends this view on the basis of child language acquisition and evidence from computational modelling. The regions of the vocal tract are differentiated by the child in response to cognitive pressure from an increasing vocabulary, and this, posits Studdert‐Kennedy, is what also occurred in evolution. As vocabulary grew in our hominin ancestors, so particulate speech emerged—forming the requisite contrasts needed to keep words separate from one another. Computational modelling supports what would otherwise be mere conjecture, showing how segments and syllables emerge via the processes of self‐organization, with conflicting pressures from speaker economy, on the one hand, and listener clarity, on the other, predicting just the kinds of systems we find in languages.
MacNeilage argues that the initial organizational unit of speech was not the phonological gesture, but the syllable, an essentially inevitable consequence of rhythmic alternations of the jaw from the closed to the open position, paired with phonation (sound). When the jaw is raised, we get consonants; when it is lowered, we get vowels. Initially (in ontogeny and phylogeny), protosyllable frames are very simple, relying on the inertial position of the tongue. This is the case in infant babbling and first words, and, like Studdert‐Kennedy, MacNeilage thus regards infant vocalizations as providing a good indication of what the earliest speech of hominins was like, since the same kinds of biomechanical constraints operate in both cases. MacNeilage argues that the ‘poverty of the stimulus’ (see below) is irrelevant in phonology, suggesting that ‘a listener hears all the sound patterns he/she needs to learn’.
Carstairs‐McCarthy (Chapter 47) turns to a related, but distinct, area: morphology. Why should languages have complex words (combinations at morpheme and word level) at all? As Carstairs‐McCarthy points out, the very existence of a morphological system separate from the combinatorial system used for syntax seems highly unlikely, yet it exists. One of the peculiarities of morphology is that it is the only part of the language faculty that appears to be virtually optional in realization—some languages have much morphological complexity, but others almost none. Although some scholars deny that there is in fact a separate system, (p. 358) Carstairs‐McCarthy argues that some features are more or less uniquely morphological: bound morphology (morphemes that can't stand alone), allomorphic variation (baked, but taken not *taked), morphophonological alternation (leave but lef‐t not *leav‐ed), and irregularity. Although most accounts seeking an explanation of the existence of morphology turn to grammaticalization (see chapters in Part V), Carstairs‐McCarthy suggests an alternative: morphology was not a late development in the language faculty, but could have originated at the protolanguage stage, well before syntax. The basis for this account is that phonological processes of assimilation would inevitably occur between proto‐words if they were frequently contiguous, and these could give rise to allomorphy and to morphophonological alternations.
The final five chapters (48–52), by Maggie Tallerman, Derek Bickerton, Andrew Carstairs‐McCarthy, and Cedric Boeckx, look at the syntactic component of language. Three authors consider the properties of modern linguistic systems, thus asking which properties must ultimately be accounted for by a comprehensive theory of language evolution. Tallerman's chapter (48) on syntax introduces the central concepts and properties of syntactic systems in the world's languages, providing an overview for the non‐specialist reader. Picking up on these themes, Bickerton (Chapter 49) evaluates existing scenarios for the origins of syntax, concluding that the main accounts are all flawed in various ways, and suggesting an alternative scenario.
Carstairs‐McCarthy's chapter (50) on complexity asks whether all languages are equally complex in their syntax and morphosyntax. Linguists have traditionally regarded known languages (living and extinct, spoken or signed) as being commensurable in terms of complexity, but some recent claims have cast doubt on this view. Two languages in particular have been argued to lack complexity in significant ways: Riau Indonesian and Pirahã. Pirahã, for instance, has been claimed to lack recursion in its syntax. The concept of recursion came prominently to the attention of researchers outside of linguistics with the publication of Hauser et al. (2002), and is also discussed in the chapters on syntax by Bickerton (49) and Tallerman (48). Both Tallerman and Bickerton conclude that recursion is not the main explanandum in syntax.
The question of complexity with respect to creole languages arises in the chapters by Bickerton (49) and by Carstairs‐McCarthy (50). Creoles are full natural languages (they have native speakers) but are claimed by many to exhibit exceptional simplicity in their grammars (see also Roberge, Chapter 56). Bickerton regards creoles as one specific type of ‘window’ on evolutionary processes (see also Botha, Chapter 30). Bickerton aims to determine an overall order in which the main processes in syntax must emerge from a simpler precursor system. These central processes (a) assemble words into hierarchical structures; (b) determine the boundaries of the resulting units; (c) displace elements from their normal position; and (d) determine the reference of phonetically unrealized elements. Bickerton suggests that since processes for marking clause and phrase boundaries (p. 359) are lacking in creoles today, ways of determining phrase boundaries may have been a very late (or even the final) property of syntactic systems to emerge.
Stepping back to pre‐modern states, both Bickerton and Tallerman (Chapter 51) assume the existence of protolanguage, a non‐syntactic pre‐language. As a concept, protolanguage is broadly supported by researchers from most disciplines, though it is not regarded favourably from within the Minimalist programme (e.g. Piattelli‐Palmarini 2010). However, the putative properties of protolanguage are disputed. The ‘compositional’ or ‘synthetic’ model of protolanguage, most widely associated with Bickerton's work (e.g. 1990), is strongly defended by Bickerton and by Tallerman on linguistic and biological grounds. Other models exist (e.g. holistic and musical protolanguage) but are shown to be highly problematic.
Boeckx (Chapter 52) returns full circle to the questions posed by Anderson's chapter (39) at the start of Part IV. Boeckx mounts a detailed defence of the concept of the ‘poverty of the stimulus’, or ‘Plato's problem’, from within the (Chomskyan) Minimalist Program. The idea is that children acquire their native languages with such ease, rapidity, and uniformity, despite receiving an incomplete and impoverished input, that they must have some kind of ‘head start’ in this process, an innate language faculty or ‘Universal Grammar’. However, in line with Chomsky's recent work (e.g. 2005, 2010), Boeckx stresses the importance in language evolution of principles that are not specific to the language faculty, so called ‘third factor’ effects arising spontaneously as a result of generic principles such as laws of form; see also Anderson's contribution.
Boeckx argues that very little genuine evolutionary novelty is required to ‘buy’ the unique language faculty in humans, despite the fact that other animals are not close to having anything similar. He argues for a recent (occurring just in H. sapiens) and sudden emergence of the language faculty, with ‘at most one or two evolutionary innovations’. The critical development bringing all the linguistic and generic properties together is the ‘lexical envelope’: the property that turns concepts into lexical items that can be combined by the operation ‘Merge’; see Bickerton (Chapter 49) for discussion. The idea that a catastrophic event could give rise abruptly to the entire language faculty ex nihilo is criticized by Bickerton, who, as we have seen, favours the approach that sees full language as a development from a simpler protolanguage via a number of stages. Bickerton also proposes that Merge was a critical development in the evolution of syntax—in fact, the first in the sequence of events that transformed protolanguage into full language. Contrary to the Minimalist approach, though, Bickerton and Tallerman both argue for a prior protolanguage stage, in which words are simply uttered in short, unstructured sequences, not arranged hierarchically into phrases.
Linguists in general have been relatively slow to join the debates on the origins and evolution of language, but fortunately, that has changed quite rapidly over recent years. Interest is now widespread, and comes from many different theoretical perspectives, ranging from the ‘biolinguistic’ investigations (from a broadly (p. 360) Chomskyan perspective) cited by Boeckx, to the perspective taken by (again, broadly) functional approaches, centring on the study of grammaticalization (see Part V). What is evident is that linguists must spell out clearly for researchers from other disciplines what exactly it is that biologists, neurologists, psychologists, and others have to account for: the observed properties of language‐as‐it‐is (in all its forms) and the fact of its acquisition (or development) in real time by human children.