Introduction

A key desideratum in the study of the biology and evolution of language ("biolinguistics") is to understand what allows humans to acquire full-blown language, but not other animals, especially our nearest relatives the chimpanzees. Studies of chimpanzees have revealed an impressively broad set of cognitive and communicative abilities, and these empirically-determined similarities are a key source of data for reconstructing the capacities of our last common ancestor with chimpanzees (hereafter, the "LCA") which lived around 6 million years ago. However, a chimp, raised in diapers in a human home, will develop no speech production capabilities, and only limited comprehension of speech.  Even with intensive training using gestures or lexigrams, chimps acquire a productive vocabulary of only a few hundred signs, unlike the child who under identical circumstances acquires a lexicon of tens of thousands of words, and a complex syntax that allows generation of an unlimited number of sentences. Taking the reconstructed cognitive capabilities of the LCA as the starting point, then, the goal is to understand what changes in preexisting capabilities, and addition of new capacities, have occurred in the hominid line since our divergence from chimps.

Detailed comparative study has revealed at least three capacities that play a key role in language and differentiate humans from chimps: speech (especially vocal imitation), syntax (especially recursive syntax) and semantics (especially intentional communication).  A recent survey of hypotheses about language evolution (Christensen & Kirby, 2003) reveals broad agreement about these three components (although authors differ in their emphasis and terminology), despite ongoing debate concerning many other issues.  This consensus list highlights the minimum that biolinguistics should be able to explain in an adequate theory of language evolution, which starts with the LCA and ends with all key components of the contemporary language capacity in place. In principle, each of these capacities could have evolved at different times, and under different selective pressures.  Further, the selective value of some capacity (say vocal imitation) might have varied during different stages of hominid evolution. Indeed, most contemporary theorists, have offered "dual-stage theories" positing different selective forces at different phylogenetic stages (e.g.,Arbib, 2003; Bickerton, 1998; Corballis, 2003; Donald, 1993).  Current consensus thus isolates three key components to be explained, and posits a multistage phylogenetic process to attain them.

Although speech, syntax and semantics are profitably separated for conceptual clarity, they obviously represent closely-linked capacities, and their interactions are as crucial as their individual function.  Here I explore some of these interactions from a comparative, evolutionary viewpoint. To the extent that synergies can be identified between components, coevolutionary accounts of language evolution gain in plausibility.  After a brief comparative review of speech and syntax, I focus on semantics, especially intentional communication in relation to theory of mind.  Intentionality in the human sense does not appear to be present in the communication systems of nonhuman primates (Seyfarth & Cheney, 2003; Seyfarth & Cheney, in press), despite a widespread current assumption (based on misinterpretation of vervet alarm calls) that alarm calls are homologous to words.  Second, while theory of mind abilities in chimpanzees remains an area of intensive investigation and debate, it is clear that human children go far beyond chimps in this arena by the age of three or four, and thus that this capacity also has either appeared, or hypertrophied, since our evolution from the LCA.  I then suggest 1) that vocal imitative skills provide a "scaffolding" for theory of mind abilities, via the mechanism of quoting, and 2) that the capacity for recursive embedding of minds within minds (high-order theory of mind) may be closely related to recursive embedding in syntax, itself related to quoting. 

Speech and Syntax

Vocal imitation, the ability to flexibly recreate novel utterances after hearing them produced by another individual, is a well-developed skill in human children by the age of three, and provides a crucial prerequisite for spoken language.  Without this imitative ability, an indefinitely-extensible lexicon would be impossible.  Although well-developed vocal imitation is observed in many nonhuman species, including birds, cetaceans (whales and dolphins), seals and perhaps bats (Janik & Slater, 1997), it has not been observed in any nonhuman primate, including chimps, despite intensive investigation.  While some birds can imitate arbitrary sounds, including human speech, chimpanzees fail to do so even with intensive training.  Indeed, primate vocalizations appear to be relatively involuntary (though audience effects exist, showing that primate calls are not entirely stimulus-driven as sometimes stated (Cheney & Seyfarth, 1990b)).  In addition, primate calls are highly genetically canalized: many are present at birth, and will reliably develop in the absence of any relevant input (Winter et al., 1973; Hauser, 1996).  Thus primate calls appear to resemble human laughter or crying (also present at birth, even in deaf children) more than human words.  This is as true for vervet alarm calls as for any other primate call: the structure and even some aspects of the meaning of these calls appears to be genetically-determined (Seyfarth & Cheney, 1980).  This is one reason that the facile equation of vervet alarm calls with human words must be avoided: the only enlargement of the vervet call "lexicon" must be accomplished by natural selection over millennia, not during individual ontogeny as in humans.  Thus vocal imitation, and the lexicon that goes with it, is a crucial component of language that has evolved since the LCA (note that the existence of non-vocal imitation, equally important for signed languages, does not change this necessity). 

Syntax in modern human languages is highly complex, and entails several potentially independent innovations.  At the simplest level, the ability to construct and process phrase structure, independent of meaning, is a crucial component of our ability to make "infinite use of finite means" in language: embedding phonemes within syllables within words within sentences, we can use a small "alphabet" to generate an unlimited set of signals.  Production of correspondingly flexible phrase structures is not observed in nonhuman primate calls, nor is it available at the perceptual level by those primates tested so far (Fitch & Hauser, 2004); however, phrase structure may be present in the "songs" of birds and humpback whales..  A second aspect of syntax, recursion, is more restricted and more powerful.  Recursion entails phrase structure, but goes beyond it in that structures of the same type may be embedded within one another.  Recursion is not present at the phonological level (we cannot embed syllables within syllables) but is clearly present at the sentential level (we can embed phrases within phrases).  We currently have no evidence for such recursive embedding within animal communication systems (Hauser et al., 2002).  Finally, long distance dependencies (connections of elements between sentences, or between multiple phrases in a sentence), as in anaphora, has no known parallel in animal communication.  However, detecting such semantically-defined dependencies in non-semantic vocalizations like birdsong may be impossible.  Below I focus on recursive syntax, because it has not yet been observed in animal communication systems, but could be in principle (a mockingbird could "quote" the phrase-structured utterance of another bird within its own phrases, or a humpback whale the song of another male).

Semantics and Theory of Mind

A virtually universal aspect of human languages is symmetry between signaler and recipient: if you can make a signal with a certain meaning, then you can perceive the same correspondence, and vice versa.  This symmetry or "parity" is central to both word meaning and Gricean maxims in pragmatics.  Speaker/listener symmetry seems so natural to us that it has taken many years to recognize that it is not necessarily present in other communication systems. A key example comes from experimental work on monkey calls (Cheney & Seyfarth, 1990a).  Japanese macaques have food and predator alarm calls which are reliably produced in the presence of food or predators, and which appear to be appropriately perceived by others.  In these experiments, a mother monkey was shown food or a predator in a separate test enclosure, while her infant was absent.  Then the infant was introduced into the test enclosure.  If the mother is aware of her offspring's ignorance, we predict an increase in her food or alarm calling when her infant enters.  No such increases were observed.  Furthermore, mothers produced virtually no alarm calls in the "infant with predator" context.  Given the strong inclination of primate mothers to protect their offspring, and the strong evolutionary incentive to do so, these results powerfully suggest that monkeys are unable to represent ignorance in their own offspring, or to vocalize in a way that remedies such ignorance.  In contrast, monkeys have a flexible and well-developed interpretive ability (on the part of receivers), easily learning proper interpretation of novel calls including those of other species (Seyfarth & Cheney, 1990). The fact that the infant clearly would respond appropriately to alarm or food calls, if they had been made, makes the absence of calling even more striking. 

The strong conclusion to be drawn from this and similar research is that there is an asymmetry in monkeys' perception and production of calls. In particular, signalers do not appear sensitive to the knowledge or ignorance of intended listeners, and thus appear unable to intentionally inform others.  The signaler reacts vocally (modulo some audience effects) to its perceptual world, irrespective of its listener's knowledge, and most of the communicative work is done, inferentially, by the perceiver (Seyfarth & Cheney, in press; Seyfarth & Cheney, 2003). Thus, functional referentiality of alarm calls (shown by listeners' proper reaction to played-back calls) appears different in kind from the referentiality of human language.  The producer's half of the Gricean bargain is not upheld.

This is perhaps less surprising given the paucity of evidence for theory of mind in primates (including chimpanzees).  In the classic "knower-guesser" paradigm, where a subject desiring food must use cues from individuals, who either could or could not know where the food is hidden, primates consistently perform at chance (Povinelli et al., 1990; Povinelli et al., 1991).  In a striking example, chimps fail to choose correctly when one of the two "informants" has a bucket over their head.  This and other evidence led (Tomasello & Call, 1997) to conclude that absence of even rudimentary theory of mind capacities was a key (perhaps the key) difference between humans and chimps.  However, new data has moderated this view.  In particular, when placed in a competitive situation, chimps do behave as if they understand that "seeing is knowing", and are able to adjust their competitive behaviour accordingly (Hare et al., 2000).  Thus, the problem may not be a complete inability to represent other minds, but a missing propensity to interpret them cooperatively.  Again, the kind of collaborative exchange of information that is at the heart of human linguistic communication appears absent in other primates (interestingly, domestic dogs DO collaboratively interpret seeing as knowing (Hare et al., 2002)).

Summarizing, the referential nature of nonhuman primates has a peculiar asymmetry in comparison with human language.  The poorly developed capacity to represent the contents of others' minds leads to a situation in which most of the communicative work is done by the receiver of a signal, based on inference, rather than intentionally by the producer.  Well-developed theory of mind thus seems to be a crucial component of human cognition that is both necessary for fully-formed language, and absent in our nearest cousins. In the rest of this essay will argue that the connection between theory of mind and language is not one way, and propose two independent co-evolutionary links.  First, the simple ability to produce a novel signal with the same structure as some other signaller's, which I will call "quoting" provides a potentially invaluable scaffolding for theory of mind.  And second, the ability to represent minds within minds (high-order theory of mind) may have provided a crucial stepping stone towards, and selective pressure for, recursive syntax.

Quoting and Theory of Mind

Children hear their parents say many things they do not understand.  By the age of three or four, they can reproduce these utterances effortlessly (and by rote, often with humorous effect).  Put another way, children exhibit a sensorimotor ability (vocal imitation) that allows them to recreate the sensorimotor component of a meaningful utterance, while comprehending the meaning dimly if at all.  However, the ability to retrieve the utterance (silently or out loud) along with its context, long after the communicative act is over, perhaps in concert with new information, provides a very important advantage in trying to make sense of others' minds.  At the very least, the ability to encode part of a communicative event in the concrete realm of sensorimotor memory aids the encoding and retrieval of the event.  More generally, the phonological and syntactic structure of the remembered utterance can act as a scaffold for a slow but steady increase in the understanding of the intended meaning.  Anchored by the sensorimotor representation, the semantic and pragmatic interpretation process can continue days or even years after the utterance itself has faded. We all remember instances where we suddenly realized "ah, that's what she meant by that", months or years post-utterance.  Thus, I suggest, the durable non-semantic sensorimotor representation (the "quote") can act as a scaffold upon which a rich interpretation of meaning can slowly build.

It is important to note that this advantage to listeners would accrue even in the absence of intentional signaling by the speaker.  Someone muttering to himself, speaking to a third party, or even producing inchoate grunts associated with some other task, might produce signals that afford quoting.  Nor must the call quoted be structured syntactically (into words and morphemes) for quoting to be advantageous.  A "frozen" holphrastic utterance like "how do you do" can still be subjected to offline rumination and interpretation to discover its appropriate context ("aha - they always do that when two unacquainted people first meet!").  Given the repeated exposures and variegated data necessary to infer context, even simple holophrastic quoting could prove advantageous.  Thus the coevolutionary relationship between quoting and theory of mind might have long predated language in its fully modern form, and provides a plausible, persistent selective force in the evolutionary path to language.

Second, and finally, I want to consider a related but considerably more sophisticated potential relationship between theory of mind and syntactic recursion.  Basic phonological phrase structure entails combination of one type of entity into larger wholes (syllables into words, for example) without blending of their properties.  This kind of structured compositional recombination seems quite ubiquitous in nature (Studdert-Kennedy, 1998).  An example is the combination of small motor actions into larger complexes, whether chews within a feeding complex or prey capture within a locomotion complex.  While the existence of compositional ability in one domain does not entail it in another (monkeys have it in their motor control, but not their vocal communication), we need not look far for possible precursor abilities of such basic phonological-level recombination, motor control being the most obvious possibility (Lieberman, 1984; MacNeilage, 1998).  Another possibility is spatial cognition: we readily represent furniture within rooms within houses within towns, and so on (or substitute fruit, branches, trees and forests).  In recursive syntax, in contrast,  entities of the same type are embedded ("self-embedding").  This provides more of an evolutionary challenge.  We string motor acts together sequentially, but we do not embed chews within chews, or syllables within syllables.  It is not clear that motor behaviour could provide a preadaptation for recursion.  Are there other cognitive capacities that might?

I join many others (Bergman et al., 2003; Byrne & Whiten, 1988; Dunbar, 1998) in suggesting that social intelligence has the requisite properties to both provide precursors for, and drive the evolution of, aspects of language..  The reasons are simple: in a complex social world there is always a premium on representing the minds of others.  Furthermore, the potential rewards are great for doing so even slightly better than the competition.  Such a Machiavellian "arms race" is the perfect evolutionary context to drive the rapid evolution of intelligence (Humphrey, 1976).  In particular, few cognitive domains seem more suited to utilize full syntactic recursion than social cognition.  The ability to represent minds within minds ("John knows that I know") is a huge step beyond simply knowing what John knows (or doesn't).  But the ability to perform such embedding at progressively higher levels may be equally valuable strategically.  The sentence "Mary was surprised that Jacques didn't know that Judith lied to Pierre" is both easy for humans to understand, and of great relevance in planning a coalition with Mary or Jacques.  But it requires a level of self-embedding that is hard to envision being possible (or important) in other nonlinguistic cognitive domains.  It should require no further argument to see that the quoting abilities discussed above are also relevant to this ability (both simply storing and processing them, and communicating them).  Thus the hypothesis that high-order theory of mind is closely related to recursive syntactic self-embedding should be given serious consideration.  More generally, detailed comparative consideration of the interactions between cognitive components of the language faculty may provide useful insights for hypotheses about language evolution.

References:

Arbib, M. A. 2003. The evolving mirror system: A neural basis for language readiness. In: Language Evolution (Ed. by Christensen, M. & Kirby, S.), pp. 182-200. Oxford: Oxford University Press.

Bergman, T. J., Beehner, J. C., Cheney, D. L. & Seyfarth, R. M. 2003. Hierarchical classification by rank and kinship in baboons. Science, 302, 1234-1236.

Bickerton, D. 1998. Catastrophic evolution: the case for a single step from protolanguage to full human language. In: Approaches to the Evolution of Language (Ed. by Hurford, J. R., Studdert-Kennedy, M. & Knight, C.), pp. 341-358. New York: Cambridge University Press.

Byrne, R. W. & Whiten, A. 1988. Machiavellian Intelligence: Social expertise and the evolution of intellect in monkeys, apes and humans. Oxford: Clarendon Press.

Cheney, D. L. & Seyfarth, R. M. 1990a. Attending to behaviour versus attending to knowledge: Examining monkeys' attribution of mental states. Animal Behaviour, 40, 742-753.

Cheney, D. L. & Seyfarth, R. M. 1990b. How monkeys see the world: Inside the mind of another species. Chicago, IL: Chicago University Press.

Christensen, M. & Kirby, S. 2003. Language Evolution. Oxford: Oxford University Press.

Corballis, M. C. 2003. From hand to mouth: The gestural origins of language. In: Language Evolution (Ed. by Christensen, M. & Kirby, S.), pp. 201-218. Oxford: Oxford University Press.

Donald, M. 1993. Origins of the Modern Mind. Cambridge, Massachusetts: Harvard University Press.

Dunbar, R. 1998. Theory of mind and the evolution of language. In: Approaches to the Evolution of Language (Ed. by Hurford, J. R., Studdert-Kennedy, M. & Knight, C.), pp. 92-110. New York: Cambridge University Press.

Fitch, W. T. & Hauser, M. D. 2004. Computational constraints on syntactic processing in a nonhuman primate. Science, 303, 377-380.

Hare, B., Brown, M., Williamson, C. & Tomasello, M. 2002. The domestication of social cognition in dogs. Science, 298, 1634-6.

Hare, B., Call, J., Agnetta, B. & Tomasello, M. 2000. Chimpanzees know what conspecifics do and do not see. Anim Behav, 59, 771-785.

Hauser, M., Chomsky, N. & Fitch, W. T. 2002. The Language Faculty: What is it, who has it, and how did it evolve? Science, 298, 1569-1579.

Hauser, M. D. 1996. The evolution of communication. Cambridge, MA: MIT Press.

Humphrey, N. K. 1976. The social function of intellect. In: Growing points in ethology (Ed. by Bateson, P. P. G. & Hinde, R. A.), pp. 303-317. Cambridge: Cambridge University Press.

Janik, V. M. & Slater, P. B. 1997. Vocal learning in mammals. Advances in the study of behavior, 26, 59-99.

Lieberman, P. 1984. The Biology and Evolution of Language. Cambridge, MA: Harvard University Press.

MacNeilage, P. F. 1998. The frame/content theory of evolution of speech production. Behavioral and Brain Sciences, 21, 499-546.

Povinelli, D. J., Nelson, K. E. & Boysen, S. T. 1990. Inferences about guessing and knowing by chimpanzees (Pan troglodytes). Journal of Comparative Psychology, 104, 203-210.

Povinelli, D. J., Parks, K. A. & Novak, M. A. 1991. Do rhesus monkeys (Macaca mulatta) attribute knowledge and ignorance to others? Journal of Comparative Psychology, 105, 318-325.

Seyfarth, R. M. & Cheney, D. in press. Constraints and preadaptations in the earliest stages of language evolution. Linguistic Review.

Seyfarth, R. M. & Cheney, D. L. 1980. The ontogeny of vervet monkey alarm-calling behavior: A preliminary report. Zeitschrift fur Tierpsychologie, 54, 37-56.

Seyfarth, R. M. & Cheney, D. L. 1990. The assessment by vervet monkeys of their own and another species' alarm calls. Animal Behaviour, 40, 754-764.

Seyfarth, R. M. & Cheney, D. L. 2003. Signalers and Receivers in Animal Communication. Annual Review of Psychology, 54, 145-173.

Studdert-Kennedy, M. 1998. The particulate origins of language generativity: from syllable to gesture. In: Approaches to the Evolution of Language (Ed. by Hurford, J. R., Studdert-Kennedy, M. & Knight, C.), pp. 202-221. New York: Cambridge University Press.

Tomasello, M. & Call, J. 1997. Primate Cognition. Oxford: Oxford University Press.