Current Biology Vol 15 No 21 R884

hardwired and independent of the environment and how much they are dependent on incoming information. Infants are sophisticated learners and it seems that, in language and music, their perceptual abilities are driven by innate mechanisms and learning by experience [8]. In both domains their learning is driven by the ability to extract statistical regularity from stimuli. Their responses to strings of words and sounds depend on the probability that one element of a string will follow another based on previous experience. In other words, it seems that the infants are making statistical inferences regarding the external world. What we now need to know is the extent to which the algorithms for extracting these regularities and building predictions are similar or even shared at some stages of development across domains such as music and language. The issue is a deep one: We make predictions about visual objects, about others emotional responses, about speech, music and the movement of objects in the world. The ground has shifted, then, from thinking about hardwired knowledge to thinking about hardwired ways of

acquiring knowledge. Whether adults can recapture the early power of these learning mechanisms or whether more developed mechanisms can be adapted to learn things in different ways is also a question opened by this line of work. What of the adults in Hannon and Trehubs experiment [7]? Would they have learned like the 12 month olds if they just had more time, or had they missed a critical time window after which they could not use the same learning mechanism? Again a halfway house might be the right place to stop: it may not be that the window is slammed shut, but that the adults continue to use predictive learning mechanisms only on a different, usually less rich and more fixed body of neural representations than are available to children. Adults can learn, of course, but they have to overcome the limitations set by their early experience and subsequent neural sensitivities. Balkan dance classes are now open for enrolment.
References
1. Hubel, D.H., and Wiesel, T.N. (1963). Receptive fields of cells in striate cortex of the very young, visually inexperienced kittens. J. Neurophysiol. 26,9941002.

2. Kuhl, P.K. (1994). Learning and representation in speech and language. Curr. Opin. Neurobiol. 5, 812822. 3. Werker, J.F., and Tees, R.C. (1984). Cross-language speech perception: Evidence for perceptual reorganisation during the first year of life. Infant Behav. Dev. 7, 4963. 4. Werker, J.F., and Lalonde, C.E. (1988). Cross-language speech perception: initial capabilities and developmental change. Dev. Psychol. 24, 672683. 5. Lynch, M.P., Eilers, R.E., Oller, D.K., and Urbano, R.C. (1990). Innateness, experience, and music perception. Psychol. Sci. 1, 272276. 6. Hannon, E.E., and Trehub, S.E. (2005). Metrical categories in infancy and adulthood. Psychol. Sci. 16, 4855. 7. Hannon, E.E., and Trehub, S.E. (2005). Tuning in to musical rhythms: Infants learn more readily than adults. Proc. Natl. Acad. Sci. USA 102, 1263912643. 8. Streeter, L.A. (1976). Language perception of two month old infants shows effects of both innate mechanisms and experience. Nature 259, 3941. 9. Hirsch, H.V., and Spinelli, D.N. (1970). Visual experience modifies distribution of horizontally and vertically oriented receptive fields in cats. Science 168, 869871. 10. Saffran, J.A., Aslin, R.N., and Newport, E.L. (1996). Statistical learning by eight month old infants. Science 274, 19261928. 11. Saffran, J.A., Johnson, E.K., Aslin, R.N., and Newport, E.L. (1999). Statistical learning of tone sequences by human infants and adults. Cognition 70, 2752.
1University

of Newcastle and Wellcome Department of Imaging Neuroscience, University College London, London, UK. 2Institute of Cognitive Neuroscience and Department of Psychology, University College London, London, UK.
DOI: 10.1016/j.cub.2005.10.019

Language Evolution: What Do Chimpanzees Have to Say?

Although unique in important ways, language shares some properties with other animal communication systems. Comparative analyses of nonhuman primate vocalizations can shed light on the evolution of languages special features. Adam Clark Arcadi Human language exhibits many unique features compared with the communication systems of other animals. The most obvious of these is its expressive power the grammatical structure of language permits an infinite number of meaningful utterances [1]. Language behavior depends on the ability to model the mental states of conspecifics [2]. And the ability to produce and process speech involves specialized orofacial, respiratory and perceptual abilities [3]. But these potentially unique features are also supported by capacities that show continuities with other species. For example, members of some species partition continuous acoustic variation categorically, exhibit lateralization in perceptual processing, require auditory feedback to learn species-specific vocalizations, engage in timed vocal interactions, vary call production depending on their audience, and

encode information about external events in their calls [4]. Tracing the evolution of language requires clarifying the nature of continuities between human and nonhuman cognitive structures and communication systems in order to specify likely pathways by which languages unique features could have emerged [5]. One intriguing research area, explored by Slocombe and Zuberbhler [6] in a recent issue of Current Biology, concerns the possibility that some animal signals refer to objects or events external to the signaler, and may therefore be similar to words. The first evidence of such referential potential came from the observation that wild vervet monkeys (Cercopithecus aethiops) produce acoustically distinct alarm calls in response to their three most important predators

Dispatch R885

eagles, leopards and snakes and that listeners respond differently to each [7]. Broadcasting recordings of these calls in the absence of actual predators elicits the same behavioral responses from listeners as the predators themselves, suggesting that the vocalizations in some way stand for the predators [8]. Similar observations have been made for the alarm calls of ring-tailed lemurs (Lemur catta) [9], Diana and Campbells monkeys (C. diana, C. campbelli) [10], and domestic chickens (Gallus gallus) [11]. There are, however, important ways in which such calls differ from human words. Human words are built according to phonological rules from a finite number of sound contrasts characteristic of a given language community [12]. The consequence of this structural property, whereby meaningless units are combined into meaningful ones, is that languages contain a vast number of words. By contrast, vocal repertoires in animals are limited to a few dozen calls, and evidence for phonological rules is scarce [13]. In addition, unlike the alarm calls of some primate species, a large percentage of words refer to concepts rather than physical entities, or have no referents at all, functioning instead as grammatical items (articles, prepositions and so forth). Finally, there are substantial philosophical and empirical obstacles to specifying what animal vocalizations might mean to signalers and receivers, and thus how similar they could be cognitively to human words. For example, does hearing a leopard alarm call produce a mental picture of a leopard in the vervet listener, or simply an escape routine [14]? In recognition of the fact that animal signals cannot be the same as words, the term functionally referential was coined to characterize those calls that appear to encode sufficient information about referent characteristics to allow conspecific receivers to respond appropriately [15]. Two empirical requirements for demonstrating

Figure 1. Liberius, the juvenile male chimpanzee tested with food grunt playbacks [6]. Photograph courtesy of Katie Slocombe.

functional reference are recognized: high production specificity (a call is always associated with the same stimulus), and high response specificity in the absence of other cues (listener responses to playbacks match responses to referents) [16]. The empirical demonstration of functional reference is most compelling when the vocalizations under consideration are from different call types, rather than being variants of the same vocalization, and when listener responses differ in kind rather than in degree. When these conditions are met, it is more likely that the motivational component of calls, which can operate as a contextual cue to supplement the referential component of signals, can be experimentally controlled [15]. As they reported in Current Biology last month, Slocombe and Zuberbhler [6] conducted experimental playbacks of chimpanzee food calls to a captive chimpanzee in order to examine the referential potential of these calls (Figure 1). Many animals and birds are known to produce specific vocalizations upon discovery of food, and experimental studies have suggested that some of these signals meet the criteria of functional reference [17]. Wild

chimpanzees produce one type of call, the rough grunt, upon discovering food and in no other contexts, thus fulfilling the requirement of production specificity [18]. Experimental studies have shown that the rate of rough grunting depends on the amount and divisibility of food present [19]. Field workers have noted that rough grunts are acoustically variable, especially in being relatively high or low pitched, but field observations have not revealed differences in food quality or quantity associated with this acoustic variability. Instead, acoustic differences have been associated with social context [20]. Thus, suggestions of functional reference for this call have been restricted to food discovery in general. Based on acoustic analyses of 82 calls delivered in 19 bouts, Slocombe and Zuberbhler [6] were able to distinguish two rough grunt variants produced by three captive chimpanzees: a higher pitched one given to a preferred food type, and a lower pitched noisier one given to a less prized food. The chimpanzees were habituated to two feeding stations, a tree baited with the preferred food and a tree baited with the less desirable food. Experimental playbacks (17 tests

Current Biology Vol 15 No 21 R886

and 10 controls) of the two rough grunt variants elicited differential searching behavior by a juvenile male. In four of the six initial trials, this male first searched the tree containing the food corresponding to the playback stimulus. In the 11 remaining trials, the male searched the tree with the preferred food first irrespective of playback stimulus, but searched more intensively after hearing grunts given to the preferred food. The authors conclude that the two rough grunts meet the criteria of functional reference, containing sufficient information about the nature of food sources to guide their subjects searching behavior. This conclusion goes beyond previous suggestions of functional reference for chimpanzee rough grunts in claiming not just that the rough grunt is functionally referential in general, but that different rough grunt variants refer to different foods or food preferences. Although the demonstration of functional reference in a chimpanzee vocalization would be an important contribution to language origins research, Slocombe and Zuberbhlers [6] findings should be viewed with considerable caution for several reasons. First, responses to the playbacks were not sharply differentiated (low response specificity, unlike monkey responses to alarm calls), a problem also encountered in the analysis of rhesus monkey food calls [17]. Second, the authors were unable to control for individual, age, and sex differences in call structure. Field analyses have indicated that younger individuals give higher pitched rough grunts [20]. In this study, a subadult male and adult female contributed 6 of 9 bouts to the preferred food, but only 5 of 10 bouts to the less preferred food. Moreover, it is unclear what effect social context had on call production. Again, field analyses have indicated that social

excitement affects call structure [20]. Since the hypothesized referential signals are variants of the same call type, it is therefore difficult to evaluate the contribution of motivational state as a contextual cue embedded in the vocal playback [15]. Taken together, these ambiguities mean that production specificity, like response specificity, has not been clearly demonstrated. Larger sample sizes of both callers and test subjects are needed to disentangle the possible sources of variation in rough grunt acoustic structure and to isolate their effects on listeners. Despite these shortcomings, this first attempt to test for functional reference in a chimpanzee vocalization should stimulate further research. Efforts in captivity can profit from the opportunity to manipulate key aspects of social context and motivational state in both senders and receivers. Renewed attention to food calling in the wild can help specify its functional significance. Tracing the evolution of language will always be hampered by the fact that early hominid species that likely possessed proto-linguistic abilities are extinct. Nevertheless, it is a fair hypothesis that at least some of the substrate on which language is founded can be explored through comparative analyses of the vocal behavior of our closest phylogenetic relatives.
References
1. Chomsky, N. (1980). Rules and Representations. (New York: Columbia University Press). 2. Seyfarth, R.M., and Cheney, D.L. (2003). Signalers and receivers in animal communication. Annu. Rev. Psychol. 54, 145173. 3. Lieberman, P. 1991. Uniquely Human: The Evolution of Speech, Thought, and Selfless Behavior. (Cambridge, Massachusetts: Harvard University Press). 4. Evans, C.S., and Marler, P. (1995). Language and animal communication: Parallels and contrasts. In Comparative Approaches to Cognitive Science. H.L. Roitblat and J-A. Meyer, eds. (Cambridge, Massachusetts: MIT Press), pp. 341382.

5. Hauser, M.D., Chomsky, N., and Fitch, W.T. (2002). The faculty of language: What is it, who has it, and how did it evolve? Science 298, 15691579. 6. Slocombe, S., and Zuberbhler, K. (2005). Functionally referential communication in a chimpanzee. Curr. Biol. 15, 17791784. 7. Struhsaker, T.T. (1968). Auditory communication among vervet monkeys (Cercopithecus aethiops). In Social Communication among Primates. S. A. Altmann, ed. (Chicago: University of Chicago Press), pp. 281324. 8. Seyfarth, R.M., Cheney, D.L., and Marler, P. (1980). Monkey responses to three different alarm calls: Evidence of predator classification and semantic communication. Science 210, 801803. 9. Macedonia, J. (1990). What is communicated in the antipredator calls of lemurs: Evidence from playback experiments with ring-tailed lemurs and ruffed lemurs. Ethology 86, 177190. 10. Zuberbhler, K. (2003). Referential signaling in non-human primates: Cognitive precursors and limitations for the evolution of language. Adv. Stud. Behav. 33, 265307. 11. Evans, C.S., Evans, L., and Marler, P. (1993). On the meaning of alarm calls: Functional reference in an avian vocal system. Anim. Behav. 46, 2338. 12. Jakobson, R., and Halle, M. (1980). Fundamentals of Language, 4th edn. (The Hague: Mouton). 13. Mitani, J.C., and Marler, P. (1989). A phonological analysis of male gibbon singing behavior. Behaviour 109, 2045. 14. Owren, M.J., and Rendall, D. (2001). Sound on the rebound: Bringing form and function back to the forefront in understanding nonhuman primate vocal signaling. Evol. Anthrop. 10, 5871. 15. Marler, P., Evans, C.S., and Hauser, M.D. (1992). Animal signals: Motivational, referential, or both? In Nonverbal Vocal Communication. H. Papousek, U. Jrgens, and M. Papousek, eds. (Cambridge: Cambridge University Press), pp. 6686. 16. Macedonia, J.M., and Evans, C.S. (1993). Variation among mammalian call systems and the problem of meaning in animal signals. Ethology 93, 177197. 17. Hauser, M.D. (1998). Functional referents and acoustic similarity: Field playback experiments with rhesus monkeys. Anim. Behav. 55, 16471658. 18. Goodall, J. (1986). The Chimpanzees of Gombe: Patterns of Behavior. (Cambridge, Massachusetts and London: Belknap Press of Harvard University Press). 19. Hauser, M., Teixidor, P., Field, L., and Flaherty, R. (1993). Food-elicited calls in chimpanzees: Effects of food quantity and divisibility. Anim. Behav. 45, 817819. 20. Marler, P., and Tenaza, R. (1977). Signaling behavior of apes with special reference to vocalization. In How Animals Communicate. T. A. Sebeok, ed. (Bloomington: Indiana University Press), pp. 9651032.

Department of Anthropology, 264 McGraw Hall, Cornell University, Ithaca, New York 14853, USA.
DOI: 10.1016/j.cub.2005.10.020