THE SOCIAL ORIGINS OF LANGUAGE

Robert M. Seyfarth and Dorothy L. Cheney

Human language poses a problem for evolutionary theory because of the striking discontinuities between language and the communication of our closest animal relatives, the nonhuman primates. How could language have evolved from the common ancestor of these two very different systems?

The qualitative differences between language and nonhuman primate communication are now well known (see Fitch 2010 for review). All human languages are built up from a large repertoire of learned, modifiable sounds. These sounds are combined into phonemes, which are combined into words, which in turn are combined according to grammatical rules into sentences. In sentences, the meaning of each word derives both from its own, stand-alone meaning and from its function role as a noun, verb, or modifier. Grammatical rules allow a finite number of elements to convey an infinite number of meanings: the meaning of a sentence is more than just the summed meanings of its constituent words. Languages derive their communicative power from being discrete, combinatorial, rule-governed, and open-ended computational systems, like the number system or the use of 1s and 0s in a digital computer (Jackendoff 1994; Pinker 1994).

By contrast, nonhuman primates (prosimians, monkeys, and apes) — and indeed most mammals — have a relatively small number of calls in their vocal repertoire. These calls exhibit only slight modification during development; that is, their acoustic structure appears to be largely under genetic control (see Hammerschmidt and Fischer 2008 for review). Furthermore, while animals can give or withhold calls voluntarily and modify the timing of vocal production (reviewed in Seyfarth and Cheney 2010), different call types are rarely given in rule-governed combinations (but see Ouattara, Lemasson, and Zuberbuhler 2009; Zuberbuhler 2014). When call combinations do occur, there is little evidence that individual calls play functional roles as agents, actions, or patients. As a result, primate vocalizations, when compared to language, appear to convey only limited information (Bickerton 1990; Hurford 2007; Fitch 2010).

Differences between human language and nonhuman primate communication are most apparent in the domain of call production. Continuities are more apparent, however, when one considers the neural and cognitive mechanisms that underlie call perception, and the social function of language and communication in the daily lives of individuals. Here we begin by briefly reviewing the evidence that homologous brain mechanisms in human and nonhuman primates underlie the recognition of individual faces and voices; the multimodal processing of visual and auditory signals; the recognition of objects; and the recognition of call meaning. These results are relevant to any theory of language evolution because they suggest that, for much of their shared evolutionary history, human and nonhuman primates faced similar communicative problems and responded by evolving similar neural mechanisms.

What were these similar communicative problems? In the second part of this essay we compare how language functions in human social interactions with the function of vocalizations in the daily lives of animals, particularly baboons. We argue that, while the two systems of communication are structurally very different, they share many functions. These shared functions help explain the evolution of homologous neural mechanisms.

To understand the function of primate vocalizations, one must understand what primates know about each other. In baboons, for example, this includes knowledge of individual identity, dominance rank, matrilineal kin membership, and the use of different vocalizations in different social circumstances. In chimpanzees, it includes knowledge of other animals' alliance partners. In the third part of this essay we show that selection has favored in baboons — and, by extension, other primates — a system of communication that is discrete, combinatorial, rule-governed, and open-ended. We argue that this system was common to our prelinguistic primate ancestors and that when language later evolved from this common foundation, many of its distinctive features were already in place.

SHARED BRAIN MECHANISMS

An area in the human temporal cortex, the fusiform face area, responds especially strongly to the presentation of faces and appears to be specialized for face recognition (Kanwisher, McDermott, and Chun 1997). A similar area, consisting entirely of face-selective cells, exists in the macaque temporal cortex (Tsao et al. 2003, 2006; Freiwald, Tsao, and Livingston 2009). Humans also have a region in the superior temporal sulcus that is particularly responsive to human voices and appears to play an important role in voice recognition (Van Lancker et al. 1988; Belin et al. 2000; Belin and Zattore 2003). Petkov et al. (2008) document the existence of a similar area in the macaque brain.

When communicating with one another, humans exhibit multisensory integration: bimodal stimuli (voices and concurrent facial expressions) consistently elicit stronger neural activity than would be elicited by either voices or faces alone (e.g., Wright et al. 2003). Human infants are sensitive to the "match" between speech sounds and their corresponding facial expressions, responding more strongly to incongruent than to congruent vocal and visual stimuli (Kuhl and Meltzoff 1984; Patterson and Werker 2003). A variety of studies document similar multisensory integration in monkeys (Ghazanfar and Logothetis 2003; Ghazanfar et al. 2005; Ghazanfar, Chandrasekaran, and Logothetis 2008; Sliwa et al. 2011; Adachi and Hampton 2012).

In both humans and macaques, neurons in the ventral premotor cortex exhibit neural activity both when performing a specific action and when observing another perform the same action (see Ferrari, Bonini, and Fogassi 2009 for review). These "mirror neurons" seem likely to be involved in the development of novel behaviors and may constitute a shared, homologous neural substrate for imitative behavior (Ferrari, Bonini, and Fogassi 2009; de Waal and Ferrari 2010).

We take it for granted that humans can classify words according to either their meaning or their acoustic properties. Judged according to their meaning, treachery and deceit are alike whereas treachery and lechery are different; judged according to their acoustic properties, these assessments would be reversed. The "ape language" projects were the first to suggest that, like humans, nonhuman primates can classify communicative signals...