We welcome Christiansen & Chater's (C&C's) bold proposal that “language is easy for us to learn and use, not because our brains embody knowledge of language, but because language has adapted to our brains” (target article, sect. 1, para. 3). Whether our language capabilities derive from a Universal Grammar (UG) or draw on more “domain-general” mechanisms, we believe there is somewhat more potential than C&C want to allow for genetic adaptations to language or language processing.

Modern humans' unique social system makes demands on language unknown in any other species. We are the only animal with complex systems of cooperation and exchange among unrelated individuals. We help people we may never see again; we have an elaborate division of labour; and we sometimes act in ways that benefit the group but at considerable cost to our own fitness, even to the point of causing death. Our social system is one based on reciprocation and trust and is therefore vulnerable to cheats who take advantage of the goodwill of cooperators, without returning aid themselves.

Human language was almost certainly required to manage the social complexity of modern human society (Pagel Reference Pagel2008). Some of the most frequently used words in spoken communication are those relating to “social coordinates” – the pronouns I, you, he/she, we, they, who, and verbs to say and to know. As language became more sophisticated in our ancestors, so also grew the opportunities to use it for personal gain – to enhance one's reputation, to tell lies, and to spread rumours. Individuals with poor language skills would have been markedly disadvantaged in a society of shrewd, guileful, and self-interested speakers.

It does not matter for our view whether the selective pressures for language and language skills built something like a Universal Grammar or whether they modified elements of a more “domain-general” apparatus (whatever form it may take). If the latter, general traits (such as arms and legs or general cognitive abilities) are typically shaped by competing demands of different functions and environmental contingencies. Our position is that a facility with language would have been one of the key sets of demands shaping some of these domain-general mechanisms.

Bird-song is not a recursive symbolic language, but it illustrates how the foundations for producing a culturally influenced and geographically variable trait may be encoded genetically (Nottebohm Reference Nottebohm2005). In humans, evidence for genetic variation related to language learning and processing may be all around us. Our brains appear to be “wired for language” (Glasser & Rilling, in press), brain structures related to language are likely to be heritable (Thompson et al. Reference Thompson, Cannon, Narr, van Erp, Poutanen, Huttunen, Lönnqvist, Standertskjöld-Nordenstam, Kaprio, Khaledy, Dail, Zoumalan and Toga2001), and people seem to vary greatly in their abilities to express themselves verbally. The pervasive phenomenon of dyslexia, which seems to affect both language learning and production, and the phenomena associated with FOXP2 variants also suggest that genetic variance for language capabilities is common.

C&C argue that language changes too rapidly for genes to get a fix and adapt to it. Possibly, but the human immune system has diverged genetically around the world in response to rapidly evolving parasites. We also note that not all elements of language evolve at a high rate. Dunn et al. (Reference Dunn, Terrill, Reesink, Foley and Levinson2005) report phylogenetic signals in Papuan language typology that may have been preserved for many millennia. We have shown that highly expressed or frequently used words in the Indo-European lexicon evolve as slowly as some genes (Pagel et al. Reference Pagel, Atkinson and Meade2007). The word for the number “two” is highly expressed and is cognate across the entire Indo-European language family. Thus, here is a lexical item that has retained homology throughout roughly 130,000 language-years of evolution. By comparison, a genetic tolerance in adults for lactose probably arose and spread to high frequencies within the last 6,000 years in some populations of this same language family (Burger et al. Reference Burger, Kirchner, Bramanti, Haak and Thomas2007).

We do not suggest that there is a gene for the number “two.” But these examples from empirical studies may indicate that there are general features of linguistic systems that, either on their own or because of their links to other systems, change slowly enough that we could expect to see genetic adaptations for the cognitive mechanisms that process them. The frequency-effect we document also depends on the part of speech. For a given frequency of use, conjunctions evolve fastest, followed by prepositions, then adjectives, verbs, nouns, special adverbs, pronouns, and finally numbers, which evolve very slowly. Why parts of speech influence the rate of evolution is unclear but may be relevant in a very general way to C&C's discussion of Frequency×Regularity interactions. We wonder if, for example, the different parts of speech are linked with (or overlap with) other features of language that may make them less prone to change over time. Alternatively, might the rank ordering of parts of speech correspond to memory and production effects related to word “concreteness” (e.g., Bleasdale Reference Bleasdale1987; Jessen et al. Reference Jessen, Heun, Erb, Granath, Klose, Papassotiropoulos and Grodd2000), with numbers the most concrete and prepositions the least?

We see human language as a trait on which we have stamped our cognitive and psychological signatures along the road to ensuring that language is a useful tool for us. This is, in some sense, the “flip side” of C&C's view of languages as semi-autonomous systems that adapt to us. Some of the signatures of our shaping of language may emerge from historical analyses of the variation among languages. C&C discuss some, and the frequency-effect discussed above may be another. Social factors may also play a role. We have recently been able to show that languages often exhibit a rapid or punctuational burst of evolution around the time of their divergence from a sister language (Atkinson et al. Reference Atkinson, Meade, Venditti, Greenhill and Pagel2008). One interpretation of our findings is that humans use language as a tool for establishing a distinct social identity. If social forces such as this can leave their imprint on languages over historical time, we should not be surprised to find that features of our information processing systems can do the same. Equally, for a trait of such importance to our lives we should not be surprised that we have also adapted to it.