news and views Survival of the clearest

There are no fossils to show how language evolved. But is revealing how some of the defining features of human language could have been shaped by natural selection.

he study of the of language, Figure 1 According to recent famously banned by the Société de Lin- work in evolutionary game Tguistique de Paris in 1866 and dismissed theory — the latest example of as idle story-telling ever since, has returned which appears elsewhere in this to respectability. Recent years have seen a issue9 — minimizing flurry of articles and books1–5, a biannual communication errors was a research conference and now several papers major selection pressure in the by Martin Nowak and collaborators apply- evolution of human language. ing evolutionary game theory to the prob- lem6–9. Their latest offering appears on page 495 of this issue9. Language would seem to be a natural tar- get of evolutionary analysis. Universal, heri- table, rapidly acquired by children and well suited to communicating complex informa- tion, language is plausibly an adaptation that allowed our ancestors to share knowledge and negotiate agreements, fuelling the technological and demographic explosions that led our species to infest the planet millennia ago. But the very idea of evolutionary linguis- tics, like in general, draws mixed reactions from scientists. On the one hand, “nothing in biology makes sense except in the light of evolution.” The

major parts of our psychology could no THE FAR RESERVED ALL RIGHTS INC. USED BY PERMISSION. SIDE © 1994 FARWORKS, more be understood without considering their biological function than could the heart or eye. For example, visual perception does not just entertain us with pretty colours marks can serve as predictions for how Dar- mation between partners provides them and shapes, but delivers usable information winian organisms ought to work. The more with an advantage (say, by exchanging know- about the surfaces and materials of the uncannily the engineering specifications for how or coordinating their behaviour), and world. And it is no surprise that people a system match the facts of the org- that the advantage translates into more off- everywhere fear heights and snakes more anism, the more confidently we can infer spring, with similar communicative skills. than comfy chairs and bunnies. On the other that the organism was selected to perform The question then is how a stable communi- hand, other evolutionary ‘explanations’ that function. cation system might evolve from repeated of behaviour — “the function of humour With many behavioural systems, the rele- pairwise interactions. is to relieve tension”, “music evolved to vant environment consists not of predictable In 1989 the linguist James Hurford12 used foster group cohesion” — can seem glib things like light waves and snakes but of evolutionary game theory to show that a and lame. other behaving individuals co-evolving their defining property of human language, the Fortunately, good theories of adaptation own strategies. Evolutionary game theory11 arbitrary, bidirectional sign (in which a com- can be distinguished from bad ones10. The has allowed biologists to analyse such munity of speakers and hearers tacitly agree bad ones try to explain one bit of our psy- dynamics, predicting how Darwinian org- to use particular signals to refer to particular chology (say, humour or music) by appeal- anisms ought to treat their antagonists, concepts), will drive out other schemes over ing to some other, equally mysterious bit mates and comrades. evolutionary time. Nowak and collaborators (laughing makes you feel better; people like Language, like sex, aggression and coop- have done the same for two other defining to make music with other people). The good eration, is a game it takes two to play. Many properties of language. ones use some independently established origins have been proposed for it — a substi- Last year, Nowak and David Krakauer6 finding of or mathematics to show tute for grooming, a way to tattle on adulter- pointed out that real organisms never trans- that some mechanism can attain some goal ers, a by-product of a big head — but they mit information perfectly: errors in sig- in some environment. For instance, projec- have been based on intuitive appeal. Game nalling or perception are inevitable, espe- tive geometry shows how stereoscopic imag- theory can provide the external criteria for cially when signals are physically similar ing can compute the third dimension; her- utility enjoyed by the rest of evolutionary (Fig. 1). Imagine organisms that use a differ- petology has documented that many snakes biology. Modellers make the minimal ent sound (say, a vowel) for every concept are venomous. These engineering bench- assumption that the transmission of infor- they wish to communicate. As they commu- © 2000 Macmillan Magazines Ltd | VOL 404 | 30 MARCH 2000 | www.nature.com 441 news and views

nicate more concepts, they will need addi- cry from showing how it did evolve in this 1. Pinker, S. & Bloom, P. Behav. Brain Sci. 13, 707–784 (1992). tional sounds, which will be physically closer world. But the game theorists have demon- 2. Pinker, S. The Language Instinct (HarperCollins, New York, and hence harder to discriminate. At some strated the evolvability of the most striking 1994). 3. Hurford, J. R., Studdert-Kennedy, M. & Knight, C. (eds) point adding new signals just makes the features of language — arbitrary signs and Approaches to the Evolution of Language: Social and Cognitive whole repertoire more confusable and fails duality of patterning — breaking the logical Bases (Cambridge Univ. Press, New York, 1998). to increase its net communicative power. circle that cripples many evolutionary expla- 4. Deacon, T. W. The Symbolic Species: The Co-evolution of Nowak and Krakauer showed that this nations. And by showing how the abstract Language and the Brain (Norton, New York, 1997). 5. Calvin, W. H. & Bickerton, D. Lingua ex Machina: Reconciling limitation can be overcome by capping the components of language (lexicon, phonolo- Darwin and Chomsky with the Human Brain (MIT Press, number of signals and stringing them gy, syntax and semantics) might be related to Cambridge, Massachusetts, 2000). together into sequences, with one sequence more tangible skills, such as the articulation 6. Nowak, M. A. & Krakauer, D. C. Proc. Natl Acad. Sci. USA 96, per concept. The sequences are what we call and perception of speech and the cognitive 8028–8033 (1999). 7. Nowak, M. A., Krakauer, D. C. & Dress, A. Proc. R. Soc. Lond. B words, and the combination of meaningless categorization of the world, they suggest 266, 2131–2136 (1999). vowels and consonants into meaningful ways of connecting the evolution of language 8. Nowak, M. A., Plotkin, J. B. & Krakauer, D. C. J. Theor. Biol. words (d + o + g = dog) by rules of phonology to other topics in human evolution, allowing 200, 147–162 (1999). 9. Nowak, M. A., Plotkin, J. B. & Jansen, V. A. A. Nature 404, is a universal property of language: half of each to constrain the others. Perhaps some 495–498 (2000). what linguists call ‘duality of patterning’. day the Société will reverse its ban. ■ 10.Williams, G. C. Adaptation and Natural Selection: A Critique of Nowak and Krakauer demonstrated that its Steven Pinker is in the Department of Brain and Some Current Evolutionary Thought (Princeton Univ. Press, evolution is likely among communicators Cognitive Sciences, Massachusetts Institute of 1966). 11.Maynard Smith, J. Evolution and the Theory of Games with a large number of messages to convey, a Technology, Cambridge, Massachusetts 02193, USA. (Cambridge Univ. Press, New York, 1982). precondition that plausibly characterizes e-mail: [email protected] 12.Hurford, J. R. Lingua 77, 187–222 (1989). our brainy ancestors. Now, Nowak, JoshuaPlotkin and Vincent Jansen9 analyse another hallmark. For a word Genome sequencing to survive in a community, it must be used frequently enough to be heard and remem- bered by all the learners. As new words are A view of Mount Drosophila added to the vocabularies of speakers, old Jonathan Hodgkin words must be used less often, and they are liable to fade, leaving the language no more expressive than before. iologists live in privileged times. In the Nowak et al. point out that this limitation next few years, the complete genome can be overcome by communicators who use Bsequences of representatives from all compositional syntax: rather than pairing the major groups of organisms on Earth will each word with an entire event, they pair be determined. Like a range of mountains each word with a component of an event (a emerging from cloud, successively larger and participant, an action, a relationship), and larger genomes can now be seen in their string the words together in an order that entirety, as each major genome project reflects their roles (for example, dog bites reaches its goals. Every one of these projects man). Such communicators need not mem- creates a gigantic and breathtaking archive of orize a word for every event, reducing the solid data and new information, which can word-learning burden and allowing them to be analysed in detail for decades to come. LIBRARY PHOTO DR JEREMY BURGESS/SCIENCE talk about events that lack words. Syntax and Nothing like this has happened in previous Figure 1 In the limelight — the fruitfly semantics, the other half of the duality of scientific history, or is likely to happen again. Drosophila melanogaster. The sequence of the patterning, will evolve. The latest grand achievement in genome of Drosophila has now been published1, Nowak et al.9 note that syntax has a genomics has a particular significance for giving geneticists invaluable new information cost: the requirement to attend to the order the science of . Almost a century of about their favourite organism. of words. Its benefits exceed the costs only research on the fruitfly Drosophila melano- when the number of events worth communi- gaster (Fig. 1) has culminated in the publica- licly funded human genome consortium. It cating exceeds a threshold. This ‘syntax tion of 120 million base pairs (Mb) of is still unclear whether the shotgun strategy threshold’ is most likely to be crossed when sequence1, and it seems to include the entire will work when applied to the much larger the environment, as perceived by the com- protein-coding part of the fly genome. genomes of vertebrates and economically municators, has a combinatorial structure, Although a number of bacterial (prokary- important plants4. for example, when any of a number of actors otic) genomes have been sequenced, this is The sequence determined does not corre- (dogs, cats, men, women, children) can only the third eukaryotic genome to arrive at spond to the whole 180 Mb that make up the engage in any of a number of actions (walk- this level of coverage. It follows the 12 Mb of four chromosomes of the fly, and it still con- ing, running, sleeping, biting)1. In such a the yeast Saccharomyces cerevisiae2 and the tains gaps and some stretches of low quality. world, the number of words that have to be 97 Mb of the nematode worm Caenorhabdi- The other 60 Mb (technically, heterochro- learned by a syntactic communicator equals tis elegans3, and is celebrated in a whole series matin) remains largely uncloned, unse- the sum of the number of actors, actions, of papers in Science1,4–9. These report both quenced and neglected because most of it places and so on, whereas the number that the technical aspects of the work and some of consists only of monotonous repeats and must be learned by a nonsyntactic com- the first insights to emerge from a complete dead transposons. municator equals their product, a poten- panorama of the genome. As with most other animal and plant tially unlearnable number. Syntax is handy Technical issues will not be discussed fur- genomes, determining every last nucleotide to an analytical mind in a combinatorial ther here, but are especially important in in the sequence is not justified. The authors world. view of the whole-genome shotgun strategy declare their sequence “substantially com- Of course, showing how language could used for this project, in contrast to the clone- plete”, in contrast to the “essentially com- have evolved in a hypothetical world is a far by-clone approach being used by the pub- plete” sequence of C. elegans, of which more © 2000 Macmillan Magazines Ltd 442 NATURE | VOL 404 | 30 MARCH 2000 | www.nature.com