Is coevolution of language and language genes possible?

Nick Chater, Department of Psychology, University of Warwick

Morten Christiansen, Department of Psychology, Cornell University

Nick.chater@warwick.ac.uk, mhc27@cornell.edu

Languages change fast; genes changes slowly.  This presents a puzzle for
any theory of language origins which assumes that (1) human language
capacities are underpinned by genes for language specific constraints;
(2) that these genes arose by adaptation to the linguistic environment. 
The puzzle is that the linguistic environment would seem to be too
unstable to provide a consistent selection pressure on successive
generations of learners.  Moreover, because humans are geographically
distributed, and linguistic change leads to divergence across
linguistically unconnected or loosely connected populations, different
populations of humans would appear to be under very different
selectional pressure---i.e., pressure to be adapted to learn to cope
with their own linguistic environment.  Yet, it is typically assumed
that language-specific genes do not differ across populations, but are
instead universal---indeed, one motivation for proposing language genes
is that they provide an explanation for universal properties of the
world's languages.  These difficulties appear to undermine the idea the
adaptationist picture, according to which languages and language genes
co-evolved.  Instead, they suggest that language change is too variable
to provide an environment for stable selectional pressures across over
genes.  This paper explores these arguments with a series of computer
simulations of co-evolving populations: one population that mutates
rapidly (analogous to language); and one population that mutates slowly
(analogous to genes).  The fitness function of elements in each
population is determined by the "alignment" with the members of the
other population (i.e., a language is 'fit' if it is easily learned; a
learner is fit if it learns languages easily).  We also present some
results from simulations with spatial structure to their populations,
attempting to model the geographically distributed character of human
populations.  From our theoretical arguments, and the simulations we
present, we draw conclusions for the origin of language, arguing that
the fit between language and language learners primarily arises from the
rapid evolution of language to fit learners, rather than the other way
around.  These arguments present challenge for the adaptationist picture
of the development of putative language genes with which we began. 
Indeed, we suggest, unless some plausible alternative to the
adaptationist story can be developed, the very idea of genetically
encoded language universals becomes problematic, purely on evolutionary
grounds.