Language change in individuals and populations

Historical linguists study changes that occur in the structure of languages - for example, changes in word order - that occur over many human generations. Psycholinguists meanwhile investigate (amongst other things) changes that arise from language learning and use. The link between the individual- and population-level dynamics is unclear, and provides an opportunity for investigation within the framework of statistical physics, which is set up to do the same sort of thing for condensed matter systems. In this talk I will argue that a very broad class of models of individual learning in populations predict that the system size (number of speakers) should have a big effect on the rate of historical language change, which is apparently not a feature of the empirical data. I will further propose that this conflict can be resolved if one augments language-learning biases that are universal to all humans with a set of socially-acquired biases that vary between cultures and over time.