Mark Pagel

Abstract. The words for some things change rapidly across languages whereas the words for some other meanings change very slowly if at all. I describe our recent work on models to measure the approximately 100-fold variation in rates of evolution we observe for different words in Indo-European languages, and our discovery that the frequency with which words are used in everyday speech explains this variation in rates of evolution. We have also discovered that languages appear to evolve in punctuational epochs or bursts associated with language splitting events, and I will describe a methodology for detecting this effect, and show some very recent results with Indo-European, Austronesian and Bantu languages. Some promising avenues for future research and modeling arise from these studies and from other observations on linguistic evolution. I describe four: i) we can now make predictions about a classes of ‘ultra-conserved’ linguistic characters that should be found in all human languages. These might inform studies seeking long-range relationships; ii) how does the ‘frequency effect’ influence the rates at which words evolve within populations of speakers? Does it alter the rate at which new words arise (in effect alter the mutation rate) or does it alter the probability that a new form will be accepted, conforming to an expectation of purifying selection acting against rare word forms; iii) the same words are likely to evolve at different rates in different languages and so flexible models of evolution are needed to detect and accommodate this effect; iv) words may evolve in a contingent fashion changing more rapidly in the presence of some other factor and vice versa. I describe one way this phenomenon might be modelled.