Presented by the ARC Centre of Excellence for the Dynamics of Language

There are a number of claims regarding why linguistic complexity varies, for example: i) different types of societal structure (e.g. Wray & Grace, 2007), ii) population size (e.g. Lupyan & Dale, 2010), and iii) the proportion of child vs. adult learners (e.g. Trudgill, 2011). The difficulties in defining and/or measuring linguistic complexity make it hard to assess these proposals. I will present a simple agent-based model which assumes that complexity is proportional to the amount of information which learners must acquire. The results partially support the conclusions of previous studies, but several subtleties arise. Firstly, differences in the capacity or opportunity to learn determine how much complexity is stable. Secondly, small populations are susceptible to large amounts of 'linguistic drift' and the subsequent loss of complexity, unless either innovation is frequent or learners are strongly disposed towards maintaining complexity. Conversely, large populations should in theory lead to more stable complexity, unless there is too much innovation, which leads to a paradoxical collapse in complexity. Third, large populations with many adult learners only lead to a collapse in complexity when learning is highly restricted. I will then discuss this analysis in the light of empirical work, and in particular recent studies of language change in Murrinhpatha (Mansfield, in preparation).