Supervisor: Don Ringe
Dissertation proposal committee: Charles Yang (chair), Robin Clark, Rolf Noyer
A fundamental question regarding language is whether it has evolved through adaptation (Pinker & Bloom 1990, Jackendoff & Pinker 2005, Pinker & Jackendoff 2005) or is a byproduct of a single genetic mutation (Hauser et al. 2002, Fitch et al. 2005, Chomsky & Berwick 2015). The analogy between language evolution and biological evolution goes back to Darwin and has been applied frequently to this domain, in particular as an explanation of language change (e.g. Yang 2000, Niyogi & Berwick 2009). However, while there have been many attempts to build evolutionary models that model language change in terms of selection, fewer attempts have been made to formalize neutral models of evolution (cf. Baxter et al. 2009, Kauhanen 2017 and Newberry et al. 2017).
In this work, I address the question of whether sound change is better explained in terms of adaptation for communication purposes or in terms of stochastic processes. The starting points of my dissertation are the last three chapters of Labov (1994), where functionalist explanations of sound change are ruled out on the basis of empirical investigations: as the Neogrammarians put it, sound change is essentially a mechanical and phonetically determined process. The need to write a dissertation on a scientific debate which appeared settled after Labov's contributions arises after the emergence of a new `functionalist' wave, inspired by information theory and by mixed-effect regression models. The recent works of Piantadosi et al. (2011), Kaplan (2011), Cohen Priva (2012), Graff (2012), Wedel et al. (2013), Bouchard-Côté et al. (2013), Dautriche et al. (2017) and Mahowald et al. (2018) provide arguments in favor of functional pressures on language change, which in some cases (Sóskuthy 2013, Cohen Priva 2017) are presented as solutions to the actuation problem in Weinrich et al. (1968).
Two parallel strategies will be adopted to address this question. On the one hand, I will use historical data in every circumstance in which they are needed to determine the behavior of a model of sound change and to test the predictions of existing models. On the other hand, I will develop a null model of sound change, in order to establish a baseline against which claims and results can be evaluated.