Mini-Seminar: Models, data, and their limits: What can we know about the deep history of language families? with Philipp Rönchen

Abstract

Computational methods are increasingly used to reconstruct the deep history of language families, yet different models often produce strikingly different answers. This reflects a general challenge in the historical sciences: we must draw conclusions from fragmentary data shaped by complex processes that cannot be directly observed or experimentally repeated.

In this talk, I summarise the main ideas of my PhD thesis, which examines how computational methods can be evaluated rather than simply applied. I argue that good statistical fit to the available data is not, on its own, evidence of reliable historical inference, because results depend strongly on modelling assumptions and on how we represent processes of change. I will discuss why more elaborate models do not necessarily yield more trustworthy conclusions, and how simulation can be used to probe the robustness of our inferences and clarify what computational methods can, and cannot, tell us about the past.

Philipp Rönchen is a PhD candidate in general linguistics at the Department of Linguistics and Philology, Uppsala University