What are cognates?

Abstract

The popularity of computational methods in historical linguistics has primarily been motivated by mere access to the new methods themselves, rather than by looking for tools to solve problems. Investigators have looked for problems with which to showcase their tools. This dynamic is one reason why eye‐catching but long solved problems, such as the homeland of the Indo‐Europeans (Gray & Atkinson 2003) have received more attention than genuinely unsolved or controversial questions, such as how to incorporate the Hittite ḫi‐conjugation into an understanding of the Indo‐European verbal system (Jasanoff 2003). One assumption of Bayesian methods is that cognacy can be conceptualized as binary. Although this is how historical linguists themselves often speak, it is not how they work. The goal of this essay is to more precisely delimit what is meant when we call two words cognate, to emphasize that this is not a binary relation, but to suggest that this relationship can still be modeled formally.