Language is interesting by itself. Understanding the hidden structure of language, both the things that make different languages distinctive and the shared structure across all languages, that’s exciting stuff. In this sense, I’m a lucky guy to get to research things that are obviously cool.

But what’s really neat about linguistic structure, to me, is that it isn’t some simple optimization problem. Language doesn’t just have to work; it has to work for us. We need to be able to learn language as babies and keep updating it, or even learning new languages, as we age. We need to be able to store and represent or language knowledge inside our limited brains. We need to use language communicatively, to achieve our social goals and express social identities. It’s a lot to put onto a system, and there’s so much about language that we still don’t understand.

In my research, I want to formalize and hopefully quantify the ways that the goals of language shape its form. If language needs to be learnable by babies, what influence does that have on its structures? If language needs to be used in real time, does that limit its possibilities? And what do our successes in using language reveal about the rest of us — our minds, our social goals, etc.?

I use probabilistic (generally Bayesian) mathematical models to try to probe into our language and our minds. I mostly work with bounded rational models, which examine the best possible ways of understanding language, and compare them against actual human performance. Perhaps surprisingly, people are often right up there with the rational models! And that’s great, because then we can use these models to examine what factors affect language use and cognition.

I’ll be porting my papers over to this site over time, but for now, you can check out my page on Google Scholar to see what I research.