[Layout still needs to be changed; right now just uses blog post layout]

My overarching goal is to develop an understanding of language as a whole, stitching together the various components of the system that are usually studied in isolation. My current research interests can be categorized into two main lines of inquiry: language learning and change.

First, I am interested in an approach to language learning where a learner builds up their complex knowledge of language by generalizing patterns from the input data in an incremental fashion. This is in contrast to, for example, approaches that treat language learning as the process of selecting the correct grammar among existing hypotheses on the basis of input data. In these approaches, the burden of explaining the complexity of language is resolved by positing a complex initial state (often taken as part of the innate component of language). This is exemplified by the Principles & Parameters approach (Chomsky 1981), which takes learning to be the specification of a set of innate parameters, but is also implicitly assumed in many computational models of learning (e.g., Niyogi 2006 and Bayesian approaches). My approach instead shifts the explanatory burden onto learning itself; learners incrementally generalize rules from their input data, which can together create the complex system that is language. This is the view of learning that underlies Yang 2016 and related work.

The second area of my current interest lies in language change. Most of the time, a learner’s input data consists of the output of adult speakers with largely identical grammars; thus the learner will arrive at a largely identical grammar as well. However, if a learner’s input data is different, for various reasons [note TBW], they may arrive at a different grammar. This is the way in which individual learners carry out language change. I am interested in connecting this individual-level (acquisition-centered) understanding of language change to population-level behavior. More specifically, I hope to better understand historical phenomena through this framework of change, as well as develop statistical methods for population-level data in a more mechanistically meaningful way.