Harry (Hal) Tily

My research tries to understand how human languages come to be the way they are, by looking at what the humans who use them find easy or hard, natural or unnatural. Languages emerge through the interactions of people who need to solve common problems, and who all have similar cognitive capabilities for processing the input they receive from the world around them. I combine computational and statistical language modelling with behavioral experiments using human participants, with the goal of identifying the features of languages which make them well-suited for human communication.

I am currently working with Ted Gibson on an NSF funded project, evaluating models from machine learning and NLP as approximations of human language processing. I am also particularly interested in the question of why languages place their words in the orders they do, and how speakers make use of what flexibility their language gives them. My PhD dissertation looked at changes in word order fixedness in English over the last 1200 years, as well as word order variation in modern Japanese and English. I am now extending that work using several paradigms: behavioral measures of comprehension difficulty such as reading times; the learnability of specially constructed artificial languages; statistical modeling of speaker choice in large corpora; and computer simulation of language change.