Professor Berwick’s research spans the areas of learning, evolutionary biology, and complex systems, applied to the study of human cognition and its evolution, designing algorithms that can mimic human language acquisition in children and its evolution over time, as well as how human language might have first evolved.
In the area of language and its origins, along with Professor Shigeru Miyagawa of MIT, Prof. Berwick hypothesize that the origin of human language can be accounted for by the combination of two abilities that evolved independently: one analogous to the vocal learning of songbirds, and the other the ability to name objects. The reflex of these two distinct can be observed in modern human syntax. This research also connects human language to birdsong in a new way, by probing the commonalities in the rhythmic structure of both human language and birdsong as a ‘window’ into their common neurological and genetic basis.
In the area of language acquisition, Prof. Berwick investigates the role of statistical inference such as Bayesian analysis in the way that children learn to pair words with objects, and beyond that, more complex syntax. For analysis he uses actual videotapes of children interacting with parents. One of his findings is that complex graphical modeling does not seem to be required.
Tying language acquisition to language change, Prof. Berwick models the way in which languages can change over time, as dynamical systems. One key finding is that there are stable accumulation points that allow change in some directions, but not all. For example, English used to look more like German, but then changed to look more like French, in a way that is not easily reversible.