23 February 2014

Pater at Berkeley on Monday

Joe Pater will give the colloquium talk at University of California at Berkeley on Monday entitled “Structural Bias in Phonology.” An abstract of the talk follows.

Chomsky and Halle (1968) propose an Evaluation Procedure that prefers featurally simple rules and grammars, and Bach and Harms (1972) propose that this bias for structural simplicity can explain instances of historical change in terms of rule simplication in learning. In this talk, I discuss an alternative model of structural bias (Pater and Moreton 2012) cast in a Maximum Entropy framework (Goldwater and Johnson 2003, Hayes and Wilson 2008), which does not invoke an explicit simplicity metric. In the first part of this talk, I present this model, and the results of a phonotactic learning experiment that supports its predictions (Moreton et al. 2013). 

I also present the results of an ERP study (Moore-Cantwell et al. 2013) that provides insight into the nature of the phonotactic knowledge acquired in an experimental setting. After participants are trained on a small set of pattern-conforming words, we find that novel words that violate the pattern elicit a larger Late Positive Component (LPC) than the novel conforming items. LPCs have been observed in response to syntactic violations in language, and also violations of musical expectation. Finding an LPC rather than an N400 effect for the novel words is consistent with the view that participants in these experiments are forming an abstract generalization about the phonotactic pattern rather than directly judging the similarity of novel and trained words. It is also noteworthy that the response to a laboratory learned phonotactic constraint is similar to that for a naturally learned one (Domahs 2009).

In the last part of the talk, I present a model of how structural bias can impact phonological typology, which uses MaxEnt learning in the context of agent-based modeling. I present simulation results (Pater and Staubs 2013) that show the emergence of featural economy (Clements 2003) in systems produced by agent interaction. In these results the tendency towards simplicity is balanced by a tendency to maintain contrast between potentially homophonous words. No constraints or principles specifically demanding economy or contrast are required to obtain these results, suggesting that it is possible to maintain the standard view that phonological grammars evaluate individual representations, rather than entire systems.