How are your mental strategies for understanding language influenced by your experience? Research suggests that people track their language experience, and they know for example that the more common interpretation of "ear" is as a body part and not a piece of corn. But a lot less is known about how language experience changes the way we interpret larger chunks of discourse. For example, an important part of language understanding is interpreting reference. In "Ana saw Liz. She waved", people need to figure out whether "She" refers to Ana or Liz. Do people pay attention to the frequency of referential patterns, and use this to interpret ambiguous pronouns? Recent findings suggest they might. In our example, most people tend to assume that "she" refers to Ana, who is the subject of the previous clause. But this bias toward the subject is weakened if people have recently read many stories where the non-subject gets re-mentioned, as when the pronoun is rendered unambiguous in context by the gender of the discourse participants (e.g., Matt saw Lyn. She waved.). This suggests that people shift their strategies on the basis of the most frequent reference patterns in context. But researchers do not know exactly how these biases are learned. For example, do people learn biases separately for each verb or verb class, or do they assume that frequencies of reference can be generalized across sentence types? How much exposure is necessary, and of what type?
The proposed project tests these questions with adults. Participants read numerous "priming" stories that display one type of reference, for example where the subject character gets re-mentioned, or the non-subject character gets mentioned. The critical question is whether this changes the way people interpret ambiguous pronouns, like "she" in the Ana/Liz example above. All priming stories use unambiguous reference, for example in mixed-gender stories so the pronoun referent is clear. The first set of experiments tests the types of generalizations people draw, for example across verbs or verb types, or based on syntactic vs. semantic information. The second set of experiments tests the conditions necessary for learning, for example the amount or type of input needed. This project is one of the first studies to examine how the frequency of discourse structures influences the strategies people use for pronoun comprehension. This builds on other recent evidence that language comprehension strategies can be learned from the frequencies of language input. This project will help demonstrate how the mind uses statistics from the environment to shape cognitive processing. Results will be important for developing models of how higher-level language comprehension is influenced by experience, and for developing interventions to address problems with poor reading and listening comprehension.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
