Perception requires interpretation. The information provided by the senses is often incomplete, conflicting, and noisy; we understand the world by evaluating sensory data and weighting it more or less heavily according to specific biases. Many of these adaptive biases are learned through experience. For example, we perceive speech according to our knowledge of the language's sounds, words, and grammar, and according to what we think the speaker is likely to be saying. All this interpretive guesswork is necessary partly because as listeners, we try to understand someone's speech while they are still talking; and as speakers, we only enunciate when we think we have to. This research investigates the use of adaptive perceptual biases by closely monitoring people's interpretation of speech. The studies use eyetracking techniques in which listeners' visual fixations to sets of pictures reveal what listeners think sentences mean. For example, in evaluating a word, how important is the precise acoustic realization relative to the likelihood of a given word? The research also studies how biases in language are learned by infants and young children, who are surprisingly proficient interpreters of speech. These developmental studies examine how infants and toddlers can use what they have learned about language so far to help them learn and understand more.

The research promotes understanding of how linguistic information of several kinds is brought to bear on interpretation. Speech is inherently ambiguous. What strategies do our minds recruit in perception? Understanding how normal adults interpret speech helps in the design of intervention strategies for patients with compromised language, second-language learners, and the hard of hearing; understanding how infants and children make use of their knowledge of sound structure in language and their growing vocabularies yields a better understanding of language acquisition, early communicative deficits, and phonological bases of reading difficulty. In addition, the research uses phonological learning and interpretation as a test case for understanding the temporal dynamics of human interpretation at multiple time scales: in the moments when speech is being heard; while learning of specific probabilities or associations proceeds; and while the infant grows into an adult.

Project Start
Project End
Budget Start
2004-09-01
Budget End
2010-08-31
Support Year
Fiscal Year
2004
Total Cost
$697,451
Indirect Cost
Name
University of Pennsylvania
Department
Type
DUNS #
City
Philadelphia
State
PA
Country
United States
Zip Code
19104