It is well known that the speech stream contains rapidly unfolding, locally ambiguous information. Human infants and adults are highly sensitive to statistical regularities in linguistic input. Recent research has demonstrated that skilled language users capitalize on myriad sources of information, including statistical regularities, to make specific, real-time predictions about upcoming linguistic events. Work in other domains, such as probability learning, has shown that predictions are a rich substrate for learning: Mismatches between predicted and actual outcomes generate error signals, which can then be used to refine future predictions. Little is known, however, about how this process of refinement takes place, how it applies to predictions made during language processing, and how it changes over development. The proposed research is a novel set of studies exploring the hypothesis that error-driven learning is a domain-general learning mechanism that subserves linguistic and nonlinguistic processing and learning. The three aims of the proposal are 1) to determine whether error signals are incorporated into behavior immediately in linguistic and nonlinguistic tasks, 2) to determine whether behavioral change driven by error signals is sensitive to the magnitude of the expectation that led to the signal and 3) compare behavioral change driven by error signals in infants and adults. We will use a probability learning framework, in which adult and infant participants will learn probabilistic associations between (linguistic or nonlinguistic) elements. We will use reaction times (adults) and anticipatory eye movements (infants and adults) to measure participants'learning and adaptation to unexpected elements.

Public Health Relevance

The goal of the proposed research is to investigate error-driven learning processes in linguistic and nonlinguistic contexts in typically developing infants and young adults. Error-driven learning is a basic, domain-general mechanism that could help learners track statistical regularities in language, a skill important for multiple components of language acquisition and use. Understanding the role of error-driven learning in language processing and acquisition is essential to a complete picture of language learning mechanisms, and will allow researchers to determine how disruption of those mechanisms leads to atypical language development and use.

Agency
National Institute of Health (NIH)
Institute
National Institute on Deafness and Other Communication Disorders (NIDCD)
Type
Predoctoral Individual National Research Service Award (F31)
Project #
5F31DC009940-03
Application #
8066360
Study Section
Communication Disorders Review Committee (CDRC)
Program Officer
Cyr, Janet
Project Start
2009-06-01
Project End
2011-08-31
Budget Start
2011-06-01
Budget End
2011-08-31
Support Year
3
Fiscal Year
2011
Total Cost
$11,154
Indirect Cost
Name
University of Wisconsin Madison
Department
Pediatrics
Type
Other Domestic Higher Education
DUNS #
161202122
City
Madison
State
WI
Country
United States
Zip Code
53715
Romberg, Alexa R; Saffran, Jenny R (2013) All together now: concurrent learning of multiple structures in an artificial language. Cogn Sci 37:1290-320
Romberg, Alexa R; Saffran, Jenny R (2010) Statistical learning and language acquisition. Wiley Interdiscip Rev Cogn Sci 1:906-914