- UNDERGRADUATE PROGRAMS
- GRADUATE PROGRAMS
Language acquisition, infant speech perception, psycholinguistics
Beginning academic life as a linguist with interests in language processing and computation, I switched over in graduate school to become a (developmental) psychologist. In my youth, claims about innate bases and properties of language predominated. I am not altogether unsympathetic with that viewpoint, but it has always seemed to me that the most powerful argument for language preprogramming must be made by considering the strongest possible empirically supportable assumptions about richness of language input and the power of learners' perceptual, representational, and analytic capacities, and then determining specific aspects of language where these fall short. I have devoted my career to exploring the nature of language input (the auditory and, more recently, visual experiences of infants) and the nature of infants' language processing abilities. I have focused particularly on infants' spoken word recognition – a set of complex perceptual and computational skills fundamental for language comprehension and acquisition, involving arguably the most central unit of language structure.
I am interested in how the speech that infants and young children hear affects early language acquisition. My current research focuses on complementary questions of the nature of such speech, particularly with regard to properties that may cue aspects of language structure, and the nature of early perceptual capacities for extracting and representing the structural information that is cued. This work bears on the theoretical characterization of the initial state with regard to language learning: To the extent that input speech is rich in cues to structure that infants can represent appropriately, the need for imputing abstract grammatical knowledge to infants (as many recent theories have done) will be reduced.
Recent work in my laboratory on the properties of speech to infants shows that repeated tokens of words tend to be less phonetically variable (though more prosodically variable) in speech to infants than in speech to adults. This may aid infants in recognizing different tokens as exemplars of the same word and may assist in the establishment of the mental lexicon. Different grammatical categories of words possess varying constellations of acoustic, phonetic, and phonological properties; hence, these properties of speech may provide a basis for infants to begin to assign words to appropriate categories.
For her doctoral dissertation at Brown, Rushen Shi (now assistant professor at the University of British Columbia) showed that sets of cues in typologically distinctive languages are sufficient to allow untutored learners to categorize lexical and grammatical words accurately. At present, Heather Bortfeld, a National Research Service Award (NRSA) post-doctoral fellow, is investigating the prosodic and phonological correlates of the given/new distinction in English and Spanish infant-directed speech, and Leher Singh, a first-year graduate student, is analyzing the differences between "happy talk" and "baby talk" as the initial step toward studying whether infants' listening preferences are determined by speech affect or register. As a follow-up to her dissertation, Rushen Shi has been studying whether very young infants can categorically discriminate sets of lexical and grammatical words - categories that appear to be based on correlated acoustic/phonological properties. Shi, Werker, & Morgan (in submission) show that both 6-month-olds and 3-day-olds(!) succeed in this discrimination. Three-day-olds recover from habituation symmetrically, but 6-month-olds have already learned that lexical words are more interesting (or perhaps gramatical words aren't worth conciously attending to), so that they display asymmetric recovery from habituation.
Our work on infant speech perception focuses on the problems of how infants solve the word-segmentation problem and the nature of early lexical representations. Although words must initially be no more evident to an infant than they are to an adult listening to a foreign language, well before the end of the first year infants are capable of recognizing at least some words in fluent speech. Our studies show that infants exposed to English begin between 6 and 9 months to deploy a bias for the strong-weak pattern that predominates in the language in grouping syllables and segmenting words. Other studies show that between these two ages infants also begin to utilize language-specific knowledge of phonotactics (concerning permissible sequences of sounds and their relative likelihoods) as an additional means of identifying word boundaries. Ongoing research is examining in detail the development of infants' recognition of familiar words in fluent speech. Some of this research was described in an article in the Brown Alumni Monthly.