PsyLing Home Page

Brian MacWhinney, Professor of Psychology at Carnegie Mellon University

LDR papers | Papers | Talks | Vita | Photo | TalkBank |

Among the many puzzles that language presents to the developmental psychologist, one of the most fascinating is the relative ease with which a toddler picks up a first language. Although infants know nothing of the rules of grammar and have only a fragmentary understanding of the physical and social world, they are able to master the core structures of language by the age of three. The extent to which children can master both a first and second language contrasts with the more incomplete process of learning a second language in adulthood.

My approach to this problem views language acquisition as an emergent process. Instead of working with the traditional opposition between nativism and empiricism, I believe that we can better understand language learning as arising from competitive processes that operate across five major time/process scale groups, including scales for online processing, memory consolidation, development, social diffusion, and evolution.

For the time scale of online processing , I view utterances as providing cues that resolve the competition between alternative interpretations. Beginning in 1978, Elizabeth Bates and I worked with over 20 colleagues studying processing in 18 different languages to elaborate what we call the Competition Model. The Competition Model views language processing as a series of competitions between lexical items, phonological forms, and syntactic patterns. Competition Model studies have shown that learning of language forms is based on the accurate recording of many exposures to words and patterns in different contexts. If a pattern is reliably present in the adult input, the child picks it up quickly. Rare and unreliable patterns are learned late and are relatively weaker even in adults. Online processing is also responsive to the system of perspective shifting which allows the mind to construct an ongoing cognitive simulation based on the use of grammatical devices such as pronouns, case, voice, roles, and attachment.

For the time scale of memory consolidation , we can rely on connectionist models, such as the DevLex model I have worked on, to show how constructions form mutually supportive groups, allowing for resonance in memory retrieval.

For the time scale of development, we can examine language emergence in at least two ways. One methodology uses neural network models to simulate the acquisition of detailed grammatical structures. Beginning in 1989, I have worked on building connectionist models for the acquisition of morphology, syntax, and lexicon in English, German, and Hungarian. More recently, I have examined the ontogenetic emergence of language from a more biological viewpoint, using data on language processing from child with early focal lesions. The results of studies of these children using reaction-time methodologies and standardized tests indicate that, although they have completely normal functional use of language, detailed aspects of processing are slower in some cases. Using functional magnetic resonance imaging technology, we have pinpointed areas of activation involved in specific linguistic tasks. These results have allowed us to evaluate a series of hypotheses regarding sensitive periods for the emergence of language in the brain.

For the time/process scale of social diffusion , we can rely on ideas from sociolinguistics and corpus linguistics. For these the methods of the TalkBank databases I have been constructing can provide initial insights.

Finally, for the time scale of evolution , researchers have begun to examine the ways in which language has emerged gradually through competitive Darwinian processes, working within a series of cognitive, physical, and social constraints.. My work on perspective-taking, competition, and brain mechanisms suggests that the most likely account of the origin of language is one grounded on social mechanisms. In this sense, the elaboration of an emergent account of perspective-taking suggests a Vygotskyan approach to language evolution.

Papers on these topics can be found here.

In addition, to this empirical and theoretical work, I have helped construct the TalkBank system for the analysis of spoken language, the STEP system for teaching experimental psychology, and the PsyScope (Mac) and E-Prime (Windows) experiment generation systems. The classes I teach are:

My postal address is:

Brian MacWhinney
Department of Psychology 254M Baker
Carnegie Mellon University
5000 Forbes Ave.
Pittsburgh PA 15213
phone: 412 268-3793
GDPR Legal Information