Converging Models of the Influence of Experience upon Perception

We all understand what it means to speak a native language, but studies with infants show that we are native listeners too. Experience with our native language actually shapes the way in which we hear speech sounds. Until about 10 months of age, infants show a striking ability to hear differences between speech sounds, surpassing even adults in their ability to hear differences among foreign speech sounds. However, by their first birthday, infants are have become native listeners and, like adults, are able only to discriminate among speech sounds used by their native language. Japanese infants, for example, can discriminate [r] from [l] (as in "rock" and "lock") throughout their first few months of life despite the fact that [r] and [l] are not distinguished in the Japanese language. However, by the end of their first year Japanese infants, like Japanese adults, no longer discriminate [r] from [l]. They have become native listeners.

So far, research has been able to document this milestone of development across a number of languages and speech sounds. However, very little is known about the mechanisms that guide this change. How does experience with a native language change the way that we hear speech?

This is a difficult question and it remains unanswered. The rapidity with which infants' perception of speech changes has led many scientists to propose that infants are born with specialized mechanisms shaped by evolution to decode speech sounds. By this view, infants possess specialized brain mechanisms to recognize all of the possible speech sounds of the world's languages. This explains their early ability to hear differences between both foreign and native sounds. Experience with the native language then shapes these mechanisms via a "use it or lose it" principal. The regions not used by the native language are inactive and thus are lost, causing infants' decline in the ability to hear differences among foreign sounds.

This project is a departure from accounts that suggest infants are born with specialized brain mechanisms. The account favored in these studies relies only on the properties of the auditory system and general mechanisms of learning. To understand how such an approach might account for infants' declining ability to hear the difference between foreign sounds, it is helpful to consider what infants are hearing during their first year.

Across the world, there are over 5000 distinct languages. These languages make use of well over 800 different speech sounds (phonemes). Individual languages differ dramatically in the groups of phonemes they use to communicate meaning. English employs about 50 phonemes (phonemes are different from the letters of our alphabet - some letter combinations are spoken in several ways, thus there are more phonemes than letters). Different languages have different patterns. There exist languages that use as few as 11 and as many as 141 phonemes. The phonemes that a language uses shape the pattern of experience developing infants encounter in the first months of life.

The hypothesis of the present investigation is that experience with structured patterns of speech sounds present in the language encourages infants to categorize speech. Listeners encounter many different examples of a particular speech sound in any given day. There are subtle, but very important, differences in the way that people pronounce speech. When a woman says "rock", for example, the properties of the "r" are quite different from when a man or a child pronounces the same sound. Even within a single speaker, different pronunciations have different characteristics. One of the challenges for infants is to discover the commonalities across this variety of sounds in order to recognize all of the variants as examples of "r". Adding to the challenge of this task is the fact that infants must manage to categorize speech in a manner that is appropriate for their native language. An English baby must distinguish "r" from "l", but a Japanese infant must learn to treat these sounds as equivalent.

We believe that changes in the way infants hear foreign speech sounds is a result of learning to categorize speech. The present studies investigate whether categorization may emerge from experience with the patterns of speech sounds that vary across languages.

So far, tests of these predictions have not been undertaken because it is difficult and even unethical to manipulate infants' experience with speech. However, very new research has suggested an alternative means of testing these questions. Under the right experimental conditions, animal species (e.g., birds, monkeys and rodents) learn speech categories in a manner that is strikingly similar to humans. It is unlikely that these animals possess specialized brain mechanisms for decoding human speech. Thus, this evidence indicates that the mechanisms of speech category acquisition may be very general, extending even to animals.

This finding opens the door of opportunity for testing whether the loss of the ability to hear differences in foreign sounds is brought on by categorization. It is possible to very carefully manipulate animals experience with speech. As a result, the present studies have complete control over animal listeners' experience and can test how patterns of experience that model different languages influence perception. The goal is to determine whether animals exhibit changes in their ability to hear differences in speech sounds as a function of the patterns of the "native-language" sounds they experience in the experiment. Using this method, we can uncover the mechanisms of learning that are important.

Understanding these properties will have important implications. For example, one of the most difficult things about learning a second language as an adult is the simple fact that it is difficult even to hear the differences between foreign speech sounds. Finding out the means by which perception of speech is shaped by experience with a language will provide important clues about how to develop efficient ways to train second language learners. More generally, these data will contribute to our understanding of how experience influences the way we perceive the world.

For Further Information

The following articles address the issue of how infants perceive speech and how experience with a native language influences their perception. The articles are all accessible to a lay audience:

Werker, J. F. (1989). Becoming a native listener. American Scientist, 77, 54-59.

A very general introduction to the research that demonstrated infants' change in foreign speech discrimination.

Kuhl, P. K. (1987). Perception of speech and sound in early infancy. In P. Salapatek & L. Cohen (Eds.), Handbook of Infant Perception, Volume 2. New York: Academic Press.

A somewhat more advanced paper that discusses infant speech perception more broadly.

Holt, L. H., Lotto, A. J., & Kluender, K. R. (1998). Incorporating principals of general learning in theories of language acquisition, in M. Gruber, C. Derrick Higgins, K. S. Olson & T. Wysocki (Eds.), Chicago Linguistic Society, Volume 34: The Panels. Chicago: Chicago Linguistic Society, 253-268.

A more thorough general introduction to the ideas presented here