School of Communication at Northwestern University :: Communication Sciences and Disorders

Learning phoneme categories without minimal pairs.

Learning phonemes: How far can the input take us?

Maye, Jessica, & LouAnn Gerken (2000). Learning phoneme categories without minimal pairs. Proceedings of the 24th Annual Boston University Conference on Language Development.

The present study investigates the effect of the statistical distribution of phonetic variation in the speech signal on listeners’ ability to discriminate a pair of speech sounds. We presented adult subjects with speech sound stimuli exemplifying a phonetic contrast that might potentially be used phonemically in some foreign language. The speech stimuli were resynthesized from natural speech into a phonetic continuum. During the familiarization phase of the experiment, subjects passively listened to the continuum stimuli. Subjects were randomly assigned to one of two familiarization conditions: in the bimodal condition, stimuli near the endpoints of the continuum were presented most frequently; in the unimodal condition, stimuli from the center of the continuum were presented most frequently. Subjects were then given a modified AX discrimination task in which they were tested on their discrimination of the endpoint stimuli. The bimodal group displayed greater discrimination than the unimodal group, demonstrating that brief, passive exposure to a phonetic distribution affects adults' discrimination of speech sounds.

Maye, Jessica, & LouAnn Gerken (2001). Learning phonemes: How far can the input take us? Proceedings of the 25th Annual Boston University Conference on Language Development.

In this paper we discuss the role of contrastive features in phoneme acquisition. We conducted an experiment in which participants were presented with an artificial language containing a potential phoneme contrast (e.g. /d/~/D/). They were then tested on an analogous pair of stimuli not presented during training (e.g. /g/~/G/). Training did not generalize to the untrained stimuli, indicating that mere exposure to one example of a featural contrast is insufficient for generalization of that feature to new examples of the contrast. These results demonstrate that language learners do not come to the task with the assumption that they will encounter symmetrical featural systems. We conclude by presenting testable hypotheses to account for how the featural system of a language might be learned.