The roles of phoneme frequency, similarity, and availability in the experimental elicitation of speech errors

Abstract

In two experiments subjects read aloud pairs of nonsense syllables rapidly presented on a display screen or repeated the same syllables presented auditorily. The error patterns in both experiments showed significant asymmetry, thus lending support to explanations of the error generation process that consider certain phonemes to be “stronger” than others. Further error analyses revealed substantial effects of phoneme frequency in the language and effects of phoneme similarity, which depended on the feature system used to index similarity. Phoneme availability (the requirement that an intruding phoneme be part of the currently presented stimulus) was also important but not essential. We argue that the experimental elicitation of errors provides critical tests of hypotheses generated by the analysis of naturally occurring speech errors.