The Predictability of Words as a Function of Auditory and Contextual Cues

Abstract

To assess the effects on the predictability of speech of auditory context above and beyond structural environmental context, the “cloze” technique was used with aphasic and normal speech records. It was found both with regard to verbatim and form-class accuracy that subjects who listened to tape recordings while reading printed transcripts of the same utterances performed better than those provided only with the printed transcripts; linguistically trained listeners performed better than naive college-student listeners. In each instance the differences in performance, while consistent, were quite small. However, subjects rpovided only with a tape recording of normal speech performed far worse than those presented only with a printed transcript, or than those presented with both transcript and tape. These results are interpreted as indicating the preponderant importance of structural as compared to auditory cues in the successful prediction of missing words.