Predictability and Disruption of Spontaneous Speech

Abstract

Three experiments were conducted to test the general hypothesis that disrupted spontaneous speech is less predictable than fluent speech. Predictability was estimated by means of the Cloze procedure and disruption by Mahl's non- ah speech disturbance ratio. The hypothesis was not confirmed. Instead, the results revealed significant differences in predictability among speakers and as a function of the degree of verbal interaction and topical focus of the speech. Moreover, the effect of the latter conditions also varied from speaker to speaker. A significant interaction occurred such that the speech of certain speakers was more predictable when it was highly disrupted, while that of others was less predictable when highly disrupted.