Distributional characteristics of simultaneous speech in conversation

Abstract

Seventy-two half-hour conversations—the interactions of 36 female dyads on two successive occasions—were automatically analyzed for instances of simultaneous speech (SS), defined as speech uttered by the participant who does not have the floor while the participant who does is talking. On the basis of outcome, SS segments were divided into two categories: nonencompassed SS(NSS) culminates in a change of which speaker has the floor while encompassed SS(ESS) does not. Duration frequencies of ESS and NSS segments were distributed exponentially over subjects. Comparing ESS and NSS occurrence frequencies in the first and second halves of the conversations yielded consistency estimates of 0.77 and 0.61 for occasion 1 and 0.72 and 0.54 for occasion 2. Comparing the occurrence frequencies of both occasions yielded stability estimates of 0.54 for ESS and 0.50 for NSS. However, comparing ESS and NSS frequencies over subjects and occasions yielded an r of 0.36, indicating the two categories share only about 13% of their variance. The results suggest that the frequencies with which speakers engage in ESS and NSS represent different but equally reliable aspects of their conversational styles, and provide a basis for examining lexical and psychological correlates of SS outcome.