Abstract

The following investigation is part of a !arger psycholinguistic project on the use of transcription systems in research on spoken discourse. Notation of pause occurrence and duration has become common in such transcription. To date, however, there has been a Jack of empirical research to evaluate the accuracy and efficiency of such notations based on transcribers' perceptual reports. The present research examines one English and two German excerpts of spoken discourse transcribed according to the prosodic transcription of the Survey of English Usage (SEU), the Halbinterpretative Arbeitstranskriptionen (HIAT), and the Gesprächsanalytisches Transkriptionssystem (GAT), respectively. The tape recordings were subjected to acoustic measurement and the measured pauses compared with the pauses notated on the basis of perceptual judgement. This comparison revealed that long (::: 1.0 sec) and medium(> 0.30 - < 1.0 sec) pauses were transcribed relatively accurately, but short pauses (::: 0.12 - $; 0.30 sec) were not; both false alarms and misses occurred. Scaling of pause duration was found to be ordinal, but not in accord with the interval and ratio scales specified within the SEU and GAT notation systems, respectively. The results suggest that the three transcription systems overburden even trained transcribers beyond their capacity.