Notation systems for transcription: an empirical investigation

A 21-syllable question posed by Bernard Shaw in a CNN television interview with Margaret Thatcher was presented to 90 participants, either as an audio recording or as a typed transcript or as both. Participants were asked to speak it, as closely as possible, as Shaw had (or, in conditions without the audio recording, as he might have). The typed version was either an ordinary transcript or a transcript in one of three transcription systems used currently in research on spoken discourse, all of which incorporate notations for prosody. Hence, there were nine conditions in all, with five women and five men in each. Contrary to the experimental hypothesis, approximations to Shaw's original temporal measures of performance were not degraded but were instead improved significantly by the addition of a prosodically notated transcript to the audio recording and significantly more in the absence of the audio recording. Presentation of the ordinary transcript alone produced the worst approximation to Shaw's temporal measures. The usefulness, accuracy, and readability of transcripts prepared according to detailed notation systems are discussed.