An empirical investigation of pause notation

The current article examines the limitations of perceptual judgment for the transcription of pause occurrence and duration. This investigation extends Kowal and O’Connell’s (2000) previous research, which examined pause occurrence and duration notated by perceptual judgment with measurements from acoustic instruments for three corpora. Kowal and O’Connell found that, across corpora, there were a notable number of errors in perceptual detection of pauses, including both failure to notate pauses measurable with instrumentation (misses) and notation of pauses that were not measurable (false positives). In this study we focus on pause transcription in a uniquely English-language database, examining four excerpts from the London-Lund Corpus (LLC). Pauses which had been notated perceptually in LLC were compared with pause measurements from a Siemens Oscillomink L. As in the previous research, it was found that a notable number of pauses detectable with the acoustic instruments were not notated in LLC, which relied on only perceptual judgment. Errors in pause detection, both false positives and misses, accounted for 86 cases over 257 perceptually notated pauses, an error rate of one in three. We also examined two assumptions of LLC: (1) The assumption that perceptually notated pauses would adhere to interval scaling was not substantiated by instrumental measurement. (2) The assumption (Crystal and Quirk 1964: 49) that “impressionistic relative length varies with the tempo norm of a given speaker” was also not substantiated insofar as all correlations of tempo (operationalized as articulation rate) with ambient pause duration were nonsignificant.