Abstract
The time required to transcribe and analyse lengthy conversation puts conversational discourse analysis (CDA) out of reach for most practising clinicians. However, standards have not been established for appropriate conversation sample size. Data are presented supporting the use of conversation samples of 5–10 minutes when studying conversation repair, speaking rate, and utterance length. Ten minute samples adequately represented ‘parent’ conversations from which they were derived when measuring conversation repair for six of eight cases. For measuring speaking efficiency (length of utterance and speaking rate) 5 minutes was adequate for all eight cases. For variables occurring once per minute, 10 minute samples were adequate, and for variables occurring three times per minute, 5 minute samples were adequate.