A comparison of techniques for measuring intelligibility of dysarthric speech

Eight techniques for quantifying intelligibility of dysarthric speech were compared. Eight dysarthric speakers who represented a wide range of severity were recorded producing single words and sentences. Thirty-two college students performed the following intelligibility quantification tasks: percentage estimates, rating scale estimates, work and sentence transcriptions, word and sentence completions, and word and sentence multiple-choice tasks. Intelligibility scores for transcriptions were compared to estimates and to other objective tasks with the following results: (1) all measurement techniques, except word completion, rank ordered speakers similarly to transcriptions, (2) mean estimates of intelligibility closely parallel transcription scores, but dispersion of listener estimates was large, and (3) objective tasks form a hierarchy with speakers receiving lowest scores on transcriptions, intermediate scores on completions, and highest scores on multiple-choice tasks. Mean scores for words and sentences were similar. Implications of results for clinical management of dysarthria are discussed.