Individual and consensus judgments of disfluency types in the speech of persons who stutter.

Previous research has suggested that the reliability with which judges identify individual disfluency types, such as repetitions or prolongations of speech sounds, may be very poor. The use of disfluency types judgments in research and clinical applications is also complicated by important differences among the several disfluency-based characterizations of stuttered speech. In an attempt to address these problems, this study arranged for 30 judges to identify all disfluency types that they perceived to be present in 5-s audiovisually recorded speech stimuli, each in an Individual task and then with a partner in a Consensus task. Intrapair agreement and interpair agreement for occurrences of disfluency types (from Consensus conditions) were significantly higher than intrajudge and interjudge agreement for occurrences (from Individual conditions). Despite being higher than individual values, however, intrapair and interpair agreement for occurrences both averaged less than 50%. Results also showed that disfluency types judgments, interpreted in terms of three common disfluency-based definitions of stuttering, were not strongly related to previous assessments of whether these speech tokens contained or did not contain stuttering. When combined with previously available data, the present findings suggest caution in the use of disfluency types to describe or define stuttered speech.