Stuttered and normal speech events in early childhood: the validity of a behavioral data language.

A set of 200 utterances from stuttering and normally speaking children aged 2-4 years was obtained. Each utterance contained a disfluency. A group of 5 sophisticated listeners assigned one of Johnson's eight disfluency categories to each of the 200 utterances. These clinicians showed poor agreement in the categories they assigned. Subsequently, the 200 disfluencies were presented to a group of generalist clinician listeners and a group of unsophisticated listeners, who were asked to judge whether each disfluency was "stuttering" or "normal." The disfluencies judged with high agreement to be "stuttering" and the disfluencies judged with high agreement to be "normal" were not categorically distinguished by the disfluency categories assigned previously by the sophisticated listeners. Further, judged presence of various disfluency categories accounted for only a small portion of the variance in numbers of "stuttering" judgments assigned to disfluencies. It is concluded that it is justifiable to question the validity of the data language used by researchers to describe stuttered and normal speech in early childhood. Several implications of this conclusion are discussed.