Fifteen-second samples of speech were recorded from 20 adult stutterers. The samples were played to two groups (I and NI) of unsophisticated observers. The 18 observers in the NI Group were instructed to underline any word that was stuttered. The 18 observers in the I Group also were instructed to underline stuttered words, but these observers were given a "standard definition" of stuttering (Wingate, 1964); repetition of a sound, syllable, or one-syllable word; silent or audible prolongation; or both. On the average, observers who were given a definition of stuttering marked more words as stuttered than observers who were told only to mark stutterings. In addition, the observers for whom stuttering was defined also displayed significantly more variability in terms of the average number of words underlined. Both interobserver and intraobserver agreement were significantly higher for the NI Group than the I Group observers. The data from this and other identification studies were used to support the suggestion that the identification of stuttering is a threshold phenomenon.