Clinical diagnostic procedures to distinguish stuttering from non-stuttering individuals partially rely on the observation of characteristics of stuttered words or syllables. The purpose of this study was to examine the sensitivity and specificity of such a procedure, and to describe the observed characteristics of stuttered words. Methods involved the recording of the frequency, duration, and physical tension of stuttered words in conversational speech samples (total 43 100 words) and in an oral reading task (total 43 100 words). In this manner, 351 native Dutch speaking people, aged 8-53, who stutter were compared to a group of non-stuttering peers (n = 80). Findings suggest that applying a criterion of 3% stuttered words in conversational speech makes it possible to distinguish stuttering from non-stuttering individuals with high sensitivity (0.9345) and perfect specificity (1.0000). Frequency, duration, and physical tension accompanying stuttered words are described for the group of stuttering participants, and these characteristics were compared to the group of non-stuttering speakers. (Contains 2 tables and 3 figures.)