The specification in concrete behavioral terms of theoretical constructs is one difficulty associated with behavioral assessment. For example, the construct of heterosocial skill has not been successfully defined in operational terms. A linear modeling technique was used to identify valid behavioral referents of heterosocial skill ratings for both men and women. Videotapes of heterosocial interactions of male (N=30) and female (N=30) university students representing a wide range of scores on the Social Avoidance and Distress Scale were rated by untrained peers (N=67), who then supplied lists of the behavioral cues they considered most useful in discriminating skill levels. The rate of occurrence of the most widely endorsed cues were scored for the videotaped subjects and used to construct linear models of the heterosocial skill ratings. Valid behavioral referents of molar heterosocial skill ratings were identified for both men and women. Men rated by female peers as high in heterosocial skill gazed upward, asked questions, and used appropriate hand gestures during conversations. Women rated by male peers as high in heterosocial skill also gazed upward, made eye contact, and avoided speaking too quietly. The findings suggest that the experimental method employed is able to identify behavioral referents of reliable peer ratings of molar heterosocial skills. (NRB)