The rationale underlying the use of such a program with delinquent populations is that delinquent behavior can be viewed in terms of a failure to learn acceptable and appropriate means of responding in social situations. The subjects resided in a community home in Birmingham, England, were 10 to 15 years of age, and from lower class working families. Training was restricted to the basic skills of eye contact, lack of fiddling movements, appropriate head movements, attention feedback responses (such as 'yes' and 'yes I see'), and question-type feedback responses (such as 'did you?' and' really?'). Four skills were selected for each subject, and each skill was trained for two or three 30-minute training sessions. Training effects were assessed 3 days after each session and 2 weeks after the end of training. Assessment measures included a staff questionnaire, a self-report questionnaire concerning difficulties in social interractions, and direct behavioral observation which used videotaped conversations between each subject and a previously unknown adult. The program proved to be particularly effective in increasing eye contact and in decreasing fiddling movements. Training the use of head movements, acknowledgement, and question feedback responses was less successful, although some subjects showed greater improvements than others. The use of question-feedback responses may be too advanced a skill for adolescent offenders. Conclusions could not be drawn about the generalizability of the training effects outside the training environment. Data graphs and a 17-item reference list are included.