Self‐recording procedures were used by four adolescent girls to increase work and comments (cues) that evoked staff praise during vocational training sessions in a maximum‐security institution for offenders. The girls were selected on the basis of their not responding to a staff‐directed token program. The self‐recording procedures were directed by a therapist who saw the girls outside the vocational training sessions. According to a multiple‐baseline design, self‐recording of work was introduced sequentially to each of the two or three settings the girls attended each day. A few days after work had increased, self‐recording of cues was introduced. Tokens were delivered by the therapist for work and cues recorded by the girls. Work and cues increased following self‐recording for three of the girls and increased cues evoked higher rates of staff praise. Girl and staff behaviors were maintained during short follow‐up periods when tokens were not given for the girls' records. The procedures failed to effect desirable changes with a fourth girl's work, and self‐recording of work was terminated without