A rating scale developed by Griffiths (1973) was used to assess four separate areas of the work behaviour of patients in a psychiatric rehabilitation unit. Assessments were made at an early and a late stage of rehabilitation. The ability of ratings on this scale to predict resettlement was confirmed in a prospective study. The scale was also used to provide information about four separate areas of work behaviour. Social relationships provided the strongest predictor of resettlement in employment, response to supervision and enthusiasm predicted only in psychotic patients; task competence failed to predict. Changes in work behaviour during rehabilitation, and the relationships between self-reported attitudes and work behaviour were also examined.