One of the biggest problems currently facing the nursing profession is that of implementing the nursing process. Innovations in health care often elicit obstructions, and objective data which might help us to understand and overcome these implementation problems are lacking. The present study considered these problems by contrasting two groups of psychiatric nurses both of whom had received an in-service course in behaviour therapy. One group went on to implement this training by conducting behavioural nursing care plans with their patients, whereas the second group did not do so. A retrospective analysis of 25 variables associated with these groups of nurses indicated that neither subject variables, test scores nor learning during the course were good predictors of the nurses' implementations of the nursing process. An alternative basis for predicting and facilitating implementation is discussed. This focuses on institutional variables which may exert more control over the use of the nursing process than the personal characteristics of individual nurses. The implications of this analysis are important for the selection, training and motivation of nurses.