A nursing model selected by a population of ward managers was implemented on a long-stay psychiatric ward in Northern Ireland using an adapted action research approach within a quasi-experimental design. A range of dependent variables were appraised at one pre-test and two post-test evaluation points on an experimental (Ward X) and a control ward (Ward Y). A thorough review of the literature and methods used can be found in Part 1 (McKenna et al., 1995, Int. J. Nurs. Stud., 32, 79-94). The findings and their interpretations are presented here. Results indicate that on the experimental ward there were statistically significant improvements in "Psychiatric Monitor", patients' and staff's perception of ward atmosphere, patient satisfaction, staff's views about nursing models, and patient dependency levels. No significant changes were noted in nurse satisfaction levels nor nurses' perception of patients' behaviour. Particular emphasis is placed on the possible threats to the internal and external reliability of these findings and on the attempts made to control these threats.