Introduction: Behavior-based safety programs have been well tested under controlled circumstances but less is published on their effectiveness in uncontrolled conditions. The aims reported in this paper were to evaluate a behavior-based safety program in terms of effectiveness to improve safe work practices and good housekeeping, and the effect of mode of administration upon program effectiveness. The compliance of companies in implementing the intervention was also investigated.Methods: Nine work areas in seven companies were selected in Western Australia A quasi-experimental within-group before-after design with a multiple baseline was employed. The behavior-based safety intervention was based closely upon that described by Komaki, Barwick and Scott, J. Appl. Psychol. 63(4) (1978) 434-445. Baseline observations continued for a minimum of four weeks and post-intervention observations for a minimum of ten weeks. Observations were conducted by employees, and feedback administered jointly by the company and the researchers. Observer agreement was monitored throughout. The outcome variables were the safe practice rate and the good housekeeping rate.Results: Three out of nine work areas demonstrated a statistically significant improvement in both safe practices and good housekeeping. Three had a significant improvement in housekeeping but a nan-significant improvement in safe practices. One was worse in safe practices and improved in housekeeping (both non-significant). Two were worse in both safe practices and housekeeping (not significant). There was no association between programs coordinated by employees on the floor versus supervisor-coordinated programs.Discussion: The behavior-based safety program did not perform as well in this trial as has been previously documented. Further field trials in uncontrolled industrial conditions are needed to better understand the practical usefulness of behavioral-based safetyprograms. (C) 1997 Elsevier Science Ltd.