The intent of the present study was to assess the effects of discrimination training only and in combination with frequent safety observations on five participants' safety-related behavior in a simulated office setting. The study used a multiple-baseline design across safety-related behaviors. Across all participants and behavior, safety improved from 53% during the baseline phase to 74% during the training phase, and improved further to 89% during the training plus observation phase. The results show that both discrimination training only and in combination with frequent observations improves safety-related behavior with a medium effect size. The study also measured participants' accuracy of observation, which was the extent of their correct classification of safe and at-risk behavior. Participants' accuracy of observation was above 80% and had little correlation with safety-related behavior. Based on the various measures, conducting observations appears to serve more than a skill acquisition function for the safety-related behavior. (Contains 6 figures and 3 tables.)