Outcome measures for safety training effectiveness research often do not include measures such as occupational injury experience. Effectiveness mediators also receive sparse attention. A new safety training curriculum was delivered to workers in a stratified random sample of food service facilities across three companies. A similar group of facilities received usual training. We collected post-test measures of demographic variables, safety knowledge, perceptions of transfer of training climate, and workers' compensation claim data for one year after the initial training activities. Knowledge test scores were apparently higher in the new-training units than in the usual-training units. Some demographic variables were inconsistently associated with these differences. Evidence for reduction of the injury rate associated with the new training was observed from two companies but only approached significance for one company. A second company revealed a similar but non-significant trend. Knowledge scores were not significantly associated with lower injury rates. We found evidence that safety training increases knowledge and reduces injuries. We found almost no evidence of effects of mediators of training effectiveness, including no relationship between safety knowledge and injury experience. Methodological issues related to conducting a large study may have influenced these results. Impact on Industry: Although safety training leads to greater knowledge and, in some cases, reduced occupational injuries, the influence of mediating variables remains to be fully explained.