Evaluating the effects of interventions is among the most critical issues faced by the field of human resource development today. This study addresses the potential of multisource feedback for training evaluation. The central research question is whether or not it is possible to measure change in work behavior, as a result of training, with multisource feedback in a reliable and valid way. A procedure for training evaluation was developed and three quasi-experimental studies were carried out. The psychometric properties of the multisource feedback instruments used are discussed as well as the training effects that were found. Our conclusion is that in most cases coworkers do not perceive positive changes in the work behavior of training participants. Whenever a training effect is observed, this is observed by the ratee's peers, rather than by the ratee's superiors or subordinates. Overall, the results do not support multisource feedback as an efficient method for training evaluation. However, this does not mean that multisource feedback cannot be used in a training context at all. Two alternative applications are presented.