Measurement issues relating to behavior modeling research are discussed. Researchers have successfully demonstrated the ability of behavior modeling to produce significant behavioral change in the training milieu, but have been less successful in showing the transfer of such change to everyday situations, a problem partially caused by measurement deficiencies. Comparison of pre-training and post-training role playing is a common basis for assessment. Role-playing success may have potential for a high level of ecological validity, but this is not ascertainable unless assessment is based on multiple samples of behavior. Other problems commonly associated with behavior modeling research include: (1) inadequate behavior sampling; (2) insufficient consideration of lagged effects; and (3) inadequate attention to the problem of proactive inhibition. A four-page list of references is included. (SLD)