Evaluating Training During Training: Obstacles and Opportunities

on that rwinski, zlatively people tionistic .r ticular 3perties of both .teristics ch find-.ays that training T he conditions of training can be manipulated in a great variety of ways, producing multiple possible configurations of a given training program. Trainees' posttraining performance, in turn, is heavily influenced by such manipulations: Some configurations of the conditions of training are far better than others. It is of obvious importance to choose those configurations that optimize the conditions of training, but doing so depends on accurate assessment, which itself can pose formidable difficulties. One problem is that trainees' performance during training is an unreliable indicator of posttraining performance. Manipulations that enhance performance during training can yield poor long-term post-training performance, and other manipulations that seem to create difficulties and slow the rate of learning can be optimal in terms of long-term performance (see Christina & Bjork, 1991; Schmidt & Bjork, 1992). Another problem is that trainees' own subjective evaluations of their knowledge and capabilities can be misguided, leading them to prefer nonoptimal training regimens (see Bjork, 1994b; Jacoby, Bjork, & Kelley, 1994). Given those problems, it may seem obvious that one should assess training programs in terms of trainees' posttraining performance in the real-world settings that are the target of training. Optimizing on-the-job performance is, after all, the goal of typical training programs. For a variety of institutional reasons, however, assessing on-the-job performance in a meaningful way is often impractical, if not impossible. It is the mission of this chapter to consider a question of theoretical