In modal training studies, the comparative groups differ in the use of a particular cognitive process; training to overcome the deficiency is provided for the deficient group only, and their performance is then compared to the standard set by the untrained group. Improvements are necessary in the design of training studies, as are a reconsideration of the types of skills that are the subject of training, and a serious consideration of the relation of the training and transfer tasks to the everyday functioning of the cognitive process under investigation. Specifically, an adequate training program must include the following: careful selection of the cognitive skill to be examined; sensitivity to the actual beginning competence of the learner; stringent analyses of the requirements of the training and transfer task so that transfer failures can be interpreted properly, training in multiple settings to alleviate the problem of welding, direct feedback concerning the effectiveness of the trained skill; and direct instruction concerning the generalization of the trained skills. Until a research program can provide answers to these basic questions, it is premature to conclude that retarded or deprived children and unschooled nonwestern children cannot show flexible effects of training. (DF)