The aim of this chapter is to consider the possibility that cognitive-behavioral theories can be developed to a point at which they will be able to predict at least some classes of behavior in at least some situations. Skinner (1990) pointed out that behaviorists have responded to an unsympathetic mainstream by developing their own separate professional organizations, journals, and terms of reference. Before endorsing this negative reaction, however, one should examine the possibility that mainstream research could begin to achieve consistency and accuracy at the descriptive level. If theories relating cognitions to behaviors were refined so that they made precise predictions, it might then be possible to debate the relevance of cognition to behavior. It might be possible, for example, to identify the types of circumstance in which particular cognitive variables are useful predictors of behavior and those in which they are not. Thus, without denying the basic problems inherent in models that require nonmaterial events to cause events in the material world (Eysenck, 1978), it might be possible to find a more limited but more useful role for these theories.