The rise of the cognitive approach in psychology can probably be dated to the publication of Miller, Galanter and Pribram (1960), while its position in social psychology was most clearly stated by Zajonc (1968a). Miller et al.'s book did not so much present any new data about psychology, but rather, it gave psychologists permission to speak about events which they could not observe. This was antithetical to the dominant behaviourisms of Hull and Spence, but not to that of Skinner (1974). Giving such permission had some benefits, but it also had some costs which are perhaps only now being appreciated in social psychology.

The benefit of speaking about unobservable events was that such events could be modelled in words, symbols or equations, and predictions made of observable behaviour. Such a strategy had helped the development of nuclear physics, for example, by allowing modelling of (then) unobservable atoms and electrons. Since being applied to social psychology, there has been a rapid proliferation of cognitive models (Markus and Zajonc, 1985).

The problem with cognitive models is that unless strict controls are kept on theorizing (see Skinner, 1950), the models can become underdetermined or indeterminate, such that many different models can be supported by the same observed data. We have, in fact, already seen some of this occurring in the last chapter, when each of the theories could account for the same data in seemingly different ways.