Social psychological inquiries into attitude-behavior relations have come a long way. The pessimism of the early 1970s gave way to a decision theoretical reformulation of the question by the TRA and its variations and a process-oriented theory of attitude accessibility. Yet, neither formulation asked the question of where the behavior comes from in attitude-behavior relations. The behavior generation is an issue fundamental to any attempts at understanding behaviors. For an intention to perform a behavior to be formed, that behavior must come to one's mind. For an attitude to be used to decide whether a behavior is to be performed, that behavior must be generated to be selected. The behavior generation is an issue conceptually prior to the decision making.
It should come as no surprise that Doob ( 1947, p. 136) referenced the behavior generation process in his behaviorist approach to attitudes. In fact, his definition of attitude is almost what we have called behavior representations: "An attitude is an implicit response which is both (a) anticipatory and (b) mediating in reference to patterns of overt responses [emphasis added], which is evoked (a) by a variety of stimulus patterns (b) as a result of previous learning or of gradients of generalization and discrimination, which is itself cue- and drive-producing, and which is considered socially significant in the individual's society." Yet, our formulation is different from Doob's in that we used a connectionist model of mental processes in modeling the behavior generation. In Doob's formulation, it is unclear how responses are generated, whether serially one by one or in parallel processes. By contrast, we postulated parallel processes in which a composite behavior representation is generated.
Clearly, the current formulation requires further specification and development. Most of all, the theory needs to be tested empirically. Perhaps our formulation should be taken as a call to pay closer attention to action. The inquiry into attitude-behavior relations is one of very few areas in contemporary social psychology, if any, in which the prediction of action is an explicit aim. A strongly action-oriented approach to the issue may be needed. Miller, Galanter, and Pribram ( 1960) in their Plans and the Structure of Behavior, criticized Tolman's cognitive map approach to learning (p. 9):
It is so transparently clear to [cognitive theorists] that if a hungry rat knows where to find food . . . he will go there and eat. What more is there to explain? The answer, of course, is that a great deal is left to be explained. The gap from knowledge to action looks smaller than the gap from stimulus to action--yet the gap is still there, still indefinitely large. Tolman . . . leaps over that gap when he infers the rat's cognitive organization from its behavior. But that leaves still outstanding the question of the rat's ability to leap it.