ABSTRACT
Because the Appendix provides a convenient synopsis of the match-mismatch theory presented in this book, I will let it serve in place of a retrospective summary. However, in closing, I would like to make a few observations about the relationship between theoretical constructions and empirical investigations in experimental psychology.
A distinction should be made between the theoretical approach followed in this book and the results of that approach as realized in the specific theoretical constructs which have been presented. The theoretical accuracy or heuristic usefulness of match-mismatch principles is dependent on my own particular capacities or weaknesses and any deficiencies in this respect do not necessarily imply that the methodological approach taken is without value. I have argued strongly on behalf of a process of convergent and divergent validation of theoretical constructs across different areas or levels of behavioral analysis. The success or failure of this particular implementation of that approach would not affect my own belief in its necessity for theoretical constructions in behavioral analysis.
To a great extent, psychologists have very demanding (or optimistic) views concerning the degree of prediction capable by a theory of behavior. There is not simply a demand for what might be called predictions of a probabilistic or statistical nature but rather for predictions about the precise behavior that will occur in each specific situation. This is not infeasible when the particular parameters of the situation have been studied; however, in some cases it may be difficult to predict the exact behavior that occurs on an abstract basis. To take the example mentioned in the introduction, it is possible to predict the rate at which an idealized object will cool, given a few basic variables such as temperature, pressure, etc. A physicist would be hard pressed, however, to predict the precise rate at which a cup of coffee will cool when placed on a kitchen table. The number of extraneous factors, of course, is vastly greater in the latter situation. As it turns out, the physical sciences can usually study or create situations which reduce the number of incidental variables and these artificial situations still can have great applicability to applied problems (e.g., a controlled environment is usually created for manufacturing processes involving chemical reactions). The problems that arise when such control is not possible can be seen in meteorology, a field, incidentally, which is rarely mentioned in discussions of the relative merits of the physical sciences over the behavioral sciences. Psychology, like meteorology, faces the problem that the particular situations for which prediction is desired are often those termed naturalistic or real-life where extraneous or idiosyncratic variables are rampant.
This is not a counsel of despair, but simply a statement that certain behavioral relationships may not be entirely due to the workings of an abstract principle which pervades the psychological domain, but rather to the operations of idiosyncratic organismic or environmental factors which cannot be estimated a priori. I should note that the domain of such factors is probably no greater in psychology than in any other area, but due to the focus of the field, it is possibly somewhat more visible.
It would be exceedingly helpful if psychologists would be less dominated by task variables in talking about psychological processes. Many tasks which are in general use are at best quasi-analytical since they have no specific relationship to structural processes or even to major contingent factors (e.g., passive avoidance, paired associate learning, etc.). The emphasis on task factors is understandable psychologically since these tend to be concrete while the underlying processes are abstract. The problem is amplified by the non-standardized nature of the procedures which can be used in any single task by different investigators, who nevertheless use the same label to name the task.
To state more positively the concerns expressed in the preceding paragraph, I would advocate that new tasks or procedures be developed which permit a more direct examination of the variables being measured. For example, if encoding processes are being studied, the task should permit the measurement of the encoding operations at the time they are occurring rather than estimate their characteristics indirectly at the time of retrieval when their effects are confounded with other processes (e.g., Johnston & Uhl, 1976). In the same vein, investigators should consider the process implications of the particular task procedures and modifications which they employ.
It should always be remembered that a structural process will exert an effect on behavior whenever the task conditions permit its expression, regardless of whether the task has been designed to measure that process.
