Hull(ID:6950/)Scientific language for expression psychological theorems about rote learning Related languages
References: The work of C. L. Hull (and his associates) has been notably subject to external criticism. In his most substantial logical publication, Mathematico-Deductive Theory of Rote Learning, a system is presented requiring sixteen extralogical primitives and eighteen postulates with corollaries. The portion of psychological theory embraced is admittedly rather small, indeed almost infinitesimal when compared with the broad areas of mathematics formalized from relatively meager beginnings. It is not surprising, therefore, that some critics have expressed such reservations as, for example, "pseudo-clarity," "too costly rigor," "too costly clarity," and so on.?? It is easy to subject a formal system to a kind of eyebrowraising criticism in which one cites apparently damaging statistics. As in the case of Woodger, it is of much greater interest to probe into the why-and-wherefore of those statistics. Hull is subject, I believe, to the same basic criticism raised against Woodger. He has chosen to formalize an area of theory heavily reliant upon other areas of unformalized theory. Since that objection was elaborated in such detail for Woodger, however, we shall devote our attention in this section to a different topic. The chief criticism we shall argue against Hull (and many of his disciples) is that he has sought to mold his formal construction too rigidly into a preconceived pattern. In so doing, he has endeavored to apply certain generally accepted, valuable methodological principles outside their appropriate spheres of usefulness. As a result, his actual formal construction has suffered. Hull urges: (1) that "in any truly scientific system, all unobservables must be linked to one or more observables by unambiguous logical relationships;"?" and (2) that "qualitative" postulates are not so satisfactory as "quantitative" ones. Ultimate appeal to "unobservables" and the use of "qualitative" language are, then, to be avoided so far as possible. In his most important application of logic to the theory of rote learning, however, Hull does not succeed in totally eliminating the unobservables from the language, which we shall call H. ?Stimulus trace?, ?excitatory potential?, ?inhibitory potential?, and so forth, are all concepts falling within that category. He does indicate (perhaps motivated by Haldane?s proposal that Woodger?s Axiomatic Method should be rewritten, if possible, "in terms of observable phenomena only") that much effort has been expended with that general purpose in mind. Various remedies have, in fact, been propounded by others. Koch, for example (in remarks directed primarily toward Hull?s earlier "miniature" systems,"? but applicable to N also), suggests that by the use of Carnap?s reduction sentences such concepts can be defined operationally and the general difficulty alleviated. Unfortunately, the latter has never been accomplished. But does Hull fully appreciate the problem? Einstein remarks, in reply to a comment by Bridgman : "In order to be able to consider a logical system as physical theory it is not necessary to demand that all of its assertions can be independently interpreted and ?tested? ?operationally?; de facto this has never yet been achieved by any theory and cannot at all be achieved. In order to be able to consider a theory as a physical theory it is necessary only that it implies empirically testable assertions in general. This formulation is insofar entirely unprecise, as ?testability? is a quality which refers not merely to the assertion itself but also to the co-ordination of concepts, contained in it, with experience." Specifically, it is often asserted that the ideal scientific theory is one completely formalized, having only observables or operationally defined concepts as extralogical primitives. As indicated above in the remarks by Einstein, this is an ideal involving a basic extension to the essential function of a scientific theory, the latter being that it imply empirically testable assertions in general. If we endeavor to make the connection between our primitive concepts and the content of experience too immediate, as I believe Hull does, we may totally divorce from our construction those very fundamental, underlying statements and theoretical concepts which we normally expect any substantial theory to possess. I t is, in fact, generally taken as a sign of progress in a scientific theory that its more primitive language does not deal directly with every-day objects and "middle-sized?? facts. if one does literally attempt to start "in the middle," one cannot be expected to go far on only a small beginning; and the possibility of relating the specific area under advisement to other larger areas will be diminished. Abstract: The extent to which scientific languages can be formalized is an important problem, especially if it is assumed that a theorem-proving machine will deal most effectively with formal systems. In Part I, the axiomatic attempts of Woodger in genetics and of Hull in the theory of rote learning are examined. In Parts II and 111, to be published later, the more prominent efforts to formalize physical theory will be considered and a general study will be made of related questions. |