Kellogg's research in conditioning and learning while at Indiana resulted in some 50 published articles. These papers, many of which describe work carried out in the Indiana Conditioning Laboratory, cover abroad range of topics. Among them are the bilateral transfer of conditioning, the necessity of making a motor response in order to condition that response, learning in dogs suffering varying degrees of cortical loss, the effects of various drugs on learning, the relationship of forward and back-ward conditioning, spinal conditioning, and the nature of the response inflexion conditioning. As well, there are reports describing methods and apparatus that Kellogg used in his conditioning work. It is outside the intent and scope of this essay to review all of Kellogg's research in this field. Instead, we shall attempt to present its general nature and purpose. This will serve to illustrate further the kind of scientist that Kellogg was and to help us understand what posterity's judgment of him is likely to be.

From the array of problems that Kellogg investigated, one may get the impression that his conditioning research had no focus. However, such an impression is mistaken. His chief concern was the nature of learning, and virtually all of his conditioning work was aimed at illuminating that issue. Kellogg's views on the matter are set out in five articles published in the Psychological Review from 1938 through 1940. In retrospect, one can see how his empirical research flowed from the ideas expressed in these papers. This is not to say that they contained a comprehensive theory of learning. Rather, they presented what Kellogg considered to be a scientifically useful conception of learning, that is, one that did not stray far from what could be observed, namely, behavior, and one that generated scientifically researchable questions. Moreover, the articles nicely illustrate Kellogg's atheoretical bias that we mentioned earlier .

The first three papers are the most important. In " An Eclectic View of Some Theories of Learning" Kellogg (1938b) attempted to minimize the differences among four different theories of learning by showing that each simply emphasized different parts of the learning situation and by highlighting continuities among them. Another paper (Kellogg, 1938c) criticized Cason (1937) for defining learning in part as the strengthening of neural connections. The themes of this article were sounded later at greater depth by Kellogg and Britt (1939). They argued for a definition of learning that stressed function (behavior change) and not structure (changes in the nervous system), as Cason had proposed. The major part of the essay raised objections to the structural viewpoint. Essentially, these were that the structural changes underlying learning were hypothetical and that the role played by the nervous system in learning was uncertain. In discussing the latter point, Kellogg and Britt mentioned the evidence of learning in dogs when the cerebral cortex was missing and the possibility of conditioning in spinal animals. Rather prophetically, both topics were ones that Kellogg was later to investigate in depth himself. In any case, it was concluded that a physiological definition of learning was far too speculative and should yield to one that emphasized changes in behavior or function. Behavioral changes were factual and observable; neurological changes amounted to little more than hypothetical inference.

The two final articles in the series (Kellogg, 1939, 1940) were brief replies to criticisms of the earlier papers. Neither reply is particularly substantive, although the later one is characteristically Kellogg. Chappell(1940) had attacked Kellogg and Britt's (1939) behavioral definition of learning. In response, Kellogg simply reiterated that the existing scientific data did not warrant a definition of learning in terms of changes in the nervous system, and, he concluded, " As far as I am concerned, that is all there is to it" (1940, p. 97).

We have said that a review of Kellogg's empirical work in conditioning and learning is beyond the scope of this essay. Nevertheless, it is appropriate to describe briefly one of the more significant of the learning projects, specifically, conditioning in spinal dogs. For one thing, Kellogg's interest in this problem may be clearly linked to his conceptual concerns about learning since, as noted above, spinal conditioning is explicitly mentioned in the Kellogg and Britt (1939) article. Second, major text books of the 1950s and 1960s (Kimble, 1961; Osgood, 1953; Stevens, 1951) gave prominent attention to Kellogg's spinal conditioning work. Third, the work brought Kellogg into controversy. Finally, the spinal conditioning publications indicate how accomplished Kellogg was as an empirical scientist.

The issue was whether dogs whose spinal cords had been transsected could acquire a conditioned response in a limb below the point of transsection. Observations pointing to this possibility had been first described by Culler (1937). Later, Shurrager and Culler (1940) reported data which they felt met the criteria of true motor conditioning and extinction in spinal dogs. The specific response conditioned was a muscle twitch in the exposed semitendinosis muscle in the dog's hind leg. Kellogg's first reference to the Shurrager-Culler data was in an article (Pronko & Kellogg, 1942) that described a muscle twitch in a limb when an electric shock was delivered to another limb. The animals involved were not spinal dogs, but the observation suggested to Kellogg the possibility that Shurrager and Culler's semitendinosis muscle twitch was not a true conditioned response.

Kellogg does not seem to have acted on this possibility immediately. In the years 1946 through 1949, however, he and his students published seven articles and delivered three oral presentations on the problem. The upshot was that spinal conditioning in dogs could not be produced in the Indiana laboratory. The most extensive reports were by Kellogg, Deese, Pronko,and Feinberg (1947) and by Deese and Kellogg (1949). In both articles, Kellogg concluded that the muscle twitch observed by Shurrager and Culler(1940) was actually a basic response to a conditioned electric shock stimulus applied to another part of the body (either to another limb or to the tail) and that an unconditioned electric shock stimulus applied to the limb in question was unnecessary. It was argued further that changes in this muscle twitch with training should be regarded simply as sensitization of a reflex.

Although Shurrager responded to Kellogg's claims (Shurrager, 1947) and continued his interest in the topic (e.g., Dykman & Shurrager, 1956),we shall not follow the story further. Our intent has been to describe the nature and extent of Kellogg's participation in the controversy. One point worth noting is that Shurrager's work was carried out with acute preparations whereas Kellogg's research employed chronic preparations. While it is possible that this was the reason for the difference in their results, we mention the point also as a tribute to Kellogg's laboratory skills. It was no simple matter to keep dogs alive for months after their spinal cords had been transsected. The data published with Deese in 1949 were Kellogg's last word on the problem. Convinced that he had done all the experimental work necessary and that spinal conditioning was not to be found in his animals, he turned to other concerns.