In Table 1, we summarize what is convincingly demonstrated to date for the major vertebrate and invertebrate model systems attempting to elucidate cellular mechanisms of associative learning. Two major concerns are the adequacy of the behavioral demonstrations and the completeness and extent of the accompanying neurophysiology. In addressing the issue of behavior, it is important to define clearly which criteria are both necessary and sufficient to infer the involvement of an associative-learning process. Similarly, it is also important to distinguish among those primary characteristics of associative learning in general, and those secondary or tertiary features that serve to define various subclasses. In our view, it would be unreasonable to require that any given preparation exhibit all the defining features of classical conditioning, for example, in order to qualify as a "legitimate" instance of associative learning. This is especially true if the goal is to understand the more general, rather than the specific, mechanisms involved in associative learning. Hence, we emphasize the following as primary features of learned behavior: pairing specificity, stimulus specificity, long-term retention (arbitrarily defined as lasting for at least 24 hr), a moderate degree of reversibility by subsequent experience (e.g. extinction), and demonstrations that nonassociative-learning processes cannot account for features a-c. Where appropriate, we also identified other interesting features of the learned behavior. It is apparent from the table that a major unresolved issue for most of the preparations is the extent to which the behavioral changes are exclusively associative. This is no less true for the vertebrate preparations than it is for the invertebrates. The clearest example of an exclusively associative behavioral change is the rabbit NMR. The learning-produced changes in the invertebrate preparations were all shown, to varying degrees, to be pairing specific. Yet a major unresolved issue is the degree to which apparent examples of associative-learning reflect complex interactions among basically nonassociative-learning processes. The core issue is really quite simple: Does the associative training procedure result in the acquisition of new or qualitatively different behavior; and is there a strict requirement for an associative relation? In addressing the adequacy of the neurophysiological analyses, the major issue is that of localization. Logically, there are two components to this.(ABSTRACT TRUNCATED AT 400 WORDS)