Behaviorists often eschew explanations of performance (response rates and locations) that are based on psychological constructs (learning, preference, memory, impulsivity, etc.), for such constructs often lead to dualisms, homunculi, and pseudoexplanations. Their primary interest is in explaining behavior. For many others, however, this diet is too lean. Neuroscientists, for instance, often invoke psychological constructs to interpret performance. Their primary interest is psychological constructs, not behavior. Can there be a principled integration of these approaches? The conventional solution is to operationalize constructs (e.g., define preference as proportion of choices, memory as percent successes), but this often results in arbitrary and misleading indices that constrain commensurability to procedure fetishism. An alternative solution, advanced in this talk, consists of identifying constructs with hidden parameters, states, and processes defined with quantitative precision in generative models. These models specify random processes that generate instances of performance, thus providing the basis for explaining performance from psychological constructs, and for reverse engineering the principles that govern behavior. Basic psychological research, according to this approach, should focus on developing and testing generative models, and on estimating the sensitivity of model parameters to relevant manipulations. Empirical research on memory, learning, and impulsivity in animal models illustrate these ideas.

Review Federico Sanabria’s biographical statement.