In order to discover and develop drugs to treat cognitive decline, animal tests are needed which will predict clinical activity. The process of developing such animal models does not differ, in principle, from the approaches used in other areas of psychotropic drug research. One research approach involves attempts to create models homologous with the disorder to be treated. This requires a thorough analysis of the clinical problem and an attempt to reproduce the biological and behavioral aspects of the disorder in animals. Recent studies of the effects of certain brain lesions on learning in monkeys or rats as models of Alzheimer's disease may go some way towards developing an animal model for this disorder. However, it is arguable whether this approach to "modelling" has ever had much success in psychopharmacology. A more pragmatic strategy involves the development of what are sometimes called emperical models, according to which any biological and behavioral test can be used if it provides a reasonable prediction of activity in the clinic. For example, the much used passive avoidance test in rodents does not need to model human cognition if it accurately predicts clinical activity. The extent to which this approach can be used successfully for evaluating cognition-enhancing drugs is discussed in view of the many drugs already marketed as cognition enhancers.