Limitations and Power for Generalization to Humans

One danger in using animal models is “overspecifying” what is being measured—that is, interpreting the animal’s behavior anthropomorphically, without measuring different facets of the behavior in order to clearly demonstrate what behavioral system is being measured. For example, claims are made about genetic or brain mechanisms in spatial learning and intelligence when mice perform well or show deficits in a Morris water maze. In this task however, the mouse is required to do something it did not evolve to do—swim. Moreover, while swimming to avoid drowning, this non-aquatic species is required to navigate a circular pool to find a submerged platform—again, an improbable scenario. In fact, performance in a Morris water maze can be affected by the rodent’s ability to handle stress, degree of thigmotaxis (the tendency to stay close to a solid surface), and the ability to inhibit a fixed-action pattern (Day and Schallert, 1996). Thus, when an enriched environment aids recovery from a stroke, measured by improved performance in a Morris water maze, it is essential to determine which of these behavioral systems is being affected and not assume that it is spatial learning and cognitive performance, which is the most salient aspect of the test to human investigators (Ronnback et al., 2005).

Conversely, it is also a mistake to assume that human psychosocial traits that affect disease are uniquely human and that humans do not have psychological processes in common with animals. This is an error commonly made when human psychological states are measured with verbal accounts of subjective experience—for example, “I do not feel I have people I can turn to for social support” or “I feel overwhelmed.” Such verbal reports are certainly unique to humans, but nonetheless they are likely based on psychological processes and behavioral traits that have commonality with animal systems, especially when their underlying neuroendocrine mechanisms are similar. The parallel is readily accepted in nonemotional domains. The study of human hunger utilizes self-reports: “I feel hungry” or “I feel sated.” Yet, few question that animals are an excellent model for teasing apart the diverse aspects of hunger and satiety as a motivational state. Indeed, rodent models have been a powerful tool for teasing apart multiple facets of hunger, ranging from taste, chewing, insulin, leptin, and hypothalamic activity to gastrointestinal activity; there are far more independent factors than have been intuitively obvious (White, 1986; Morley, 1990; Hall and Swithers-Mulvey, 1992; Williams et al., 2001; Changizi et al., 2002; Oka et al., 2003). Thus, social animals can be powerful models of psychosocial effects on disease and gene expression, enabling the identification of transduction pathways from the social world to disease as well as the multiple functions of such pathways. Even such seemingly unique hu-



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement