This chapter questions the received view that in medical research extrapolation from animal models mainly consists in establishing mechanisms of human pathological states in organisms, thanks to a step by step comparison of causal pathways. Mechanistic extrapolation takes the form: (1) cause C brings out effect E in animal through causal pathway M, (2) M is similar in animals and humans, (3) therefore C will likely bring out E in humans. As the example of psychiatric research shows, such mechanistic extrapolation may be replaced by statistical extrapolation, an inference of the form: (1) An animal model A has been successful in predicting the effects E of drugs D1…Dn of a certain class; (2) A will be successful again in predicting the effects of a new drug Dn+1 of the same class. Statistical extrapolation relies on the predictive validity of a given animal model, without any knowledge of the mechanisms involved, on the sole ground of past successes of the model in predicting the effects of a class of drugs on their human target.
Read more here.