ABSTRACT
Learning about predictive contingent relationships allows a window into the underlying causal structure of the environment. The alternative might be to work out the true physical mechanism of any caused event. For example, that flowers on fruit trees precede the arrival of fruit can be understood on the basis of a complete understanding of the ontogenesis of fruit, but even this knowledge is based on a chain of predictive contingencies. The epistemological nature of causal knowledge has been questioned for centuries (see also chap. 2, this volume). The idea that causal knowledge might be founded on an understanding of correlation is attributed to Hume (1739/1960). He suggested that detecting a positive contingency between flowers and fruit is enough information to infer a causal connection or, minimally, to allow us to act as if there was an understanding of the causal relation. Unlike working out the complete physical model, learning a contingency simply requires remembering the likelihood of fruit following flowers and comparing this memory with one of the likelihood of fruit in the absence of flowers.
