Associative learning and animal cognition

Abstract

Associative learning plays a variety of roles in the study of animal cognition from a core theoretical component to a null hypothesis against which the contribution of cognitive processes is assessed. Two developments in contemporary associative learning have enhanced its relevance to animal cognition. The first concerns the role of associatively activated representations, whereas the second is the development of hybrid theories in which learning is determined by prediction errors, both directly and indirectly through associability processes. However, it remains unclear whether these developments allow associative theory to capture the psychological rationality of cognition. I argue that embodying associative processes within specific processing architectures provides mechanisms that can mediate psychological rationality and illustrate such embodiment by discussing the relationship between practical reasoning and the associative-cybernetic model of goal-directed action.

Figure 1.

An illustration of the design of a mediated learning experiment by Dwyer et al. [14]. See text for details. P, peppermint flavour; A, almond flavour; S, sugar solution.

Figure 2.

A cartoon of the common-cause experiment of Blaisdell et al. [25] and the hypothetical inferences mediating performance by an intervening and observing rat.

Figure 3.

A cartoon of the associative-cybernetic model of goal-directed lever pressing by a rat for food pellets that identifies the associations that instantiate the cognitive components of practical reasoning. Black lines, learned association; grey lines, fixed connection.