An autoassociative model of causal reasoning and causal learning: Reply to Van Overwalle's (1998) critique of Read and Marcus-Newhall (1993).

Abstract

The authors respond to F. Van Overwalle's (1998) critique of the explanatory coherence (ECHO) model of causal reasoning (P. Thagard, 1992) and its application to social reasoning (S. J. Read & A. Marcus-Newhall, 1993). Contrary to Van Overwalle's doubts about its possibility, a feedback model similar to ECHO that learns covariation information appropriately is presented. This model simulates Read and Marcus-Newhall's results and can also simulate findings from the causal learning literature. In contrast, because of limitations in feedforward networks, Van Overwalle can simulate many of Read and Marcus-Newhall's results only by making incorrect assumptions about their procedures. Further, the model presented handles several issues that Van Overwalle's cannot, such as asymmetries in reasoning between cause and effect, evaluation of explanatory coherence, and causal chains. Finally, limitations of feedforward models of causal reasoning are discussed.