Computational levels and conditional inference: Reply to Schroyens and Schaeken (2003).

This article is a reply to W. Schroyens and W. Schaeken's (2003; see record 2002-08431-014) critique of M. Oaksford, N. Chater, and J. Larkin's (2000; see record 2000-08540-005) conditional probability model (CP) of conditional inference. It is argued that their meta-analysis does not falsify CP because the evidence may bear on more than one computational level of explanation. Moreover, it is argued that CP provides a rational account of more of the data than W. Schroyens and W. Schaeken's mental models theory. Other points are also addressed. It is suggested that W. Schroyens and W. Schaeken's model and CP converge on the importance of probabilistic prior knowledge in conditional inference. This is consistent with the normative literature, which (like CP) treats conditionals in terms of subjective conditional probabilities. (PsycINFO Database Record (c) 2010 APA, all rights reserved)