ABSTRACT
Reasoners revise their beliefs in the premises when an inference they have made is contradicted. We describe the results of an experiment that shows that the belief they revise depends on the inference they have made. They revise their belief in a conditional (if A then B) when they make a modus tollens inference (from not-B to not-A) that is subsequently contradicted (A). But when they make a modus ponens inference (from A to B) that is contradicted (not-B) they revise their belief in the categorical assertion (A). The experiment shows that this inference contradiction effect occurs not only for factual conditionals but also for counterfactual conditionals. However, reasoners revise their beliefs in factual conditionals more than counterfactuals.
