Pragmatic versus syntactic approaches to training deductive reasoning

Two views have dominated theories of deductive reasoning. One is the view that people reason using syntactic, domain-independent rules of logic, and the other is the view that people use domain-specific knowledge. In contrast with both of these views, we present evidence that people often reason using a type of knowledge structure termed pragmatic reasoning schemas. In two experiments, syntactically equivalent forms of conditional rules produced different patterns of performance in Wason's selection task, depending on the type of pragmatic schema evoked. The differences could not be explained by either dominant view. We further tested the syntactic view by manipulating the type of logic training subjects received. If people typically do not use abstract rules analogous to those of standard logic, then training on abstract principles of standard logic alone would have little effect on selection performance, because the subjects would not know how to map such rules onto concrete instances. Training results obtained in both a laboratory and a classroom setting confirmed our hypothesis: Training was effective only when abstract principles were coupled with examples of selection problems, which served to elucidate the mapping between abstract principles and concrete instances. In contrast, a third experiment demonstrated that brief abstract training on a pragmatic reasoning schema had a substantial impact on subjects' reasoning about problems that were interpretable in terms of the schema. The dominance of pragmatic schemas over purely syntactic rules was discussed with respect to the relative utility of both types of rules for solving real-world problems.