A process model of human transitive inference

If you are told that job A pays more than job B and job B pays more than job C, then you can conclude that job A pays more than job C. This conclusion follows deductively because pays-more-than()—or, more generally, more-than()—is a transitive relation. How does the human mind reason about transitive relations such as more-than(), taller-than(), etc.? Transitive relations are logically well-behaved: Given that some relation, R(), is transitive, and given the facts R(A, B) and R(B, C), we can conclude R(A, C) with certainty. As such, there are many ways in principle to reason about them. For example, it is easy to imagine a cognitive "module " devoted to transitive reasoning. Like a program written in PROLOG, the module might take statements about transitive relations as input—for example, statements of the form pays-more-than(A, B) and pays-more-than(B, C)—and generate various inferences—such as pays-most(A), pays-least(C), and pays-more-than(A, C)—as output. However, the behavioral evidence suggests that this is not how the human mind processes transitive relations. A logical module would not provide any inherent basis for predicting that some transitive inferences will be more difficult to make than others. But it has been known for a long time that people solve some transitive-inference problems more quickly an