Abstract
In this chapter, we argue that the currently available evidence on how people interpret conditionals favors the probabilistic view, whereas evidence on how people reason with conditional premises supports the mental-model view. This motivated our search for a way to integrate the best parts of both approaches. The interpretation of conditionals in terms of conditional probabilities is incompatible with a truth-functional interpretation of conditionals. Therefore, we propose to give up the idea that the meaning of conditionals can be represented by the set of mental models that meet their truth conditions, or which are compatible with the conditional. Conditionals are not truth-functional, that is, they don't have truth conditions by which we can determine whether a given conditional is true or false simply from what is the case in a world. We believe that the core meaning of 'if p then q' is best expressed by a production that adds a representation of 'q' to every representation of 'p' in working memory. This production can be used in two ways when it comes to reason from conditionals. When the minor premise matches the conditions of application of the production (i.e., the antecedent of the conditional it represents), the conditional's consequent can be generated as a conclusion directly. Alternatively, reasoners can build mental models of situations compatible with the minor premise and use the production that incorporates the conditional to constrain these models. The two inference routes map onto the two processes in the dual-process model. They reflect not two separate systems of information processing but simply two ways of using the same system.