Following Shanks and Darby (1998), participants in our experiments had to learn that single cues that signalled the same outcome (e.g., A-1/B-1) predicted the opposite outcome when presented in compound (e.g., AB-2). Some cues were only presented in compound during training (e.g., CD-2) to see whether, at test, participants tended to respond according to rule-based (i.e., C-1/D-1) or according to feature-based generalisation (i.e., C-2/D-2). The generalisation test was based on two different tasks: A predictive judgment, and a cued-response priming task. In the judgment task, participants’ responses were consistent with rule-based generalisation. However, participants’ reaction times in the cued-response priming task were consistent with feature-based generalisation. This dissociation contradicts Jan De Houwer and Stefaan Vandorpe’s (2010) results based on the implicit association task, and indicates that, when a priming task with a short SOA (200 ms) is used, the expression of very simple processes is favoured over more complex reasoning processes. Conversely, when participants have unlimited time to think, complex reasoning processes may override simple processes. Consequently, our study supports the idea that priming techniques with demanding time requirements favour the expression of activation-based processes such as associative processes, and prevent to a great extent the operation of inferential reasoning.
