A Discrimination-Association Race Model for the Implicit Association Test

A discrimination-association race model for the analysis of the IAT is presented. Discrimination regards the amount of information that a target (attribute) stimulus contains for the target (attribute) categories. Association regards the amount of information that a target (attribute) stimulus contains concerning an attribute (target) category. Four parallel and independent Poisson processes, each of which concerns a specific category, have been considered. Information about a specific characteristic of stimuli accumulates on the counter of each process until a termination criterion is reached and a response is given. Speeds of information accumulation vary with processes and stimuli. Termination criteria vary with blocks, stimulus categories and response categories (correct and incorrect). The model takes into account both latency and accuracy of the responses. The latency is determined by the time at which a process wins. The accuracy depends on the winning process, the category of the displayed stimulus, and the test block. In an empirical application, parameter estimates and goodness-of-fit were computed for each of one-hundred respondents to a Conscientiousness-IAT. Practical implications of the model for understanding the IAT measure are discussed. In particular, it is shown that, for each respondent, the speed-accuracy task can be decomposed into four components: the discriminability of the stimuli (discrimination), the association of stimuli to categories (stimulus-driven target/attribute association), the association of categories to stimuli (category-driven target/attribute association), and the difficulty of different blocks involved in the procedure (termination criteria). Advantages and disadvantages of the model are considered in comparison to other models for the analysis of the IAT.