ABSTRACT
This paper examines how we use choice RT measurement to make inferences about mechanisms for processing stimulus information. Distinguishing strategy effects from task effects is a major problem. One approach, which is widely used, supposes that if the probability of a correct response remains constant across experimental conditions, strategy also remains constant and so differences among RTs are purely task effects. It strongly appears that this approach is based on tacit acceptance of a special case of the fast guess model. Specifically, it can be shown that this approach is consistent with the fast guess model but inconsistent with the deadline model, for example. The adjustable timing model, which is suggested by the Schouten and Bekker (1967) study, commends the opposite approach: if the distribution of RTs remains constant across experimental conditions, then strategy remains constant and so differences in the probability of a correct response arise purely from task effects. Particular methods for distinguishing effects are based on particular models and there is no universally accepted model at this time. To decide which method is best suited to a particular problem, it is necessary to conduct factorial experiments in which both “task” and “strategy” variables are manipulated. A particular method for distinguishing types of effects should not be chosen until after alternative models have been fitted and tested. An approach to the statistical problem of testing the adjustable timing model is briefly sketched.
