ABSTRACT
Response time (RT) data are widely used in studies of cognitive abilities, both in typical college-aged populations and (increasingly) in aging populations. Although many of these studies focus on aspects of processing capacity (as it is affected, e.g., by variations in stimulus conditions, cuing conditions, etc.), the typical analyses of RT data do not map well onto the construct of capacity. We suggest (following Townsend & Ashby, 1978) that hypotheses framed at the level of the hazard function of the RT distribution do map well onto the notion of capacity. We then consider, in a set of simulations, the extent to which a set of well–known models for the hazard function—proportional hazards models—can be applied to data that possess the properties of typical RT distributions. We find that the models are sensitive to the presence of true orderings on hazard functions, but are consistently conservative in their estimates of the magnitude of the ordering. We illustrate the use of these models on a data set from a free and cued recall task administered to a set of older adults.
