Redundant-target detection and processing capacity: the problem of positional preferences.

Although the capacity issue is fundamental to the understanding of human performance, there has been considerable difficulty in designing methodologies that distinguish between limited-capacity and unlimited-capacity processing models. The mathematical complexity of this enterprise has been discussed in a series of papers by Townsend and his colleagues (Snodgrass & Townsend, 1980; Townsend, 1971, 1972, 1974). In these papers, Townsend demonstrated that models of various capacities can be made to mimic the predictions of one another. For example, Townsend notes that serial and parallel models are not strictly identifiable on the basis of the relationship between reaction time (RT) or accuracy and the number of elements in visual search tasks. One task for which differential predictions are made by limited-capacity and unlimited-capacity processing models is the redundant-target detection task (e.g., Biederman & Checkosky, 1970; Egeth, Folk, & Mullin, in press; Mullin & Egeth, in press; Snodgrass & Townsend, 1980; van der Heijden, 1975; van der Heijden, La Heij, & Boer, 1983). In this task, the number of simultaneously presented items is varied. On a given trial, only items assigned to a single response are displayed. Mean RTs are then compared to determined if there is a decrease in RT with increasing numbers of targets, an effect known as a redundancy gain (for related work with redundant targets using accuracy measures, see Eriksen, 1966, and Santee & Egeth, 1982). Snodgrass and Townsend (1980) have shown that limited-capacity models (both serial and parallel) are incompatible with a redundancy gain, but a spatially parallel, self-terminating, unlimited-capacity model remains plausible if response latency is determined by the first input to f’mish processing, since one of several inputs will finish processing faster on average than the average processing time for a single input. One difficulty with the use of the redundant-target detection task is that limited-capacity processing can result in an artifactual redundancy gain if subjects have preferences or differential abilities to process certain types of information (Biederman & Checkosky, 1970). Although