Postcategorical filtering in a bar-probe task

Abstract

To investigate the causes of errors in partial-report bar-probe tasks, a set of (color) stimuli that allowed the use of a new method of analysis was devised. With these stimuli and this method of analysis, two important results were obtained. First, the number of location errors (i.e., errors consisting of naming the color of an item in a nonindicated position) was far below chance level with short probe delays and moved in the direction of chance level with increasing probe delays. Second, the number of genuine visual confusions first increased and then somewhat decreased with increasing probe delays. The results make it possible to choose among three current models of information processing for this type of task. The "orthodox" account and the "dual-buffer" account are not consistent with the data obtained. Postcategorical filtering, that is, the selection of identified information by using precategorical information, is compatible with the results reported. © 1984 Psychonomic Society, Inc.