Illusory conjunctions: Does inattention really matter?

Treisman and Schmidt (1982) found that under conditions of load, there were more conjunction errors, namely errors resulting from erroneously conjoining features from different stimuli, than feature errors, namely errors resulting from the false perception of a feature that was not presented at all. That did not occur when attention was not loaded. That interaction was argued to support feature integration theory. The attention manipulation in that study was, however, confounded with spatial uncertainty, exposure duration, and report immediacy. The present study was meant mainly to remove those confoundings. Subjects had to search for a probe in an array of three stimuli. In addition, they either had or did not have to match two outflanking digits. Exposure duration, probe type, stimulus-probe delay, and stimulus-digit delay were manipulated. Response order in the dual-task condition, and presence of digits in the single-task one were varied between subjects. Conjunction errors were found to be more frequent than feature errors, but that obtained in both attention conditions alike. That is shown to be a straightforward corrolary of the nature of these types of error that does not have implications about feature integration or perception of feature origin. Other results suggested that performance was affected by post-perceptual factors more than by perceptual ones, which casts further doubts on accounts that resort to feature integration theory.