Attending to stimulus features: Contrasting results from reaction time and masked accuracy

Purpose. Reaction time (RT) data from conjunctive visual-search tasks (e.g., look for a red X among red O's and black X's) suggest that people can limit their search to a subset of items that possess one of the relevant features (e.g., they can look for an X among the red items, ignoring all the black items; Egeth, Virzi, & Garbart, 1984). In contrast, accuracy (ACC) data from a visual-search task in which people reported the identity of a single digit among letter distractors suggest that people cannot limit search on the basis of stimulus features (Farell & Pelli, 1993). We sought to identify the source and significance of this apparent conflict. Method. Four experiments are reported in which subjects searched for and identified a digit among letter distractors. The distractors within the display were randomly blue or green/large or small, and the probability that the target would be one of those colors/sizes (e.g., blue/large) was varied across blocks of trials from 0 to 1. If subjects can limit their search, then they should be able to use the probability information to improve their performance. In two of the experiments, displays were not masked and the main dependent measure was RT; in the others, the displays were masked and the main measure was ACC. Results. Subjects were able to use the probability information when displays were not masked and RT was the primary dependent measure. They did not seem able to use the information, however, when the displays were masked and ACC was the primary dependent mesure. No other factor that was tested influenced whether the probability information could be used. Conclusion. The results suggest that attending to a stimulus feature, does not facilitate early perceptual processing (as reflected in masked ACC), but rather, prioritizes items within a later limited-capacity process (as reflected in RT). Three additional experiments are reported in which this conclusion was supported.