Abstract
What are the units of visual attention? O'Craven, Downing, and Kanwisher (1999) have demonstrated that attending to one feature of an object (e.g., motion) enhances the cortical response to this feature as well as other features (e.g., form) of the same object, even when the latter were wholly task-irrelevant. On this basis, O'Craven et al. concluded that attention automatically spreads to all features of an object. However, the behavioral task used in these studies did not explicitly discourage observers from attending multiple features of the same object. Thus, it remains possible that observers voluntarily selected task-irrelevant features of an object, even though they had no direct incentive to do so. In the present experiments, we sought to provide a more stringent test of the hypothesis that attention automatically spreads to all features of an object. Observers performed a change detection task in which they were required to discriminate differences in either the orientation or color of a single object. To encourage a narrow focus of attention on the relevant feature, task difficulty was continuously titrated for each observer. Using fMRI and a multivoxel pattern analysis (MVPA), we show that when the observer was attending orientation, the pattern of cortical activity in V1 discriminated the currently viewed orientation, but not the currently viewed color. In contrast, when the observer attended color, the pattern of cortical activity in V1 discriminated the currently viewed color, but not the currently viewed orientation. These results are inconsistent with the notion that attention automatically spreads to all features of an attended object. Instead, they suggest that attention may be voluntarily restricted to a single behaviorally relevant feature.