Abstract
Attention plays an important role in object recognition by helping observers select features (e.g., shape, motion, or color) that best identify objects in a cluttered and dynamic environment. At the neural level, attention to specific features leads to increased hemodynamic responses in brain regions that selectively process those features. However, the neural fate of unattended features remains largely unknown. One possibility is that unattended features are implicitly processed. This implicit processing may not lead to increased regional responses, but may instead modulate responses in other feature-selective regions thereby integrating multiple object features. To test this, we adapted a Garner paradigm for functional imaging. Observers (N = 12) were shown novel objects that had a unique combination of shape, non-rigid motion, and color. They were instructed to attend to each feature in separate blocks, and to respond only when the attended feature was repeated. For each attended feature, there was a baseline block in which both unattended features were repeated on every trial, and a filter block in which both unattended features changed from trial to trial. We expected that fMRI-adaptation would decrease regional responses to unattended features on baseline blocks but not on filter blocks. Consistent with previous work, a whole-brain analysis showed that attention to shape, motion or color led to increased responses in shape-, motion-, and color-selective regions. Surprisingly, there was no main effect of filter versus baseline blocks despite the large perceptual differences between these two conditions. A psychophysiological interaction analysis, however, revealed that motion- and color-selective regions identified by the whole-brain analysis modulated responses in shape-selective regions more so on filter blocks than on baseline blocks. Taken together, our findings suggest that implicitly processing unattended features functionally connects feature-selective regions. This network underlies successful recognition in a dynamic environment.