Abstract
Are distinct cognitive and neural mechanisms involved in the processes that make up object recognition, such as image segmentation, categorization, and the discrimination of different members within a basic-level category? To answer this question, we concurrently measured performance and fMRI signal from 12 subjects while they carried out three different tasks on briefly presented masked photographs. Subjects were instructed to either detect an object (vs. texture), perform basic level recognition (e.g., dog vs. bird), or identify subordinate exemplars (e.g., pigeon vs. parrot). In Experiment 1 we orthogonally varied presentation duration (17, 33, 50, 67 or 167ms) and task on the same masked stimuli. Our behavioral results showed that the functions relating performance to exposure duration were nearly identical for the detection and basic level tasks, suggesting a close link between the processes involved in image segmentation and basic level categorization. Performance in the subordinate task was significantly lower at all intermediate durations. The fMR data revealed substantial overlap in the object-selective regions (LOC and FUS) activated across tasks. The signal intensity from these regions, for a given image duration, was affected both by the task performed by the subjects and accuracy at that task. In Experiment 2 we directly searched for regions that were correlated with successful recognition in a 2AFC design while holding image exposure constant at 33ms (masked). Voxels where MR signal was significantly correlated with correct object detection (in LOC and FUS) overlapped substantially with those correlated with correct subordinate recognition; non-overlapping voxels had a similar response profile, indicating that they are part of the same functional area. Our results suggest that object segmentation is not separable from recognition and different recognition processes are mediated by common regions in the object recognition pathway.
Supported by: HFSP LT0680 and NEI 13455-05A1