Abstract
Recently Bar (2003) proposed a mechanism for triggering top-down facilitation during visual object recognition where a low-spatial frequency version of an input image is rapidly projected from occipital visual areas to the orbitofrontal cortex (OFC). Using this blurred representation, the OFC activates predictions about the most likely interpretations of the input image in the temporal cortex. This top-down projection facilitates recognition by significantly reducing the number of candidate objects that need to be considered.
For the OFC to facilitate recognition, differential activity would have to develop earlier there than the corresponding activity in the temporal cortex. Using magnetoencephalography (MEG), we have revealed that both recognition-related and spatial-frequency related activity in the OFC do develop 50 ms earlier than in the fusiform gyrus, within the temporal cortex. However, is this early OFC activity a result of early projections from the occipital cortex? Does it result in a direct projection from the OFC to the temporal cortex? And are these interactions governed by low-spatial frequencies? We have analyzed the MEG data using phase-locking statistics to test trial-by-trial covariance to determine whether the relevant regions communicate during object recognition.
The results support early communication between the occipital and orbitofrontal cortices. Occipital activity preceded OFC activity, thus indicating this is a feed-forward interaction. Furthermore, the OFC subsequently phase-locked with the fusiform gyrus, with OFC activity peaking before fusiform activity, signifying a feed-back projection. Additionally, significantly stronger phase-locking was found between the OFC and fusiform for low-pass than high-pass filtered images, indicating that this back projection is associated with low-spatial frequencies. These results confirm our hypotheses and lend critical support to our top-down model of object recognition. Supported by R01 NS44319.