Abstract
Both monkey neurophysiology and human fMRI studies have shown that in occipital-temporal cortex, neural response to multiple visual objects shown simultaneously can be reliably predicted by the averaged neural response of the component objects shown alone. Besides occipital-temporal cortex, visual object representation has also been shown in the primate parietal cortex. Does a similar representation for multiple visual objects exist in the primate parietal cortex? Using multi-voxel pattern analysis, we examined object representation in human inferior and superior intra-parietal sulcus (IPS), two parietal regions implicated in visual object individuation and identification, respectively. As a comparison, we also examined response from object shape processing region in lateral occipital (LO) cortex. During the experiment, participants saw either a pair of objects shown above and below the fixation, or a single object shown at either location. In all three regions, distinctive fMRI response patterns were found for the different single objects and object pairs shown, confirming visual object representation in these brain regions. Replicating previous findings, in LO, object pair patterns were indistinguishable from the averaged patterns of the component objects shown alone. In both inferior and superior IPS, however, object pair patterns differed significantly from those of the component objects. This pattern difference could be attributed to task context difference between attending to a single object and attending to a pair of objects. When such difference was removed by only examining paired objects, we found that the averaged pattern of two object pairs (e.g., A+B and C+D) was indistinguishable from those of two other object pairs containing the same four component objects (e.g., A+D and C+B). These results suggest that parietal response to multiple objects can be predicted from responses to the components objects, but only under the same task context. Human parietal cortex thus contains both object and task representations.
Meeting abstract presented at VSS 2014