Decoding what the brain 'sees' independent of what the person 'sees'

Multi-voxel pattern analysis techniques allow us to infer what the brain encodes from neuroimaging scans. During scanning we briefly present stimuli to participants to examine neural responses to seen and unseen stimuli. We then examine the accuracy of a support vector machine to classify the stimulus category based on the pattern of neural activity. Previously we have shown patterns in particular brain regions involved in visual perception of novel objects carry information that corresponds to behavioral performance. Although the spatial pattern in both retinotopic and lateral occipital cortex (LOC) in humans contains information about the category of an object, only in the LOC is the neural pattern stronger for correct than for incorrect trials. This dissociation between early visual and higher visual areas demonstrates the way in which the brain encodes novel objects, with a particular role for LOC in determining behavioural responses. Here we examined the neural patterns to seen and unseen 'real' objects, faces and scenes. Interestingly the areas of the brain that encode these unseen stimuli include regions of the temporal lobe, much further along the processing stream than LOC. This suggests that unseen familiar stimuli are processed at a higher level than unseen novel stimuli.