Activity in the lateral occipital cortex between 200 and 300 ms distinguishes between physically identical seen and unseen stimuli

There is converging evidence that electrophysiological responses over posterior cortical regions in the 200-300 ms range distinguish between physically identical stimuli that reach consciousness or remain unseen. Here, we attempt at determining the sources of this awareness-related activity using MEG. Fourteen subjects were presented with faint colored gratings at threshold for contrast and reported on each trial whether the grating was seen or unseen. Subjects were primed with a color cue that could be congruent or incongruent with the color of the grating, to probe to what extent two co-localized features (color and orientation) would be bound in consciousness. The contrast between neural responses to seen and unseen physically identical gratings revealed a sustained posterior difference between 190 and 350 ms, thereby replicating prior studies. We further show that the main sources of the awareness-related activity were localized bilaterally on the lateral convexity of the occipito-temporal region, in the lateral occipital (LO) complex, as well as in the right posterior infero-temporal region. No activity differentiating seen and unseen trials could be observed in frontal or parietal regions in this latency range, even at lower threshold. Color congruency did not improve gratings' detection, and the awareness-related activity was independent from color congruency. However, at the neural level, color congruency was processed differently in grating-present and grating-absent trials. The pattern of results suggests the existence of a neural process of color congruency engaging left parietal regions that is affected by the mere presence of another feature, whether this feature reaches consciousness or not. Altogether, our results reveal an occipital source of visual awareness insensitive to color congruency, and a simultaneous parietal source not engaged in visual awareness, but sensitive to the manipulation of co-localized features