Relative blindsight in normal observers and the neural correlate of visual consciousness

Abstract

By using a paradigm based on metacontrast masking, we created experimental conditions in which the subjective report of consciousness differs but the objectively measured ability to discriminate visual targets does not. This approach allowed us to study the neural correlate of consciousness while having performance levels carefully matched in healthy human subjects. A comparison of the neural activity associated with these conditions as measured by functional MRI showed that conscious perception is associated with spatially specific activity in the mid-dorsolateral prefrontal cortex (area 46). Further analysis confirms that this activation is not only free from any performance confound, but is also not driven by differences in the timing of the physical stimuli. Our results suggest that the prefrontal cortex is important for the essentially subjective aspects of conscious perception.

Fig. 1.

Visual discrimination task with metacontrast masking. The stimuli were presented on a black background. The mask overlaps with part of the contour of the target without leaving gaps or overlapping with the target spatially. After the presentation of the target and the mask, the participants were first asked to decide whether a diamond or a square was presented. Then they had to indicate whether they actually saw the target or whether they simply guessed the answer. Shown in the brackets are the durations of each stimulus.

Fig. 2.

Behavioral results. (Upper) Results from the behavioral study, which used the procedure described in Fig. 1, except that it also included trials during which the mask was presented before the target (paracontrast masking). The data were averaged over eight participants, and the error bars represent one SEM. Note that at the SOA values for which the performance levels (% correct) were the same (e.g., 33 ms and 100 ms), the awareness levels (% seen) differed significantly. (Lower) The behavioral results from the fMRI study. The data were averaged over 13 participants, and the error bar represents one SEM. We managed to create two conditions (short SOA and long SOA) that differed significantly in terms of subjective awareness but not in terms of discrimination accuracy.

Fig. 3.

Activity in the mid-DLPFC reflects visual consciousness (long SOA > short SOA). The activity in this area is higher in the long SOA condition than in the short SOA condition, despite the fact that the two conditions did not yield different discrimination accuracy. There were, however, more trials during which the stimuli were classified by the participants as consciously seen in the long SOA condition than in the short SOA condition. This area was the only one found to be activated in this comparison; parietal activity did not differ significantly even at liberal thresholds. The plot shows the time courses of the hemodynamic activity in the two conditions, as measured by the percentage of signal change from baseline. The data were averaged over 13 participants, and the error bar represents one SEM. Marked by an asterisk is the time point at which the signal differed significantly as reflected by a paired t test.

Fig. 4.

Activity in the mid-DLPFC across subjects under the same SOA. This correlation was performed to test whether the activation shown in Fig. 3 merely reflects a stimulus confound (long vs. short SOA). Each point represents a single subject, and the data were taken from the short SOA condition, for which the SOA was the same for each subject at 33 ms. As shown in the plot, the higher the activity in the mid-DLPFC, the more likely the identity of the stimulus was declared as consciously seen. This correlation was weak (r = 0.512, P = 0.037, one-tailed), but it shows that the fact that activity in the mid-DLPFC reflects the subjective criterion for conscious perception cannot be explained simply in terms of SOA.

Fig. 5.

Activations reflecting performance in general (correct trials > incorrect trials). Activations were found in these areas when correct trials were compared against incorrect trials, combining all trials of both SOA conditions. The pattern of these activations resembles that in a so-called frontal–parietal network, typically reported in previous studies of visual consciousness. VLPFC, ventrolateral prefrontal cortex; IPS, intraparietal sulcus; MTG, middle temporal gyrus.