Neural Correlates of Perceptual Rivalry in the Human Brain

See allHide authors and affiliations

Science  19 Jun 1998:
Vol. 280, Issue 5371, pp. 1930-1934
DOI: 10.1126/science.280.5371.1930


When dissimilar images are presented to the two eyes, perception alternates spontaneously between each monocular view, a phenomenon called binocular rivalry. Functional brain imaging in humans was used to study the neural basis of these subjective perceptual changes. Cortical regions whose activity reflected perceptual transitions included extrastriate areas of the ventral visual pathway, and parietal and frontal regions that have been implicated in spatial attention; whereas the extrastriate areas were also engaged by nonrivalrous perceptual changes, activity in the frontoparietal cortex was specifically associated with perceptual alternation only during rivalry. These results suggest that frontoparietal areas play a central role in conscious perception, biasing the content of visual awareness toward abstract internal representations of visual scenes, rather than simply toward space.

Binocular rivalry provides a useful experimental paradigm with which to study the neural correlates of conscious perception (1–3). When dissimilar images are presented to the two eyes, they compete for perceptual dominance so that each image is visible in turn for a few seconds while the other is suppressed. Because perceptual transitions between each monocular view occur spontaneously without any change in the physical stimulus, neural responses associated with perceptual processes can be distinguished from those due to stimulus characteristics. Recent neurophysiological studies in awake monkeys have established that, whereas the firing of most neurons in primary visual cortex (V1) correlates with the stimulus and not the percept during rivalry, activity of neurons at higher levels in the visual pathway, such as in the inferotemporal cortex, reflects the perceptual state (3). These findings suggest that rivalry results from a competition between alternative stimulus interpretations at a level beyond the stages of monocular processing early in visual cortex (4). Psychophysical observations also suggest that perceptual alternation during rivalry results from the same neural operations underlying other multistable perceptual phenomena, such as depth reversals and ambiguous figures, that show similar temporal dynamics to binocular rivalry (5). Although less pronounced, similar perceptual fluctuations can also be experienced in normal vision and may therefore reflect a basic perceptual strategy to resolve visual ambiguity (6). Yet despite significant interest in the neural correlates of binocular rivalry (1–3), the mechanisms underlying these perceptual alternations remain unknown.

Here we investigate these mechanisms by characterizing neural activity associated with perceptual transitions per se, rather than activity associated with perceptual state during rivalry. Our results provide evidence for an involvement not only of occipitotemporal visual areas in binocular rivalry, but also indicate a specific and previously unknown role for frontoparietal areas in mediating the perceptual transitions experienced during rivalry. These results were obtained by measuring brain activity with functional magnetic resonance imaging (fMRI) in humans who reported their percepts under two different viewing conditions (7). In the first condition, subjects viewed dichoptic stimuli consisting of a red-colored drifting grating shown to one eye and a green-colored face shown to the other eye. These images were chosen because they are highly dissimilar and readily produce full-field rivalry when viewed through stereoscopic glasses. By manipulating the contrast setting in each image, we were able to bias perceptual dominance in favor of the grating, with long periods during which the grating was seen alone interrupted by shorter incursions of the face in conscious perception (Fig. 1) (8). Subjects used key presses to signal perceptual alternations from the grating to the face or vice versa. To control for motor effects, we compared the activity evoked during rivalry to that elicited in a second, nonrivalrous viewing condition that required the same type of motor responses. In this second condition, subjects were exposed to a “replay” of their perception during rivalry. This was achieved by presenting, in a chronology specified by the key reports during rivalry, either the face alone or the grating alone to one eye, and a gray patch of comparable luminance to the other eye. At transition times, physical blends of the face and grating were shown. This stimulation was designed to produce a perception that closely mimics rivalry in both quality and timing, thus resulting in a matched sequence of motor reports in the two conditions (9). Because prolonged periods of stereoscopic fusion can cause ocular fatigue, a third, passive condition was also introduced during scanning to allow visual rest.

Figure 1

Temporal dynamics of binocular rivalry during fMRI. The frequency histograms show the distribution of dominance times for face (left) and grating (right) reported by a representative subject while undergoing functional imaging. Mean dominance times averaged across subjects are given in (8).

Functional MRI scans from six participants were analyzed as a group to identify brain areas where activity was consistently correlated with the perceptual changes reported during either viewing condition (10, 11). To distinguish transient activity associated specifically with perceptual alternation from other, nonselective effects of viewing condition, we modeled the predicted hemodynamic response to each transition event and tested for the presence of such responses in the data while treating the mean condition-specific effects as confounds. Such an event-related modulation of the fMRI signals reflects neural activity that is locked to the time of occurrence of perceptual transitions between face and grating. During rivalry, transient responses associated with shifts of perception were found bilaterally in extrastriate areas of the fusiform gyrus, in right inferior and superior parietal lobules, and in bilateral inferior frontal, middle frontal, and insular cortex (Fig. 2 and Table 1). Event-related activity was also observed in regions of the anterior cingulate cortex, supplementary motor area (SMA), and left primary motor and somatosensory cortex, consistent with the preparation and execution of appropriate motor reports. The estimated hemodynamic response to single transition events is shown for a representative region of activation inFig. 2 as a function of postevent time (12).

Figure 2

Event-related activity during rivalry and replay conditions. (A) Four views of the medial and lateral surfaces of a rendering of the T1-weighted anatomical template image in Talairach space, on which are superimposed areas where evoked activity was specifically related to perceptual transitions in either the rivalry condition (red) or the replay condition (green). A statistical threshold of Z = 3.09 (corresponding to P< 0.001, uncorrected) was used for display purposes; peaks of activation reaching statistical significance after correction for multiple comparisons (P < 0.05) are listed in Table 1. The areas modulated by perception during both rivalrous and replay viewing, and the bilateral symmetry of the evoked activity are apparent. (B) Illustrative postevent histograms of the modulation of activity produced by transition events in rivalry (red) and replay (green) conditions from three different subjects. The evoked activity (percent change in BOLD contrast) is shown as a function of postevent time (in seconds) for each subject, with the fitted models of hemodynamic response function superimposed in solid lines. The modulation of activity shown here is taken from a voxel in right anterior fusiform gyrus (x = 33 mm, y= −45 mm, z = −21 mm; Z = 8.10, P < 0.001 corrected).

Table 1

Coordinates and Z scores for event-related activation. Shown in the table are loci where event-related activity is greater during replay compared with rivalry (replay > rivalry); modulation of activity above baseline is measured when viewing dichoptic stimuli (rivalry) or replayed scenes (replay); and event-related activity is greater during rivalry compared with replay (rivalry > replay). Only the most significant peaks within each area of activation are reported in the table (P< 0.05, corrected).

View this table:

Although neural correlates of rivalrous transitions were expressed at multiple levels of the visual occipitotemporal pathway, such correlates were not observed in primary visual cortex. Perceptual transitions experienced during dichoptic stimulation were highly correlated with activity in extrastriate areas concerned with the representation of higher order properties of the visual scene (13). In particular, we detected transient responses that reflected these perceptual changes in regions of the fusiform gyri that included areas previously implicated in the perception of faces (14). By contrast to higher visual areas, early visual cortical areas showed no significant modulation of activity during rivalrous transitions. The differential involvement of early and higher visual cortical areas in rivalry was further confirmed by an analysis of activity correlated with the perceptual state rather than with the perceptual transitions reported by the subjects (15). These results in humans are consistent with recent findings in monkeys, suggesting that rivalry reflects central processes that take effect subsequent to the analysis of both monocular stimuli (3,4).

By comparing rivalry to a nonrivalrous viewing condition, the present study also allowed us to provide an answer to the central issue in multistable perception—whether a specific machinery mediates the ongoing selection among sets of neuronal events competing for visual awareness. Because the rivalry and replay conditions yield similar perception and behavior, we expected them to engage common neural pathways associated with the internal representation of visual scenes and the generation of appropriate motor responses. This was confirmed by the fMRI data: Event-related activity during replay was similar to that evoked during rivalry in visual areas of the fusiform gyri and in areas associated with movement (Fig. 2 and Table 1) (16). However, the rivalry and replay conditions differ fundamentally in the way that they achieve alternating perception. Whereas perceptual shifts during rivalry derive from an endogenous neural instability in the absence of changes in the stimulus, during replay they rely on exogenous manipulation of the visual input. Hence, we reasoned that any differential event-related activity between the two conditions would reflect these differences. Such a contrast would expose the mechanisms underlying the ongoing selection between conflicting perceptual interpretations during rivalry, a conflict that is not evoked by the replay condition. In addition, we predicted that the early visual cortex may show less transient activity during rivalry than during replay, because dichoptic stimulation causes little modulation of neuronal activity and possible inhibition at this stage of processing (3, 17), whereas repeated stimulus onset and offset as generated during replay typically evoke strong cortical responses.

In contrast to the bilateral pattern of event-related activity in common across conditions, activity specific to the rivalry condition was strongly lateralized to the right hemisphere. Selective activation during rivalrous perceptual transitions was seen in a region of right extrastriate visual cortex, Brodmann area (BA) 19, and in the right inferior parietal, superior parietal, and inferior frontal cortex (Fig. 3 and Table 1). This pattern of activation was both highly significant and consistent across subjects. We also characterized areas where transient activity associated with perceptual alternations was greater during replay compared with rivalry. Areas that showed such differential activation were located in early visual cortex (medial portion of BA 18), in accord with our prior hypothesis (Fig. 3 and Table 1). The comparison between the rivalry and replay conditions demonstrates a double dissociation; right frontoparietal regions show greater transition-related activity during rivalry, whereas early visual cortex shows greater transient responses during nonrivalrous viewing. These differences cannot be attributed to the generation of motor reports because the two conditions entailed the same sequence of motor responses and produced similar activity in cortical areas associated with movement (9). It is also unlikely that this differential pattern of activity results from nonspecific differences in attentional demands between the two conditions, such as arousal or difficulty. Frontoparietal activity has not been observed during the performance of other visual tasks in which attentional demands were varied systematically (18). Moreover, changes in attentional demands typically result in different levels of activation in ventral areas involved in representation of visual scenes (19); such differences were not observed in the present study when the two viewing conditions were compared. Instead, the present results are more consistent with the notion that a distributed frontoparietal system specifically mediates the perceptual switches experienced during rivalry.

Figure 3

Differential activation during rivalrous and nonrivalrous viewing. (A) Areas where transient activity related to perceptual shifts is greater under conditions of binocular rivalry compared with the replay condition, overlaid onto the average Talairach normalized anatomical MR image of the six subjects. Significant differential activation during rivalry (P < 0.05, corrected) is confined to the right hemisphere and involves frontoparietal structures previously implicated in the shifting of spatial attention. Distance from the anterior commissure is indicated for each coronal section. L, left; R, right. (B) A transverse section through the average normalized anatomical MR image, taken 9 mm below the bicommissural plane, on which are superimposed two foci of activation that represent early visual areas where activity related specifically in time to perceptual transitions is greater under conditions of replay compared with rivalry.

Right frontoparietal areas have been traditionally implicated in visual tasks requiring spatial shifts of attention and working memory (20–24). Visuospatial neglect syndromes occur most frequently and are more severe after lesions in the right inferior parietal and inferior frontal cortex (21). Moreover, functional imaging experiments have shown that the region of superior parietal cortex identified in the present study is also engaged by successive shifts of spatial attention (22). Finally, differential activation of right extrastriate cortex has been reported in tasks directing attention to global aspects rather than local details of figures (23). But our results show that these cortical areas are also involved in a phenomenon that exhibits a number of differences compared to visuospatial attention. In contrast to shifts of attention, there is no spatial component to the perceptual transitions elicited during rivalry; moreover, whereas attentional shifts are subject to top-down influences, rivalrous transitions recur in the absence of voluntary control; finally, spatial attention also engages visual and motor areas that were not activated during rivalry (24). Why then should both phenomena involve overlapping regions of frontoparietal cortex? One possibility is that these areas subtend separate neural mechanisms for spatial attention and perceptual rivalry. However, it is striking that both phenomena entail the suppression of visual information from conscious perception. Monocular stimuli become periodically invisible during rivalry; similarly, sensory events associated with unattended stimuli have a diminished impact on awareness during covert attention. These effects occur in both cases in spite of a rather constant retinal input. Both phenomena may therefore call upon a common neural machinery in frontoparietal cortex, involved in the selection of neuronal events leading to visual awareness.

Thus, our results suggest that the role of frontoparietal areas in conscious perception extends well beyond that of spatial processing. Consistent with this notion, lesions of parietal and inferior frontal cortex cause disorders of nonspatial forms of perceptual selection, in addition to spatial disorders (25). Further investigation of frontoparietal function in both human and nonhuman primates may lead to a better understanding of the neural processes underlying the formation of perceptual states and the awareness of sensory stimuli.

  • * To whom correspondence should be addressed. E-mail: elumer{at}


View Abstract

Navigate This Article