Mechanisms of Directed Attention in the Human Extrastriate Cortex as Revealed by Functional MRI

See allHide authors and affiliations

Science  02 Oct 1998:
Vol. 282, Issue 5386, pp. 108-111
DOI: 10.1126/science.282.5386.108


A typical scene contains many different objects, but the capacity of the visual system to process multiple stimuli at a given time is limited. Thus, attentional mechanisms are required to select relevant objects from among the many objects competing for visual processing. Evidence from functional magnetic resonance imaging (MRI) in humans showed that when multiple stimuli are present simultaneously in the visual field, their cortical representations within the object recognition pathway interact in a competitive, suppressive fashion. Directing attention to one of the stimuli counteracts the suppressive influence of nearby stimuli. This mechanism may serve to filter out irrelevant information in cluttered visual scenes.

The human visual system is usually confronted with cluttered scenes consisting of many different objects, which cannot all be processed simultaneously. Only a limited amount of what we see reaches consciousness and becomes stored in memory, which indicates that there is limited processing capacity within the visual system and that multiple object representations are in competition for access to this limited-capacity system (1). One way to resolve the competition is through spatially directed attention. If one attends, for example, to a specific location in a cluttered scene, information processing is greatly facilitated at that location, while interfering information from objects at nearby locations is efficiently filtered out. This suggests that processing is biased in favor of the attended location (2).

Results from single-cell recordings in extrastriate cortical areas in the ventral object vision pathway of monkeys are consistent with these ideas (3). Evidence for competition is provided by the finding that the response to an otherwise optimal stimulus presented within a neuron's receptive field is often reduced when a second stimulus is presented simultaneously at a different location within the same receptive field. Hence, multiple stimuli are not processed independently from each other but rather interact competitively in a mutually suppressive fashion. This competition can be biased in favor of one of the stimuli by spatially directed attention. If an animal directs its attention to one of the competing stimuli within the receptive field, the responses are as large as those to the stimulus presented alone. These results suggest that spatially directed attention to a visual stimulus cancels out the suppressive influence of nearby stimuli, thereby enhancing information processing at the attended location. If so, this could be a mechanism to filter out unwanted information in cluttered visual scenes.

We used functional magnetic resonance imaging (fMRI) in humans to test for the presence of suppressive interactions among stimuli presented simultaneously within the visual field in the absence of directed attention (experiment 1) and to investigate the influence of spatially directed attention on these suppressive interactions (experiment 2). The design for experiment 1 is presented in Fig. 1. Complex visual images were shown in randomized order in four nearby locations within the right upper quadrant under two presentation conditions: sequential and simultaneous (Fig. 1, A and B). In the sequential condition (SEQ), each of the stimuli was shown alone in one of the four locations. In the simultaneous condition (SIM), the stimuli appeared together in all four locations. Integrated over time, the physical stimulation parameters in each of the four locations were identical under the two conditions. However, suppressive interactions among stimuli could take place in the simultaneous but not in the sequential condition. Thus, on the basis of the results from monkey physiology (4), we hypothesized that the fMRI signals would be smaller during the simultaneous than during the sequential presentations because of the mutual suppression induced by competitively interacting stimuli.

Figure 1

Experimental design. Four complex images (each 2° × 2° in size) were presented in nearby locations at 6° to 10° eccentricity from a fixation point (FP) either sequentially (A) or simultaneously (B). Presentation time was 250 ms, followed by a blank period of 750 ms, on average, in each location. A stimulation period of 1 s is shown, which was repeated in blocks of 18 s. Stimulus location and order of presentation were randomized. New images were chosen out of a pool of 100 for different runs.

Functional MRI scans were obtained from eight people, and data were analyzed by means of multiple regression (5). Sequential and simultaneous conditions were presented in blocks of 18 s each, interleaved with equally long blank periods in the sequence SEQ-SIM-SIM-SEQ. The participant's task was to count T's or L's at the fixation point throughout the scan, which fully engaged the participant's attention at fixation and not at the peripherally presented stimuli (6).

The visual areas that were consistently activated in all participants in the ventral striate and extrastriate cortex during visual stimulation as compared to blank periods were in the calcarine sulcus [Brodmann area (BA) 17], the lingual gyrus (BA 18), and the fusiform gyrus (BA 19 and 37) of the left hemisphere, as illustrated for a single participant in Fig. 2A. Also shown is the assignment of activated voxels to areas V1 to TEO on the basis of meridian mapping (7), which was performed in a separate scan session for each participant (mean Talairach coordinates across all participants were as follows: V1: x = –3, y= –81, z = +8; V2: –9, –78, –10; V4: –19, –74, –14; TEO: –27, –59, –14). As predicted by our hypothesis that stimuli presented together interact in a mutually suppressive way, simultaneous presentations evoked weaker responses than sequential presentations, as shown by the averaged time series of fMRI signals (Fig. 2B, left panel) and by the mean signal differences (Fig. 3A), which were significant in all areas [repeated measures analysis of variance (ANOVA); P < 0.01 for V1, V2, and TEO; P < 0.001 for V4]. The difference in activations between sequential and simultaneous presentations increased from V1 to V4 and TEO (Fig. 3A) [interaction of cortical area and presentation condition: F(3, 15) = 25.1, P < 0.001]; this effect is also reflected in the sensory suppression index (Fig. 3C) [SSI = (R SEQR SIM)/(R SEQ +R SIM); R = averaged responses of the peak MRI intensities obtained during visual presentation blocks for a given condition]. The increase in the magnitude of the suppression index across visual areas suggests that the suppressive interactions were scaled to the increase in receptive field size across these areas. Because of their small receptive fields, individual neurons in V1 and V2 would be capable of processing information only from a very limited portion of our 4° × 4° display, resulting in minimal interaction effects between stimuli; whereas neurons in V4 and TEO, with their larger receptive fields, would process information from all four stimuli in the display, resulting in significantly greater suppressive interaction effects (SSI: V1/V2 versus V4/TEO, P < 0.0001). In further support of this idea, increasing the separation between stimuli decreased the suppressive interactions (8).

Figure 2

(A) Brain areas activated by the complex images as compared to blank presentations. Coronal slices of a single participant at a distance of 25 mm (left) and 40 mm (right) from the posterior pole. Activated voxels were assigned to areas V1, V2, VP, V4, and putative TEO by meridian mapping (7). R indicates right hemisphere. (B) Time series of fMRI signals in V1 and V4 in experiment 1 (left) and experiment 2 (right), averaged over all participants. In experiment 1, sequentially presented stimuli evoked stronger activations than did simultaneously presented stimuli. This effect was much stronger in V4 than in V1 and was replicated in the unattended condition of experiment 2 (unshaded time series). Spatially directed attention (blue shading) increased responses to simultaneously presented stimuli to a larger extent than to sequentially presented ones in V4. Presentation blocks were 18 s in experiment 1 and 15 s in experiment 2.

Figure 3

Mean signal changes and SSIs in areas V1, V2, V4, and TEO, averaged over participants. Results are shown for experiment 1 (A and C) and experiment 2 (B andD). Vertical bars indicate SEM. SSIs increased from V1 to V4 and TEO in experiment 1, which suggests that the effects were scaled to the increasing receptive field sizes of neurons in these areas. This finding was replicated in the unattended condition of experiment 2. In the attended condition of experiment 2, SSIs showed the strongest reduction in V4 and TEO.

In both the sequential and simultaneous conditions, the stimulus presentation rate at any one of the four locations was 1 Hz. However, across the visual field the overall presentation rate in the two conditions was different. To rule out the possibility that the differential responses evoked by the two presentation conditions reflected differences in overall stimulus presentation rate, we sought to demonstrate suppressive interactions directly in a control experiment, in which the presentation rate was held constant. In this experiment, we presented one of the stimuli close to the horizontal meridian in the upper visual field in the absence and in the presence of three other stimuli presented nearby in the lower visual field (Fig. 4); under both conditions, the stimuli were presented at a rate of 1 Hz (9). Because extrafoveal upper and lower visual field representations within early extrastriate areas are located in spatially separated regions, nearby stimuli placed on opposite sides of the horizontal meridian may competitively interact but evoke activations that are separable in the cortex. As shown in Fig. 4 for a single participant, the response evoked in V4's upper field by the single stimulus was significantly greater than the response evoked by the same stimulus presented together with the three stimuli in the lower visual field. The averaged signal changes for all participants tested (n = 3) were significantly different in the two conditions in V4's upper field (paired t test, P < 0.01) (10). This finding supports the idea of suppressive interactions among the stimuli and cannot be explained by stimulus presentation rate.

Figure 4

The representations of V4's upper visual field (UVF) and its lower visual field (LVF) are located medially and laterally in separated but neighboring locations on the fusiform gyrus (left panel). The activity evoked by a single stimulus (2° × 2°) presented at 8° eccentricity just above the horizontal meridian, as compared to blank presentations, was confined to V4's UVF representation (middle panel). As demonstrated in this participant, more activity was evoked in V4's UVF when the stimulus was presented alone than when it was shown together with three stimuli in the LVF just below the horizontal meridian (right panel). In all presentation conditions, stimuli were presented for 250 ms at 1 Hz.

To study the influence of spatially directed attention on suppressive interactions between stimuli, five of the eight participants were tested in experiment 2. This experiment employed a factorial design with two main factors—presentation condition (sequential versus simultaneous) and directed attention condition (unattended versus attended). During each scan, the four blocks of visual stimulation (SEQ-SIM-SIM-SEQ) were tested in an unattended and an attended condition, with the order of the two conditions being counterbalanced across scans (11). In the unattended condition, attention was directed away from the location of the stimuli by having participants count T's or L's at the fixation point, just as in experiment 1. In the attended condition, participants were instructed to covertly attend to the location of the stimulus in the array that was closest to the fixation point and to count the occurrences of a particular target stimulus at that location (12). The target stimulus was indicated by its presentation before each scan. We hypothesized that spatially directing attention to stimuli at one location in the four-element array would reduce the suppressive effects of the surrounding stimuli on the target stimulus in the simultaneous condition (13). Hence, we predicted that attention would enhance the responses to simultaneously presented stimuli more strongly than to sequentially presented stimuli.

In accordance with our hypothesis, the averaged fMRI signal with attention in V4 and TEO increased by 0.84 and 0.62%, respectively, to simultaneously presented stimuli but only by 0.48 and 0.34%, respectively, to sequentially presented stimuli (Fig. 3B). The interaction between the attention and presentation factors was significant in areas V4 (Fig. 2B, blue shaded blocks) and TEO [repeated measures ANOVA; V4: F(1, 4) = 11.2, P< 0.05; TEO: F(1, 4) = 8.5, P < 0.05] but just failed to reach significance in V2 [F(1, 4) = 7.5,P = 0.052]. Thus, the suppressive interactions were partially canceled out by attention. This is also demonstrated by the reduced SSIs in the attended as compared to the unattended condition shown in Fig. 3D. This figure also shows that the magnitude of the attentional effect scaled with the magnitude of the suppressive interactions between stimuli, with the strongest reduction of suppression occurring in V4 and TEO. The results therefore support the second hypothesis that spatially directed attention enhances processing of stimuli in the attended location by counteracting suppression induced by nearby stimuli.

We also found a general increase in activity, affecting the response under both sequential and simultaneous conditions [repeated measures ANOVA; main attentional effect: F(1, 4) = 17.2, P< 0.05] with a significant interaction between cortical area and attentional effect [F(3, 12) = 6.2, P < 0.01] (Fig. 2B, blue shaded blocks; Fig. 3B). The effect of attention was significant in areas V2 (P < 0.05), V4 (P < 0.01), and TEO (P < 0.05) but not in V1 (P = 0.83) (14). These results are consistent with single-cell, event-related potential and imaging studies that have found enhanced responses or increased baseline activity in the ventral extrastriate cortex in response to stimuli presented at attended locations (3, 4, 15).

Our results indicate that, in the absence of directed attention, multiple stimuli in the visual field interact with each other in a mutually suppressive way, as demonstrated by the reduced fMRI signals to simultaneously presented stimuli as compared to sequentially presented ones. Spatially directed attention reduces these interactions by partially canceling out their suppressive effects, as demonstrated by significantly greater effects of attention on the fMRI signal evoked by simultaneously presented stimuli as compared to that evoked by sequentially presented ones. Both the sensory interactions and attentional effects scale with the sizes of the neuronal receptive fields along the ventral object vision pathway. Modulation of suppression at several extrastriate stages may therefore be a mechanism by which attention filters out unwanted information.

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


View Abstract

Navigate This Article