Fast Backprojections from the Motion to the Primary Visual Area Necessary for Visual Awareness

See allHide authors and affiliations

Science  20 Apr 2001:
Vol. 292, Issue 5516, pp. 510-512
DOI: 10.1126/science.1057099


Much is known about the pathways from photoreceptors to higher visual areas in the brain. However, how we become aware of what we see or of having seen at all is a problem that has eluded neuroscience. Recordings from macaque V1 during deactivation of MT+/V5 and psychophysical studies of perceptual integration suggest that feedback from secondary visual areas to V1 is necessary for visual awareness. We used transcranial magnetic stimulation to probe the timing and function of feedback from human area MT+/V5 to V1 and found its action to be early and critical for awareness of visual motion.

Two hypotheses that were postulated about how brain activity mediates awareness have particular relevance to neurophysiology (1–4). There might be a class of neurons or neural pathways whose activity mediates awareness. Alternatively, awareness might be the result of specific forms of neuronal activity such as synchronous discharges or spike rate modulations. These hypotheses are not mutually exclusive, and a combination of both might be proposed (4). The role of striate cortex (V1) in visual awareness (4–8) is controversial; it may not have specialized “awareness-dedicated” neurons, but the spiking rate of V1 neurons appears to be modulated by perceptual context, correlated with awareness and dependent on backprojections from higher visual areas (4, 9, 10–13). The function and time course of these backprojections is beginning to be explored in nonhuman primates but has not been investigated directly in human subjects (9–15), although the organization and chronometry of feedforward-feedback loops provide a substrate that may give rise to visual awareness (16,17).

When transcranial magnetic stimulation (TMS) is applied to visual cortex, subjects may perceive phosphenes (flashes of light) (18, 19) and, when applied to area MT+/V5, subjects can perceive moving phosphenes (20–22). To examine the temporal relation between events in V5 and V1, we applied TMS to both areas with variable delays of tens of milliseconds (23–27). TMS was applied such that stationary phosphenes evoked from V1 and moving phosphenes evoked from V5 would overlap in visual space (28, 29) (Fig. 1). TMS was applied over V5 at 100% of V5 phosphene threshold and over V1 at 80% of the V1 phosphene threshold (30, 31). After each pair of pulses, the subjects were asked to describe their perception of phosphenes and to rate them using a fixed, forced-choice scale (32).

Figure 1

Brain MRI from a subject showing the site of stimulation for induction of stationary (V1) and moving (MT+/V5) phosphenes. The location on the scalp of TMS coil is projected onto the anatomical MRI (23,26).

If the V5 to V1 backprojection is critical for awareness, disruption of activity in V1 at the time of arrival of feedback should interfere with the perception of attributes encoded by the extrastriate area. The use of V1 stimulation below the threshold for production of a phosphene precludes a simple masking of a V5 phosphene by a V1 phosphene (33). Furthermore, the intensity of V1 stimulation used was below that required to disrupt psychophysical performance (34) or to induce scotomas (35). The use of a range of V5 to V1 TMS asynchronies controlled for nonspecific effects of TMS, such as the sound made by the discharge of the coil. In a control experiment, pairs of a subthreshold and a suprathreshold transcranial stimulus with variable interstimulus intervals were applied to V5 to assess the contribution of local V5 effects and fast V1 to V5 projections (36).

When TMS was applied to V1 before V5, there was no effect on the perceived movement of the phosphenes (Fig. 2). However, with a V5 to V1 asynchrony of +5 to +45 ms (that is, V5 TMS applied before V1 TMS), there was a marked decrease in the quantity and a change in the quality of the phosphenes elicited by V5 stimulation (Fig. 2). Five of the eight subjects reported an absence of phosphenes when TMS to V1 was applied up to 25 ms after V5. All subjects reported that in the majority of the trials the phosphene was stationary, rather than moving, when TMS was applied to V1 up to 45 ms after V5. Paired-pulse TMS to V5 had no effect on the perception of the moving phosphenes in any of the subjects, regardless of interstimulus interval (Fig. 2).

Figure 2

Mean responses of all subjects (n = 8) to combined stimulation of V5 and V1 (23) are shown by the circles. Results of a control experiment of five subjects with paired stimulation to V5 are shown by the squares (36). In the V5 to V1 experiment, negative values indicate that V1 received TMS before V5, and positive values indicate that V1 was stimulated after V5. In the V5 to V5 stimulation experiment, the conditioning stimulus was a subthreshold stimulus to V5, and the test stimulus was suprathreshold (36). In both experiments, subjects made one of four judgements (32). The phosphene elicited by V5 TMS was (1.0) present and moving, (2.0) present but the subject was not confident to judge whether it was moving or moving differently, (3.0) present but stationary, (4.0) no phosphene was observed. TMS over V1 between 5 and 45 ms after TMS over V5 disrupted the perception of the phosphene. A conditioning stimulus to V5 did not affect the effect of the V5 test stimulus regardless of interval (squares). The individual data of all subjects in both experiments are available (43).

Our results correspond with findings from physiologic studies of monkey area MT+/V5 (9–15), suggesting that the V5 projection to V1 operates with a short time course (37). This finding contradicts the chronometrically naı̈ve assumption that the “top-down” influences of feedback projections should occur late (38, 39). The latencies of some MT+/V5 neurons and the conduction times of the V5 to V1 pathway in recordings from monkeys are sufficiently fast to account for the early effects seen here (9–15,40–42).

These results demonstrate the importance of V5 to V1 backprojections for perception and awareness of visual motion (33,43–47). TMS in blindsight patient G.Y. is unable to induce moving phosphenes from the hemisphere with the traumatic lesion of the left occipital area, including V1 (3, 8, 22, 48). This supports the idea of the critical role of V1 in visual awareness. However, G.Y. can be aware of temporal change in his blind field (3, 8, 48). Functional magnetic resonance imaging (fMRI) studies reveal that MT+/V5 is always active when moving stimuli are presented to G.Y. in his blind field and that this activity correlates with G.Y.'s awareness for moving stimuli. An association between cortical activity as detected by fMRI and behavior (awareness in this case) does not establish a causal link between them. More important, awareness for moving stimuli in G.Y. may mean awareness of change (temporal change) rather than awareness of movement. Furthermore, G.Y.'s awareness of motion may well not be “visual awareness” but rather a metamodal alerting response. Finally, the inability to perceive moving phosphenes evoked by TMS, despite being able to be aware of visually presented moving stimuli, may suggest a necessary distinction between awareness for internally generated percepts, such as imagery, and externally driven visual percepts normally associated with vision.

Our results highlight the importance of the fast feedback projections from V5 to V1 in visual awareness of motion and document the chronometry of the phenomenon. Whether these findings and this TMS-based approach to study perceptual awareness can be extrapolated to externally driven percepts or points out a fundamental distinction between types of perceptual awareness remains to be explored.

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


View Abstract

Navigate This Article