Our brains create a stable view of the world even though our eyes dart around. A study of how the brain might compensate for eye movements reveals an unexpected twist in the vision-stabilizing mechanism. See Letter p.504
Vision is the great illusionist. Although scenes before our eyes seem vivid and detailed, the part of the eye dedicated to high-acuity vision, the fovea, can cover only a narrow sliver of visual space — little more than the breadth of a thumbnail held at arm's length. Vision seems so detailed because we constantly move our eyes to scan the high-acuity fovea across a scene. Saccades — quick, jerky movements of the eyes — occur several times per second, filling in the perceptual fogginess of the visual periphery.
Given this perpetual motion of the eyes, how the visual world looks so stable is an enduring mystery. The retina responds whenever the visual scene 'slips' across its surface, and is thus unable to distinguish between the scene flashing in front of the eyes (for example, a speeding train at a railway crossing) or the eyes rushing across the scene. The brain, however, generates eye movements, and thus could perceptually compensate for saccades1. On page 504 of this issue, Zirnsak et al.2 offer insight into how this compensation could occur.
In 1992, the neuroscientist Michael Goldberg and his colleagues suggested a potential neuronal mechanism for eye-movement compensation3. These researchers studied neurons in the macaque monkey's lateral parietal cortex, a part of the brain that serves as a bridge between vision and eye movements. Parietal neurons, like other visual neurons, have a receptive field (RF), a circumscribed part of the visual field for which visual stimuli activate electrical responses (Fig. 1). These RFs are defined relative to the fixation position (the point in visual space where the fovea is focused when eyes are stationary), and are usually considered spatially fixed relative to that position, reflecting the convergent hard-wired inputs to the neurons that ultimately stem from the retina. Thus, if the fixation position is displaced by a saccade, the RF should 'move' as well, in lock-step with the fixation position.
However, Goldberg and co-workers made the important discovery that parietal RFs are more fluid around the time of saccades: when monkeys made saccades to a point on a computer screen, even before the eyes moved, the neurons could be activated by a spot of light that fell at the future position of the RF relative to the upcoming fixation point after the saccade. The researchers thus suggested that the RF was “updated” or “remapped” to the new position, perhaps to compensate perceptually for the upcoming eye movement. Similar spatial updating was subsequently found in other brain structures that have mixed visual and oculomotor (eye-movement) function, such as the frontal eye fields4 (FEFs) and the superior colliculus5, suggesting a common compensatory mechanism.
A strongly held notion about the spatial-updating hypothesis is that the brain's entire representation of the visual field rigidly translates in preparation for the abrupt movement of the eyes. However, in the original studies of spatial updating, visual stimuli were focused almost exclusively on the expected upcoming location of the RF. Zirnsak and colleagues have now widened the net in the simplest way imaginable: they presented visual stimuli to macaque monkeys not only at the upcoming (post-saccade) location of the RF of FEF neurons, but at various positions all over a computer screen, to map the entire RF.
Surprisingly, when the full RFs were revealed, the authors found that these fields did not rigidly translate before the eye movement, but instead transiently shifted en masse towards the location of the upcoming saccade target (Fig. 1c; see also Fig. 3a of the paper2). Some of the shifts of individual RFs were substantial, up to 18 degrees of visual angle. Moreover, in the few cases in which the original RF was located farther away from the original fixation point than the saccade target (as measured along the direction of the saccade), the shift in the RF to the saccade-target location was actually in the opposite direction to that expected for a rigid translation of the RF relative to the fovea.
The discovery that RFs collapse onto the saccade target is drastically different from the original spatial-updating hypothesis. So why did the authors of the original studies find neuronal responses at the upcoming position of the RF? In fact, the RFs of neurons in the FEF (and related visual–oculomotor structures) are quite large, so even if the RFs collapsed to the saccade target, the fringes of many of those RFs would have probably overlapped with the upcoming RF location, and would thus respond to stimuli at that location. Indeed, in the original studies, responses at the upcoming RF location were often quite weak, which is consistent with the typically weaker responses at the periphery of RFs.
Zirnsak and colleagues' findings raise many issues. For one, RF shifts to a saccade target resemble the RF shifts that occur towards targets on which our attention is focused, even if the eyes do not move6,7. The two phenomena are probably related, inasmuch as shifts of attention to targets of interest precede eye movements. In this perspective, visual stability during saccades could result from the fact that we effectively ignore those parts of the visual scene that are away from the saccade target.
Moreover, spatial–perceptual distortions are known to occur owing to attention shifts and eye movements. These could be interesting phenomena to test further in monkeys, which can signal complex perceptions in experimental settings. More studies will also be needed to understand the cellular mechanisms of rapid RF modulation before saccades. But for experts and non-experts alike, Zirnsak et al. provide us with a valuable lesson: interesting things are often revealed when we search for our proverbial lost keys away from the streetlight.
von Helmholtz, H. A Treatise on Physiological Optics (transl. Southall, J. P. C.) (Dover, 1963).
Zirnsak, M., Steinmetz, N. A., Noudoost, B., Xu, K. Z. & Moore, T. Nature 507, 504–507 (2014).
Duhamel, J. R., Colby, C. L. & Goldberg, M. E. Science 255, 90–92 (1992).
Umeno, M. M. & Goldberg, M. E. J. Neurophysiol. 78, 1373–1383 (1997).
Walker, M. F., Fitzgibbon, E. J. & Goldberg, M. E. J. Neurophysiol. 73, 1988–2003 (1995).
Connor, C. E., Preddie, D. C., Gallant, J. L. & Van Essen, D. C. J. Neurosci. 17, 3201–3214 (1997).
Womelsdorf, T., Anton-Erxleben, K., Pieper, F. & Treue, S. Nature Neurosci. 9, 1156–1160 (2006).