Significance
Humans shift their gaze more frequently than their heart beats. These rapid eye movements (saccades) enable high visual acuity by redirecting the tiny high-resolution region of the retina (the foveola). But in doing so, they abruptly sweep the image across receptors, raising questions on how the visual system achieves stable percepts. It is well established that visual sensitivity is transiently attenuated during saccades. However, little is known about the time course of foveal vision despite its disproportionate importance, as technical challenges have so far prevented study of how saccades affect the foveola. Here we show that saccades modulate this region in a nonuniform manner, providing stronger and faster changes at its very center, a locus with higher sensitivity.
Abstract
Humans use rapid eye movements (saccades) to inspect stimuli with the foveola, the region of the retina where receptors are most densely packed. It is well established that visual sensitivity is generally attenuated during these movements, a phenomenon known as saccadic suppression. This effect is commonly studied with large, often peripheral, stimuli presented during instructed saccades. However, little is known about how saccades modulate the foveola and how the resulting dynamics unfold during natural visual exploration. Here we measured the foveal dynamics of saccadic suppression in a naturalistic high-acuity task, a task designed after primates’ social grooming, which—like most explorations of fine patterns—primarily elicits minute saccades (microsaccades). Leveraging on recent advances in gaze-contingent display control, we were able to systematically map the perisaccadic time course of sensitivity across the foveola. We show that contrast sensitivity is not uniform across this region and that both the extent and dynamics of saccadic suppression vary within the foveola. Suppression is stronger and faster in the most central portion, where sensitivity is generally higher and selectively rebounds at the onset of a new fixation. These results shed light on the modulations experienced by foveal vision during the saccade-fixation cycle and explain some of the benefits of microsaccades.
Human vision is not uniform across space. While the retina collects information from a broad field, only a minuscule fraction—less than 0.01%—is examined at high resolution. This is the area covered by the foveola, the region void of rods and capillaries, where cones are most densely packed. Because of this organization, rapid eye movements, known as saccades, are necessary to redirect gaze toward the objects of interest, abruptly translating the image across the retina every few hundreds of milliseconds. It is remarkable that the visual system appears unperturbed by these sudden visual transitions and seamlessly integrates fixations into a stable representation of the visual scene.
It has long been observed that visual sensitivity is transiently attenuated around the time of saccades, a phenomenon believed to play a role in perceptually suppressing retinal image motion during eye movements. This effect, known as “saccadic suppression”, consists of the elevation of contrast thresholds to briefly flashed stimuli, which precedes the initiation of the saccade and outlasts it by as much as 100 ms (1⇓⇓⇓–5). Saccadic suppression is typically investigated with stimuli that cover large portions of the visual field, often in the periphery. However, limitations in the precision of stimulus delivery, both spatial and temporal, have so far prevented mapping of the saccade-induced dynamics of visibility within the foveola. Thus, despite the disproportionate importance of foveal vision, little is currently known about its time course around the time of saccades.
Studies on saccadic suppression also commonly focus on large saccades under well-controlled, but artificial, laboratory conditions. However, an examination of the time course of foveal vision needs to take into account that natural execution of high-acuity tasks—the tasks that require foveal vision—normally tends to elicit saccades with very small amplitudes (6). Microsaccades, gaze shifts so small that the attended stimulus remains within the foveola, are the most frequent saccades when examining a distant face (7), threading a needle (8), or reading fine print (9), tasks in which they shift the line of sight with surprising precision. Because of their minute amplitudes, microsaccades pose specific challenges to the mechanisms traditionally held responsible for saccadic suppression (10⇓⇓⇓⇓⇓⇓⇓⇓–19). These movements yield broadly overlapping pre- and postsaccadic images within the fovea, which would appear to provide little masking in visual stimulation (20). They also result in reduced retinal smear (21), as they rotate the eye at much lower speeds than larger saccades, delivering luminance modulations that are well within the range of human temporal sensitivity. Furthermore, it is unknown whether possible corollary discharges associated with microsaccades exert similar effects to those of their larger counterparts (22).
Despite these observations, microsaccades have been found to suppress sensitivity to relatively large test stimuli (23⇓–25). However, the only two studies that specifically examined foveal vision during microsaccades reached diametrically opposite conclusions, with one arguing for a normal reduction in sensitivity (26) and the other for a complete lack of suppression (27), leaving open the question of whether suppression extends to the foveola. While several factors could have been responsible for these discrepant results, two important considerations are worth emphasizing. First, selectively testing foveal dynamics is technically challenging, since the entire foveola is comparable in size to the region of uncertainty in gaze localization resulting from standard eye-tracking methods. Second, the common intuition gained by conceptualizing the visual signals delivered by saccades as uniform—i.e., constant-velocity—translations of the image on the retina (28) does not apply well to microsaccades, whose relatively brief durations and well-defined dynamics yield substantially lower power on the retina than predicted by a uniform translation (29). Thus, even a moderate suppression may be sufficient to prevent visibility of stationary scenes during small saccades.
Recent advances in methods for gaze-contingent display control now enable determination of the line of sight with accuracy sufficient to selectively test a desired foveal region during normal eye movements. Leveraging on these recent advances, here we mapped the perisaccadic dynamics of contrast sensitivity across the foveola during natural visual exploration. We developed a gaze-contingent high-acuity task that resembles primate social grooming, a task that very naturally integrates visual search and detection of brief stimuli and that spontaneously elicits frequent microsaccades, and presented probes at desired retinal locations with high spatial and temporal resolution. Our results show that microsaccades are accompanied by an elevation of visual thresholds at the center of gaze that starts before the initiation of the movement but dissipates very rapidly as the saccade ends. The extent and dynamics of this suppression vary with eccentricity across the foveola, so that a stronger modulation occurs in the most central region, where vision is selectively enhanced after a saccade.
Results
In a simulated grooming task, observers reported the occurrence of “flea jumps” (the probes), brief changes in the luminance of otherwise dark dots located within the central 2°2° region of a wide naturalistic noise field. Subjects freely moved their eyes, searching for the locations at which these contrast pulses would occur, while their eye movements were continually recorded.
In reality, unbeknownst to the subject, the probes were activated on the basis of the position and movement of the eye to measure visibility within selected regions of the fovea and at various time lags around saccades. This was possible due to three state-of-the-art components: 1) high-resolution eye tracking achieved via the Dual Purkinje image method (30); 2) accurate gaze localization obtained by means of an iterative gaze-contingent calibration, a procedure that improves accuracy by approximately one order of magnitude over standard methods (6); and 3) real-time control of retinal stimulation, obtained via a custom system for flexible gaze-contingent display control, Eye movement Real-time Integrated System (EyeRIS) (31).
As expected, this high-acuity task resulted in the frequent occurrence of minute saccades. On average, observers executed ∼2.5 saccades per second, almost all of them smaller than 1°1° (average saccade amplitude and SD across subjects, 28′±6′28′±6′; mean 99th percentile of the amplitude distributions, 68′68′; Fig. 1E). In fact, the majority of saccades (68%) were smaller than 30′30′, an amplitude range that maintains an initially foveated probe well within the foveola. These tiny gaze shifts occurred at a rate (1.6 microsaccades per second) much higher than those normally encountered in tasks that do not involve high visual acuity (typically <0.2<0.2 microsaccades per second) (29), an observation consistent with the notion that microsaccades are normally part of the strategy for examining fine spatial detail (6, 7).
Because of their small amplitudes and stereotypical dynamics (32), the saccades performed in this task resulted in relatively slow changes in visual stimulation. This, combined with the characteristics of the visual scene, which, like natural images, possessed predominant power at low spatial frequencies, resulted in luminance signals to the retina that were well within the range of human temporal sensitivity as measured previously (Fig. 1F) (33). Yet, as happens for larger saccades, subjects were not aware of the resulting translations of the images on their retinas—the well-known phenomenon of saccadic omission (3).
To quantitatively examine the consequences of saccades on foveal sensitivity, we binned contrast pulses according to their combinations of retinal locations and lags relative to saccade occurrence and separately estimated contrast sensitivity in each spatiotemporal interval (Fig. 2A). Fig. 2B shows the psychometric functions of contrast sensitivity measured for one subject in three spatiotemporal bins. As these examples show, sensitivity varied considerably not only with the timing of the probe relative to saccades, but also with its position on the retina, reaching, in some instances, low values even at the highest possible contrast.
We first examined sensitivity far from saccades. The data points in the shaded region in Fig. 2C represent the average thresholds across observers estimated during fixation, i.e., when no saccade occurred in the surrounding ±200±200 ms of a probe. Strikingly, despite being separated by just a few arcminutes, the three considered foveal regions exhibited marked differences in sensitivity. Contrast sensitivity was always larger at the very center of gaze and decreased with increasing eccentricity, so that sensitivity in the most central region (the region within 15′15′) was on average 8% higher than in the range 15 to 30′30′, which was in turn ∼9% higher than sensitivity in the range 30 to 60′60′ (one-way ANOVA, F(2,17)=4.8F(2,17)=4.8; P=0.02P=0.02). These measurements reveal how contrast sensitivity varies across the central foveola. They show that, contrary to its anatomical homogeneity, sensitivity is not uniform within this region: Optimal sensitivity is restricted to a very narrow region around the center of gaze during normal fixation.
As the probe approaches the onset of a saccade, drastic changes in visual sensitivity occur. Sensitivity drops sharply from the fixation baseline starting ∼50 ms before the saccade and continues to be affected up to ∼100 ms after the saccade onset, a time at which the saccade has typically already ended (Fig. 2C). At all the considered foveal locations, suppression was strongest in the 25-ms interval immediately preceding the saccade, when sensitivity dropped by ∼38% on average. The dynamics of this effect were highly stereotypical across subjects, all of whom individually exhibited a similar and statistically significant attenuation in sensitivity (P<0.05P<0.05, nonparametric bootstrap; individual subject data in SI Appendix, Fig. S1). Thus, the minute saccades performed in our experiment were accompanied by a strong attenuation in sensitivity throughout the foveola, an effect qualitatively similar to the saccadic suppression observed elsewhere in the retina for larger saccades.
While suppression occurred over the entire foveola, the extent and time course of the process differed across foveal regions. All regions ended up with similar visibility levels at the peak of the suppression. However, since sensitivity in distinct regions started from different fixation baselines, the amplitude and speed of the process also varied, so that the change in sensitivity was larger and faster in the most central region of the foveola than at other locations. On average in the 100-ms interval centered at saccade onset, sensitivity was attenuated by 33% in the central region with eccentricity smaller than 15′15′, whereas it was reduced by only 23% in the 30to60′30to60′ region (P<0.001P<0.001; post hoc Tukey–Kramer comparison; Fig. 2E). Thus, given the similar overall duration of the effect across the foveola, both suppression and recovery were faster at the very center of gaze than at larger eccentricities (Fig. 2D).
These results were robust relative to the specific methods for data analysis. Very similar results were obtained by measuring sensitivity to changes in the Weber contrast of the probe relative to its surroundings rather than the Michelson contrast of the probe alone (SI Appendix, Fig. S2). Furthermore, differences in foveal dynamics were also reflected in the reaction times of manual responses, which were longer when the probes were less visible (r=−0.50r=−0.50, P<0.01P<0.01; SI Appendix, Fig. S3A). At the time of microsaccade onset, the reaction times for probes displayed at the very center of gaze were on average 23% longer than for probes just a few arcminutes away (one-way ANOVA, F(2,16)=8.59F(2,16)=8.59, P=0.004P=0.004; SI Appendix, Fig. S3B).
To better examine the temporal evolution of saccadic suppression, we recomputed the time course of sensitivity relative to two distinct temporal events, the end of a saccade and the time at which a saccade reaches its peak speed. Visibility recovers extremely rapidly following a saccade. On average across foveal regions, sensitivity has returned to about 90% of its presaccadic value less than 25 ms after the saccade ends and is fully restored within an additional 25 ms (Fig. 3A). This happens because suppression largely precedes the actual movement of the eye. Suppression is already recovering by the time a saccade is in midflight and has reached its peak velocity (Fig. 3B), an asymmetric temporal evolution that is evident when comparing sensitivity with equal speed of the retina before and after saccade peak speed (SI Appendix, Fig. S4)….
