This study was conducted to investigate fixation behavior of both eyes under binocular and monocular viewing. We wished to determine whether the eyes behaved the same or differently motivated by results of previous work showing asynchronous and unpredictable behavior of an occluded eye during midline pursuit [23].
Surprisingly, the eyes’ movements were not identical during binocular viewing, the standard protocol used to characterize eye movements, in fixation experiments where only one eye is measured. Fig 2 shows raw eye position traces for each eye from one observer (S1) during binocular fixation. Ocular “drift” is dissimilar between the eyes in both horizontal and vertical directions. However, microsaccades consistently occur at the same time and in the same direction in the two eyes.
The differences in ocular behavior suggest that the drift was independently controlled for the two eyes. However, an alternative explanation is that uncorrelated neural noise at the level of the output circuitry controlling the eye muscles was responsible for the differential eye movement behavior.
To determine whether the observed differences in ocular behavior were due to independent control or neural noise, we covered one of the eyes effectively decoupling the influence of the stimulus from the yoking-like behavior observed in oculomotor experiments. If uncorrelated noise was creating the differential ocular behavior, covering an eye should not change the general character of that eye’s behavior. However, if the eyes are controlled independently, the independent control mechanism for the covered eye should suffer without a visual anchor.
When covering an eye, we found striking differences in the ocular behavior relative to binocular viewing in terms of the excursion variability and drift speed, as well as differences in microsaccade magnitude.
Fixation variability
To quantify the differences in excursion variability between the eyes, we computed the 95% bivariate contour ellipse area (BCEA), a 2-dimensional measure of standard deviation (see Methods), for each eye during monocular and binocular viewing.
Figure 3A shows BCEAs (red dashed lines) superimposed on fixation density plots (see Methods) for each eye of a typical observer in each condition. During binocular viewing, the BCEA is more compact compared to BCEA’s during monocular viewing, demonstrating better control of fixation. However, note that even during binocular viewing the right and left eye BCEAs are not identical with the right eye’s BCEA being larger than left eye’s BCEA for this observer. During monocular viewing a strikingly different pattern emerged. The covered eye now showed much greater variability than it did during binocular viewing. Even more surprisingly, the monocular viewing eye’s BCEA was larger than that of the same eye during binocular viewing.
Figure 3B summarizes the BCEAs for monocular and binocular viewing for each subject, as well as the group means. To characterize the differences in BCEA during binocular viewing, we compared the eye with the smaller BCEA in each subject to the eye with the larger BCEA and found their difference to be significant (Fig 3B) (mean smaller BCEA = 0.498±0.08, mean larger BCEA = 0.668±0.11, t(7)=-4.76, p < .01). During monocular viewing, we expected that the covered eye would wander more, since it would be less able to maintain its position without a visual target to anchor it. Yoked control proposed by Hering would allow the viewing eye’s use of the target to keep the covered eye stable, but the Chandna et al. (2021) finding that the covered eye’s behavior was not determined by the target implied less stability here. Consistent with this, the covered eyes’ BCEAs were significantly larger than those of the fellow viewing eyes (mean viewing BCEA = 0.829±0.05, mean covered BCEA = 1.681±0.07, t(7)=-2.10, p < .05). This result suggests that the covered eye was more poorly controlled when it was unable to view the fixation stimulus, and furthermore that its control was not coupled to that of the viewing eye. Finally, the BCEA’s of the monocular viewing eyes were significantly larger than the BCEAs during binocular viewing (mean smaller BCEA = 0.498±0.08, mean monocular viewing BCEA = 0.829±0.05, t(7)=5.93, p < .001; mean larger BCEA = 0.668±0.11, mean monocular viewing BCEA = 0.829±0.05, t(7)=3.11, p < .01). Apparently, the monocular viewing eye was more poorly controlled than it was when both eyes viewed the fixation stimulus. Therefore, even though it appears that different circuitry controls each eye, a benefit was gleaned from both eyes viewing the target.
Drift Speed
A larger BCEA might result from the eye being more poorly controlled, leading to a less stable retinal image. Further evidence for poorer control with larger BCEAs was discovered when we investigated the relationship between BCEA size and the speed of the eyes’ drift.
Figure 4 shows mean drift speed in the different conditions for each observer superimposed on the group mean. Overall, drift speed was no different between binocular and monocular viewing (p = .97) and varied considerably across observers. Moreover, the drift speeds did not differ between the two eyes during binocular viewing (p = .45). While overall the covered eye moved faster than the monocular viewing eye this result did not reach significance (mean viewing = 0.879±0.04, mean covered = 0.954±0.07, t(7)=-1.91, p = .097). Despite failing to reach significance, there appeared to be a rough relationship between BCEA and drift speed. To explore this, we correlated these measures for each eye in both viewing conditions. Figure 5 shows that the correlation during binocular viewing for each eye was weak and did not reach statistical significance (Pearson correlation, eye with smaller BCEA: t(6)=1.71, p=0.14 p= 0.1588; eye with larger BCEA: t(6)=1.29, p=0.24). However, for both eyes during monocular viewing the correlations between drift speed and BCEA were significant (Pearson correlation, covered eye: t(6)=4.10, p=0.0064; viewing eye: t(6)=3.60, p=0.011). The relatively constant relationship between a smaller BCEA and slow drift speed supports the idea that when the eyes are better controlled, smaller BCEAs result. Conversely, strong correlations between drift speed and BCEA during monocular viewing suggest that the eyes were more poorly controlled than they were during binocular viewing.
Microsaccades
One function of microsaccades is to correct for the eye’s displacement from a target, presumably because drift moved it away.[34] However, experiments in previous studies were performed when only one eye was recorded, assuming drift and corrective microsaccades would operate the same in the unrecorded eye (see [35] for a review). The current study provides only limited support for this assumption. Regarding microsaccade rate, some degree of yoking is apparent, as every microsaccade we recorded occurs in both eyes. However, the size of microsaccades can vary between the two eyes, so the eyes do not appear to be strictly yoked during microsaccades, yoking is merely apparent when they are triggered.
Figure 6 shows microsaccade rate and size across conditions. Fig 6A shows that microsaccade rate was identical for the two eyes during binocular viewing, as saccades were always binocular. This was also true for the covered and viewing eyes during monocular viewing. It has been reported previously that a microsaccade in one eye is always accompanied by a microsaccade in the other eye during binocular viewing [36], and apparently this is also the case when only one eye views the target. Interestingly, there was a tendency for more microsaccades to occur during monocular than binocular viewing, but this difference did not reach significance (mean binocular = 0.717±0.13, mean monocular = 0.928±0.17, t(7) = -2.66, p = 0.11) .
Fig 6B plots the magnitude of microsaccades in the different conditions for all observers. Microsaccade magnitude was marginally significantly different between the eyes in the binocular condition with the eye with the smaller BCEA having smaller microsaccades (mean smaller BCEA = 0.391±0.034, mean larger BCEA = 0.441±0.045, t(7)=-2.32, p = .053). During monocular viewing, saccade magnitudes were also slightly larger in the covered than the viewing eyes (mean viewing = 0.404±0.029, mean covered = 0.425±0.030, t(7)=-2.291, p = .056). This difference is curious though, as it suggests that microsaccades, while apparently always occurring in both eyes, are not completely conjugate when an eye is covered and even when viewing is binocular.
Microsaccades are thought to correct for error induced in their position when drift carries the eyes away from a target. [34] If so, larger microsaccades might be required to correct error when BCEAs are larger, because larger BCEAs indicate that an eye has strayed further from the target. To determine this, we correlated BCEA size with microsaccade magnitude for each eye in the different conditions (Fig 7). As expected, microsaccade magnitude was well correlated with BCEA size and significantly so during binocular viewing, both for the larger BCEA (t(6)=2.97, p=0.025) as well as for the smaller one (t(6)=4.93, p=0.0026). During monocular viewing, microsaccade magnitude was also significantly correlated with BCEA size for the viewing eye (t(6)=3.46, p=0.013). Curiously though, this correlation was not present for the covered eye (t(6)=0.57, p=0.59). We think this is because the microsaccades in the covered eye did not have a visual target to correct to, and therefore the drift and microsaccades were decoupled in that eye.
Instead, it appeared that the covered eye’s microsaccades were driven by the microsaccades in the viewing eye, as these were highly correlated (R=0.989; t(6)=16.25 p=0.000). Since the viewing eyes drift was apparently indirectly driving the microsaccades in the covered eye, it is not surprising that these two values were highly correlated as well (R=0.774; t(6)=2.99 p=0.024).