Normality was tested using the Shapiro–Wilk test. Alpha was set at 0.05. For multiple comparisons, p-values were adjusted using the Šídák test for multiple comparisons. All secondary statistical analyses were performed using Prism 10 software (GraphPad Software LLC.).

Figure 2 shows two example head-unrestrained trials from participants P2 (Figures 2A,B) and P3 (Figures 2C,D) from the original study (Supplementary Table S&V1).

P2 typically did not initiate pursuit with the eyes prior to moving the head. When the head did begin to move, however, she was not able to sufficiently cancel the VOR for the duration of the trial, with eye-in-space (gaze) velocity being less than target velocity for most of the trial (Figure 2B). At the end of the trial, the head moved in the opposite direction of the target with a clear compensatory eye movement moving the eye in the opposite direction of the head (but now with the target). The position of the eyes in the head remained relatively constant.

Participant P3 initiated pursuit with the eyes only and moved the head later in the trial (possibly to increase pursuit range). At the initial acceleration of the head (Figure 2D, gray arrow), the eyes moved in a compensatory direction, and the overall gaze velocity was reduced. However, as the trial went on, the participant was better able to coordinate her eye and head movements, and gaze velocity returned to be closer to target velocity.

Figure 1 shows representative average traces of eye, head, and gaze velocities across all trials for the rightward (0°) direction for the head unrestrained (blue and black lines) and head-restrained (red) conditions. Key features to note include the initial pursuit by the eyes only (marked as “NO head motion”), higher gaze velocity in the head-restrained than unrestrained condition for the control participant and the eventual reduction of the eye-in-head velocity with the increase in head velocity in the head-unrestrained condition.

We characterized the eye motion during head-unrestrained pursuit as a combination of a scaled (K_fix) smooth pursuit eye command for the head-restrained condition and delayed compensatory response to head motion (K_v, see Methods, Section 2.3, Figures 1C,D). The eye and head time series used in each model for each participant in each direction lasted from target onset to the timepoint of peak head velocity.

Because target direction can significantly affect pursuit gain in individuals with CFL (17), we provide pursuit and VOR coefficient estimates for all directions individually in Figures 3A,B. Although we saw a trend for lower K_fix values for the vertical directions, there was no significant effect of direction [two-way mixed effects model with Geisser–Greenhouse sphericity correction, K_fix: F (1.45, 20.82) = 1.95, p = 0.17; K_v: F (1.25, 12.89) = 1.96, p = 0.19]. Further, there was no effect of participant group [K_fix: F (1, 32) = 0.66, p = 0.42, 95% CI: (−0.094, 0.222); K_v: F (1, 12) = 1.10, p = 0.32, 95% CI: (−0.473, 0.166)]. For additional statistical parameters, see Supplementary Table S1.

While the effect of direction was not significant, the horizontal and vertical trials were not equivalent, with shorter trajectories and trial durations for the vertical directions. Therefore, we focused on the horizontal directions, for a more direct comparison to prior work (24, 25). Having averaged the leftward and rightward conditions, we observed K_fix < 1 (i.e., a reduction in the smooth pursuit velocity for head-free versus head-restrained pursuit; CFL: K_fix = 0.90 ± 0.16; Control: K_fix = 0.74 ± 0.08; Figure 3C) for both groups, unlike in (25) (Control: K_fix = 1.01 ± 0.14, Extinction: K_fix = 0.98 ± 0.23, black and gray dots respectively).

To determine whether this reduction was indeed significant on a group level, we performed a one-sample t-test (Table 2) which indicated that our older adult control group had K_fix estimates significantly below 1 (suggesting different eye pursuit velocities between the head-free and head-fixed conditions, Row 1, K_fix columns, Table 2). We did not find a significant difference for the CFL group (Row 2, K_fix columns, Table 2) or the K_fix estimates from (25) (Rows 3 and 4, K_fix columns, Table 2). We did find a significant difference in K_fix estimates between our two groups (two-tailed t-test, Row 5, Table 2).

Examining K_v estimates, on the other hand (Figure 3D), we did not observe a deviation from K_v = −1 (i.e., VOR gain = 1) for either of our groups (one-sample t-test, Table 2, K_v columns, Rows 1 and 2), similar to the control (non-extinction) values in (25) (Table 2, K_v columns, Row 3) and unlike the extinction condition in (25), where the authors reported K_v = −0.85 ± 0.10, corresponding to a 15% cancellation of the VOR during pursuit (Table 2, K_v columns, Row 4). We did not find a significant difference in the K_v coefficients between our two participant groups (Row 5, Table 2). Additionally, there was no difference between our control participants and the control condition in Ackerley and Barnes (25), one-way ANOVA with a Holm-Šídák’s multiple comparisons test, p = 0.73. Participants with individual trials shown in Figure 2 are marked in Figure 3D for reference.

Estimates of K_fix < 1 suggest that the eye velocity response in the two head restraint conditions was different. To test this interpretation, we calculated eye-in-space velocity immediately prior to head movement onset for each trial, where pursuit would already be initiated, but no head movement contribution can yet occur (see Methods, Section 2.3). These values were then averaged for each target direction, for each participant. Figure 4 shows eye velocities in the head-free condition compared with eye velocities in the head-restrained condition for each CFL (Figures 4A,C) and control (Figures 4B,D) participant.

To determine whether eye velocities prior to head movement were indeed different from eye velocities in the head-restrained condition at that same timepoint, we performed a two-way repeated measures ANOVA and found a significant effect of head condition for both CFL [F (1, 20) = 10.18, p = 0.005, 95% CI: (−1.900, −0.398), η_p² = 0.34] and control [F (1, 23) = 49.17, p < 0.0001, 95% CI: (−2.248, −1.224), η_p² = 0.68] groups. We subsequently performed a Šídák test for multiple comparisons of head-restraint conditions at each target direction. We found no individual target direction effects for the CFL group (p > 0.07). In the control group, there was a significant difference between head restraint conditions for each target direction (Šídák multiple comparisons test, adjusted p-values: 0°: p = 0.004, 90°: p = 0.002, 180°: p = 0.01, 270°: p = 0.03).

To ascertain if this head-restraint-based change in eye velocity was velocity dependent, we performed a simple linear regression for each direction and each group. In the CFL group, the slopes all ranged between 0.86 (0°, rightward) and 2.23 (90°, downward) and were not significantly different upon comparison [F (4, 33) = 0.55, p = 0.70]. We found a similar trend in the control group, where slopes ranged between 1.01 (0°, rightward) and 1.60 (270°, upward) and were not significantly different from each other [F (4, 34) = 0.6376, p = 0.64]. Thus, we calculated a single slope for each group, both of which were near-one (CFL: slope = 0.85, s.e. = 0.20; Control: slope = 1.07, s.e. = 0.11, Figures 4C,D respectively). There was no significant slope difference between groups [F (1,47) = 0.98, p = 0.33]. See Supplementary Table S2 for detailed fit statistics.

One possible explanation of the difference in eye velocities between conditions is a difference in pursuit latencies – a longer pursuit latency in the head-unrestrained condition could lead to lower eye velocities. To test this possibility, we calculated pursuit latency offset between the two conditions for both groups. Median pursuit latencies were exceedingly close to 0 for both groups (CFL: −0.0095 s, Control: 0.004 s, One-Sample Wilcoxon test, p > 0.25, W_Control = 33, W_CFL = −78).

Given the limited published data on VOR in CFL, we performed an additional experiment where we asked participants to rotate their heads periodically while viewing an LED dot stimulus at a viewing distance of 2 meters. Figures 5A,B shows representative velocity traces for the eye-in-head (red) and head-in-space (black) from a participant with CFL (VP5, Table 1) and a control (VC2, Table 1) participant. Both show similar eye and head movement profiles with a near-equal and opposite rotation of the head and eyes in orbit. Indeed, we observed similar (near-one) gains for both groups [Figure 5C, CFL: Gain_VOR = −1.00 ± 0.05; Control: Gain_VOR = −1.03 ± 0.02; t = 1.12, p = 0.27, 95% CI = (−0.085, 0.027), Cohen’s d = 0.754].

In our study (26) of head-unrestrained smooth pursuit in individuals with CFL, we found group differences between eye, but not head, velocities during pursuit, which were also evident in the overall gaze gain during pursuit. These differences persisted in both head-restrained and head-unrestrained conditions (26). Moreover, we found that in the head-unrestrained version of the task, eye velocities in the CFL group were not only significantly lower than in the head-restrained condition but were sometimes negative [Figure 3A in (26), black circles], suggesting that the eyes were moving opposite the target. We did not observe these oppositely directed movements for the head velocity, however. Thus, the eyes were moving not only opposite the target, but also the head – a behavior suggestive of VOR. Indeed, when we examined this behavior on an individual trial basis, we observed that head velocities were sometimes accompanied by compensatory eye-in-head movements that appeared to be equal and opposite the head. In both the control and CFL groups, we found that overall gaze velocities (and therefore pursuit gains) were lower in the head-unrestrained, than restrained condition, with pursuit gains in CFL participants being lower than in controls, for both conditions (Figure 1 of the original study).

We hypothesized that the lower pursuit gains in the CFL group in both head-restraint conditions were likely due to a combination of oculomotor deficits and target disappearance in the scotoma (19). Previously, Ackerley and Barnes (24, 25) showed that participants could extract target motion information from a brief presentation to appropriately scale gaze velocity (25). In the extinction trials, the authors saw a contribution of ~15% uncompensated-for head movements to pursuit of temporarily extinguished targets for an overall improvement of pursuit gaze velocities (25). From our own and others’ work, we know that motion perception is generally not impaired in CFL (35, 36).

Further, Ackerley and Barnes found that target extinction duration or whether it was blanked at the beginning of or during the trial did not have a significant effect on either K_v or K_fix estimates (Equation 1) (24, 25). This outcome would suggest that the variability we see in our CFL participants may not be directly attributable to their inability to keep the target visible at various stages of the trial during head-unrestrained pursuit. In other words, given the highly repeatable nature of our trials and constant target velocity during the ramp, participants with CFL should be able to sustain gaze velocity using efferent/memory information. Taken together, these outcomes suggest that head pursuit in conjunction with VOR suppression can provide an effective adaptation to overcome pursuit deficits due to intermittent target disappearance in CFL.

However, while in the prior study (25), when targets were extinguished during a part of the trial, VOR suppression was estimated to be ~15%; CFL participants in our study had on average near-one VOR gains (K_v = −0.98), suggesting that the head movements were nearly-perfectly compensated for by the eyes in orbit. This strategy made head movements an ineffective contribution to smooth pursuit in CFL. Interestingly, in our control participants we did observe K_v values that were slightly greater than those estimated by (25) for their control (no extinction) trials. However, overall, this difference was not significant.

An interesting outcome of our study was the consistently lower K_fix estimates in our control participants than those observed in (25). The K_fix coefficient corresponds to the difference between eye velocity component of pursuit during head-restrained and head-unrestrained versions of the task. The authors in that study assert that because the internal program for pursuit is likely to be the same regardless of head restraint, the eye-in-head velocities in these two conditions should be driven by the same visually driven and extraretinal components and should thus be equivalent. The authors do allow, however, “that although [these eye-in-head velocities] share similar dynamic characteristics they may differ in magnitude” given response variabilities in their own data.

Taken together with our prior findings that gaze velocities are lower during head-unrestrained pursuit (26), lower eye-in-head velocities in the head-unrestrained trials (K_fix < 1), and proportionate compensatory responses to head velocities (K_v ~ 1) suggest that a potential mechanism for pursuit deficits in CFL is due to: (1) lower eye movement velocities when there is an anticipation of head movement (head-unrestrained condition), particularly in our control cohort, combined with (2) a lack of VOR cancellation in CFL (and to a possibly lesser degree controls) in smooth pursuit.

In terms of the decrease in eye velocity in the head-unrestrained condition, our data suggest that in our cohort of older adults, there is an overall attenuation in eye-driven pursuit when the head is free to move, regardless of CFL. Our analysis of latency offsets between conditions indicates that it is unlikely to be due to a change in pursuit latency between conditions. On the other hand, this eye velocity difference may be indicative of an overall increased reliance on head movements for gaze shifts in older age (37). This interpretation is consistent with the proportional attenuation we see in eye velocity between conditions where regardless of the eye velocity command in the head-restrained condition, there is a similar decrement in velocity across participants and directions. It is possible that this decrement could be due to an additional head pursuit command. Taking this conjecture one step further, a less pronounced change in eye velocities between conditions in our CFL group could suggest that head movements may not be an effective gaze-shifting strategy due to a lack of VOR cancellation in CFL. Thus, the emphasis on head movements is decreased and changes in eye velocity are less systematic between conditions.

Shifting to VOR cancellation, one may argue that the reason we found near-one K_v estimates in the CFL group is not a lack of VOR cancellation, but that these participants did not experience target extinction due to its disappearance in the scotoma throughout the trial. If that were indeed the case, their task would be more like the controls’ task in our study or the control task in (24, 25). Indeed, participant P1 (Supplementary Table S&V1 and Table 1 in (26)) had the largest scotoma, lowest pursuit gains, and one of the highest K_v estimates (K_v = −0.80). This explanation is rather unlikely, however. First, our participants in the head-unrestrained condition had pursuit gains that suggested significant retinal slip (26). Indeed, the participant for whom we observed the greatest amount of VOR cancellation [P5 in (26), K_v = −0.66] had one of the smallest binocular scotomas. Second, in our prior work using the scanning laser ophthalmoscope, we documented target disappearance in the scotoma during pursuit as rather common (17). Finally, we have shown previously that saccades during pursuit do not sufficiently make up for shortfalls in displacement (26), likely due to their misdirected nature (20) and were therefore unlikely to effectively keep the target out of the scotoma throughout all trials.

Perceptual filling in in the scotoma (18) may be another consideration. Unlike in the case of target disappearance or an occluder, individuals with CFL are often not aware of their scotoma and instead of a blanking, they experience a filling in of the missing information from their periphery. In this case, they may not be consistently aware that the target disappears during the trial. Indeed, optokinetic nystagmus (OKN) has been shown to be generated by a filling in of an OKN stimulus (38). For pursuit stimuli, however, prior work has shown that individuals with CFL have poorer motion extrapolation across the scotoma (19) and filling in is known to be less effective for non-uniform stimuli (18). Regardless of the specific perception of the target, pursuit gains were significantly below one in our population (26), which would suggest the need for head contribution for improved pursuit.

In their study, Ackerley and Barnes instructed participants to “move the eyes and head” to track the target, whereas we asked the participants to pursue targets in the most natural and comfortable manner possible. As a result, we saw lower head velocities and a greater heterogeneity of pursuit strategies in CFL and control participants. Indeed, our CFL (but not control participants) had a much broader K_v range indicating a variable ability to suppress VOR during smooth pursuit. Further, this outcome is not surprising given the variability we see in smooth pursuit gains in CFL during head-restrained pursuit (17, 39) and the similarity in neural control of smooth pursuit and VOR cancellation (40).

Finally, an alternative explanation for the variability in eye-head coordination we saw in our CFL participants is an altered VOR. However, we did not find VOR deficits in a separate cohort of participants with maculopathy (Figure 5). These findings are consistent with prior literature (16) for VOR responses to self-generated head movements for near viewing (0.57 m) and suggest similarly intact vestibular function between the two groups. Thus, any differences we observe in the K_v estimates are unlikely due to functional differences in VOR between groups.

It is not clear if the lack of VOR cancellation we observe is due to central visual field loss specifically or aging more generally. Prior work has shown that smooth pursuit under head-restrained (41) and unrestrained conditions, as well as the efficacy of VOR cancellation (42, 43) are all impaired in older adults. While we are not provided with the age range of the participants in the Ackerley and Barnes work, it is likely safe to assume that their median age is significantly lower than that in our sample. While we assumed that our controls would behave similarly to the participants in (24, 25) during control trials, having near perfect eye velocities and cancellation of head movements by the VOR, it would be interesting to assess this older cohort’s behavior in trials where the target is temporarily extinguished, and head movements may provide an enhancement of the pursuit response. Future work is needed to help differentiate between the effects of aging generally and age-related changes in visual function.

While the current study sheds light on eye-head coordination during smooth pursuit in CFL, we do not examine VOR cancellation (or adaptation) by itself in this population. While it is well established that individuals with CFL have pursuit deficits, there is very little research available on their ability to adapt their VOR. Demer et al. examined VOR plasticity in response to telescopic magnification in low vision participants and found that a 15-min exposure to magnification was sufficient for low vision participants to exhibit plasticity. However, the study included participants with a variety of vision conditions, and it is difficult to ascertain how those with CFL compared to the rest of the rather broad sample. We do know, however, that more than 50% of those with macular degeneration were not successful in adapting to the use of telescopic lenses (though not necessarily due to insufficient adaptation) (44). Additional work in CFL looking at VOR adaptation without gaze-shifting could help shed light on whether patients are able to sufficiently adapt their VOR gain to the target motion and under what conditions.

Another consideration is the variability in eye-head coordination strategies employed by our participants. We instructed our participants to pursue in a manner that was most natural to them to allow for each participant’s individual adaptation and to understand if some were more effective than others (reliance on head movements versus not). As expected, participants exhibited a wide range of eye-head movement strategies, with some using head movements on almost all the trials, and others relying much more on the eyes alone. Future studies could examine VOR cancellation during pursuit under several different instruction conditions that would compare participants’ natural behaviors versus specific pursuit conditions. Using a larger participant sample and under a range of pursuit instructions would also allow for the analysis of head-velocity matched subsets of participants (or CFL-control matched pairs) to examine whether eye-head coordination strategy might be related to VOR cancellation in either group.

Additionally, the limited sample size and large variability in K_v in the CFL group (Figure 3D) could put in question the conclusion that on average CFL participants did not cancel their VOR (K_v = −1). While it is similar to that in (24, 25) – seven in the current study versus 6 in the prior publications, and our initial sample size calculation based on prior work (25) suggested that seven participants was sufficient, the variability in our sample exceeded that reported previously (Table 2). Our CFL participants are more heterogeneous in terms of age, visual function, disease progression and adaptation to it. However, our findings are illuminating even with the limited statistical power. Patients with CFL are highly heterogeneous in their oculomotor strategies, and our results indicate that at least a subset of them do not sufficiently cancel their VOR. This finding on its own could point to important future directions for development of training and rehabilitation strategies that would target those individuals who are not able to appropriately leverage head movement to enhance smooth pursuit, or gaze shifting more broadly. Future work that would examine differences between individuals with CFL who do and do not cancel their VOR will be important to understand the source of eye-head coordination deficits and potential ways to address them in this subgroup.

VOR cancellation during smooth pursuit is essential in natural tasks. Understanding why adaptations are evident in some, but not all CFL participants will allow for a development of training strategies that may improve eye-head coordination in this population, allowing for head-based adaptations to address known deficits in target tracking. The lack of VOR cancellation in CFL suggest that these patients are not adapting to their changes in vision, but instead still use the same strategy as their age-matched healthy sighted peers who do not experience target disappearance during the trial. There could be several reasons for this behavior, including (1) a lack of awareness of the target disappearance in their scotoma (45) or an (2) inability to cancel their VOR due to a lack of oculomotor control or VOR cancellation with a non-foveal PRL. For the first, scotoma awareness training has been demonstrated to be effective in individuals with CFL (46) and those with simulated scotomas (47) for other oculomotor tasks.

In terms of the second possibility, the vestibular system’s plasticity and adaptability are some of its key hallmarks, with VOR adaptation being a prime example where VOR gain must change quickly in response to factors such as viewing distance (48) or optical correction (44). While the close links between smooth pursuit and VOR cancellation (28) may provide a pitfall in the trainability of this behavior in CFL, they may also suggest a need for a unified training paradigm for both behaviors. Research in cerebellar disease suggests that VOR cancellation and smooth pursuit may not be as tightly linked as previously thought (49). The complexity of these oculomotor links and heterogeneity of eye-head coordination deficits in CFL suggest the need for developing flexible rehabilitation strategies that would train those with CFL to use head movements more or less depending on how well they can coordinate eye and head movements (determined on individual basis). A consideration of age as a factor will also be important in training approach development.