Visually induced vertical vergence as a motion processing biomarker associated with postural instability

The present study explored visually induced vertical vergence (VIVV) as non-specific motion processing response. Healthy participants (7 male, mean age 28.57 ± 2.30; 9 female, mean age 27.67 ± 3.65) were exposed to optokinetic stimuli in an HTC VIVE virtual reality headset while VIVV, pupil-size, and postural sway was recorded. The methodology was shown to produce VIVV in the roll plane at 30 deg/s. Subsequent trials consisted of 40s optokinetic motion in yaw, pitch, and roll directions at 60 deg/s, and radial optic flow; optokinetic directions were inverted after 20s of motion. Median VIVV amplitude changes were normalized to the clockwise roll rotation, analysed, and correlated with changes in pupil-size and body sway. VIVV, pupil-size, and body sway were all affected by rotational changes. Post-hoc analyses showed significance in yaw and pitch planes. Radial optic flow also produced significant VIVV. VIVV responses were universally correlated with pupil-size and body sway. In conclusion, VIVV was expressed in all dimensions and may consequently serve as a visual motion processing biomarker. Failing to support binocularity while responding to optokinetic directionality, VIVV likely reflects an optomotor response associated with increased postural instability and stress, similar to a dorsal light reflex.


Introduction
While the amount of vertical vergence varies with horizontal eye position, which changes as we shift focus from objects that are near to us to those that are further away and vice versa, we humans lack voluntary command over this reflexive adaptation (Schor et al., 2002).The eyes regularly adjust their relative vertical positioning in response to each head tilt, as a head rotation in the roll plane will raise the ipsilateral eye over the contralateral due to input from the vestibular utricles in an effort to aid binocularity (Brandt and Dieterich 1991;Schor et al., 2002).When vertical vergence occurs at rest, it may be indicative of central vestibular damage (Halmagyi et al., 1991), or caused by exposure to visual disparities (Van Rijn and Collewijn 1994).
Vertical vergence can also be seen in response to visual motion, and is intrinsically linked to visually induced ocular torsion (Guyton 1988).However, we have in a series of studies shown that the relationship between visually induced vertical vergence (VIVV) and ocular torsion shifts depending on the nature of the visual scene, with information density favouring torsion (Wibble and Pansell 2019;Wibble et al., 2020aWibble et al., , 2020bWibble et al., , 2020c) ) and stimulus acceleration enhancing the vergence response (Wibble et al., 2020a(Wibble et al., , 2020b(Wibble et al., , 2020c)).Contrary to the hypothesis that visually induced vertical vergence serves to fuse the visual field we have recently shown that VIVV is depressed by binocular input (Wibble and Pansell 2020).VIVV, despite its physiological purpose, may consequently inhibit binocularity and we therefore hypothesised that VIVV may reflect an optokinetic activation of the vestibular system.The optokinetic reflex (OKR), which allows the eyes to follow a moving visual scene, is intrinsically linked with the vestibulo-ocular reflex (VOR), sharing several key subcortical structures including the vestibular nuclei where vertical vergence is hypothesised to originate (Wibble et al., 2022).
A concrete mechanism for how a visual activation of the vestibular system may produce vertical vergence has been put forward.It is known that patients with amblyopia may grow up to develop dissociated vertical divergence (DVD), where one eye is elevated above the other (Brodsky 1999).The mechanism for this phenomenon is believed to stem from a vestigial dorsal light reflex (DLR) (Brodsky 1999, Brodsky 2002).The DLR is common in many species of insects and fish, and allows the animal to realign itself in relation to incoming sunlight (Duke-Elder 1958) through visual input activating righting reflexes via vestibular and reticulospinal afferents (Ullén et al., 1996;Ullén et al., 1997).It is tempting to suggest that the same principle may hold true for VIVV.This hypothesis may be further supported by the fact that unlike other optokinetic responses VIVV is sensitive to motion accelerations, a typically vestibular motion parameter (Wibble et al., 2020a(Wibble et al., , 2020b(Wibble et al., , 2020c)).Clinically, patients with vestibular symptoms caused by visually induced dizziness have been shown to exhibit enhanced VIVV responses (Wibble et al., 2023).
Given its association with both visual and vestibular data we hypothesise that VIVV may represent a general sensorimotor biomarker for visual motion processing, possibly serving as a predictor for postural instability and general stress.While it is well-established that VIVV can be triggered by visual rotations in the roll plane, the response should, in such a scenario, also be observable in response to other directions of optokinetic stimulation.There is to our knowledge no study investigating if VIVV may be observed during optokinetic stimulations beyond roll plane rotations.
The present study subsequently aims to explore if visually induced vertical vergence may be triggered by directional changes in optokinetic rotations beyond roll plane rotations.This will be carried out by exposing healthy controls to optokinetic roll, yaw, and pitch rotations, as well as radial optic flow.We furthermore hypothesise that VIVV may correlate with other vestibular responses such as postural sway and stress, as indicated by centre-of-pressure displacements and pupil-size.Altogether, the present study will explore VIVV as a biomarker for visual motion processing and postural control.

Participants
Sixteen healthy participants (7 male, mean age 28.57± 2.30; 9 female, mean age 27.67 ± 3.65) were recruited through university channels.Exclusion criteria included any medical history suggesting neurological or vestibular injuries, recurring non-vestibular sensations of dizziness or vertigo, or having started any new pharmacological treatment in the previous three months or taking any drug affecting the central nervous system.All participants went through a series of testing to ascertain that they had healthy visual, proprioceptive, and vestibular function in balance control.All individuals included in the study exhibited visual acuity within the normal range or had their vision corrected to meet the minimum standard of ≥ 1.0 using the logarithm of the minimum angle of resolution (logMAR chart).Their stereoscopic vision, assessed using the Lang II stereo test, was found to be at least 200″ of arc.Participants also demonstrated normal eye motility.To evaluate their vestibular function, a head impulse test was conducted in the directions of all three semi-circular canals, which showed no refixation saccade.Proprioceptive control over balance was assessed using a Romberg's test on a soft pillow.In addition, the eye-tracking protocol applied during the trials was used to determine if any latent nystagmus was present.No participant needed to be excluded based on these testing procedures.The study was conducted in accordance with the Declaration of Helsinki (Ethics Review Board approval: 2018/1768-31/1) and written consent was obtained from all participants.

Stimulation protocols
The present study evaluated VIVV, pupil-size, and centre-of-pressure changes.All participants were exposed to a series of optokinetic stimulations.In order to assess the capacity of the methodology to adequately record VIVV a stimulation of optokinetic roll at 30 deg/s was issued.This was expected to produce recordable VIVV, as indicated by previous research (Wibble and Pansell 2019, Wibble et al., 2020a, 2020b, 2020c, Wibble and Pansell 2020).In order to stress the optokinetic system, all subsequent rotational stimulations were presented at a velocity of 60 deg/s.This was done in roll, yaw, and pitch planes, allowing multivariate analyses.In addition, a radial optic flow stimulation was issued, as the optokinetic direction was expected to produce significant vection motion sensations, i.e., vection (Bubka et al., 2008).The velocity of the radial optic flow was set to 30 in-programme units/s.The optokinetic drum represents a standard testing methodology in eye movement evaluations.By placing tests subjects within the drum, the present study was able to offer a greater optokinetic effect compared to standard clinical tests, albeit in a non-ecological, laboratory setting.
The order of presentation, i.e. roll, yaw, pitch, or radial optic flow stimulations (supplementary videos 1-5), was distributed through stratified randomization between participants.Stimulations were presented in the HTC Vive (HTC, Taoyuan, Taiwan).The HTC Vive is a virtual reality (VR) headset capable of full-field visual stimulations (110 field-of-view, resolution 2880x1700, 90 Hz refresh rate).The VR was also capable of tracking the wearer's head position and carries an incorporated eye-tracker (110 degrees trackable field-of-view, 0.5 degrees accuracy, 120 Hz recording rate).
The HTC Vive offered integration with Unity (Unity Technologies, Copenhagen, Denmark) in which each stimulation protocol was designed.Each optokinetic stimulation consisted of black and white squares of irregular dimensions in a check pattern presented in a vertical cylinder to simulate the inside of an optokinetic drum.A red dot (visual angle 1.7 degrees) was maintained at a head fixed position in the middle of the visual scene.Due to the layout of the cylinder, yaw and pitch rotations feature the check pattern approaching infinity in the X-and Yaxis respectively (see supplementary videos 4, 5 respectively).The visual pattern can be seen in the supplementary videos for all optokinetic protocols, and the respective code is available from the authors upon request.The scene was adjusted to the height of each participant, and was fixed to the central axis of the head to ensure that eyes were kept horizontally parallel.The VR was individually calibrated using the builtin 5-point calibration protocol.Each optokinetic stimulation started with the visual scene being at rest for 20 s.The optokinetic stimulation was then initiated in one direction for 20 s before suddenly switching directions.The stimulation continued for another 20 s before being stopped, leaving a static visual scene for 20 s after which the trial was ended.The order of direction within each stimulation trial was distributed between participants through stratified randomization.No subject reported a breakdown of binocularity during the trials, and the visual elements of the stimuli were considered adequately sized for producing VIVV without introducing confounding diplopia.Participants were instructed to stand on a Wii Balance Board (WBB; Nintendo, Kyoto, Japan) during the entirety of each stimulation.Participants were instructed to stand with feet at pre-determined positions on the WBB and to move as little as possible during the entire protocol.This was done to ensure that all four scales of the WBB were properly engaged and that detected movements could be considered involuntary due to the optokinetic stimulation.

Eye-tracking and postural analyses
The present study used the Tobii eye tracker integrated with the HTC Vive (Tobii, Stockholm, Sweden; see specifications above) and its associated Software Development Kit (SDK) HTC SRanipal SDK.The SDK offered output data in the form of gaze direction, collated according to a Cartesian coordinate system, and pupil-size, which were implemented as outcome variables in the present study.While pupil-size was retrieved in millimetres, the angular displacements of the eyes needed to be calculated using the formula (Imaoka et al., 2020): The vergence was subsequently calculated by subtracting the vertical angular position of the right eye from the position of the left eye, yielding the relative vertical vergence in degrees.While orbital constraints of the eyes means that vertical divergence is intrinsically linked with ocular torsion, and one may expect the eye movements to accompany one another albeit expressing different velocities and amplitudes based on the content of the visual stimulus (Wibble and Pansell 2019, Wibble et al., 2020a, 2020b, 2020c), the HTC Vive does not allow for quantifying ocular torsion and we were therefore unable to evaluate its relation to VIVV in the present investigation.Roll plane stimulations are nevertheless known to produce ocular torsion, which will lead to an associated vertical divergence of the two eyes.A roll stimulation longer than a few seconds will produce torsional nystagmus, quick resetting eyes bringing the eyes towards their original position (Howard and Templeton 1964).This will naturally also influence the vertical divergence, as the response is linked to the torsional eye position.This effect will however be seen equally across all time periods and reflect the general direction of the eyes.As VIVV will be influenced by other factors than torsion (Wibble and Pansell 2019, Wibble et al., 2020a, 2020b, 2020c, Wibble and Pansell 2020), the vergence reflex may still be considered its own distinct eye movement response, and the torsional influence is expected to affect it equally across all trials without inserting confounding elements.This would have been further alleviated as the eye-tracker traces the visual-rather than the optic axis, mitigating the secondary vergence due to torsional nystagmus.While the gold standard of eye-tracking remains the scleral search coils, video-eye-tracking can offer sufficiently high fidelity to adequately measure slow-phase eye movements, as recorded in the present study.It is important to note that the present study presents eye positions from the perspective of the viewer's gaze-direction as indicated by the pupil and can therefore not form any conclusions regarding the three-dimensional dynamics of the orbit.
The present study investigated if rotational and radial optokinetic stimulations may cause visually induced vertical divergence.For roll rotations, the VIVV is expected to adhere to a strict pattern where the VIVV amplitude will follow the direction of the roll.However, as the other trials were not expected to produce any VIVV, there was no precedence for which eye would raise above the other in the vergence response.This investigation therefore outlines the presence of changes in VIVV direction, caused by inverted optokinetic motion.For example, a yaw rotation may produce positive or negative VIVV for two different subjects when exposed to the same stimulation.If that VIVV changed direction with the inverted optokinetic rotation, that was seen as a significant response (i.e., a negative VIVV amplitude becoming positive or vice versa).In these scenarios, all negative VIVV amplitudes were placed as the first response (1st direction), and all positives as the second response (2nd direction).This allowed the data to be normalized to the clockwise direction roll, as it corresponds to such an eye movement response.If there was no change in VIVV amplitude, no normalization was carried out.There was also no normalization for roll stimuli, as these were expected to yield a distinctly directional VIVV.This means that the analysed values are not representative of how yaw, pitch, or radial optokinetic stimuli produce directional VIVV, and the eyemovement data was normalized to explore how VIVV responds to changes in optokinetic direction during the trials.The re-organization of the data was carried out on a trial-by-trial basis, and out of the sixteen subjects it was done two times for the yaw condition, and three times for pitch and radial flow conditions respectively.
Pupil-size was used as a proxy for determining the stress level of the viewer, with increased pupil-size indicating a heightened autonomic response to the stimulus.Centre-of-pressure (CoP) was used to evaluate postural sway, and was measured using a Wii balance board.The WBB employed a dynamic sampling frequency averaging 99.1 Hz.Using its four corner scales the WBB was used to quantify the centre-of-pressure using a previously established in-house software where the data stream could be translated into a two-dimensional grid of coordinates (Wibble et al., 2020a(Wibble et al., , 2020b(Wibble et al., , 2020c)).This horizontal-by-vertical grid allowed us to calculate postural displacement through fitting collected data points with a 95 % confidence-interval ellipse and calculating its area.CoP area was adopted as a reflection of postural sway as it allowed for evaluating the magnitude of postural sway during any given period against the scalar value of the VIVV.While analysing the fractal dynamics of each sway would have yielded additional information concerning the nuances of the postural instability, this was deemed to be outside the scope of the present study as the main purpose for collecting postural data was to test its correlation with VIVV, evaluating if a greater divergence of the eyes could predict poorer balance.As analyses were subsequently carried out with regards to absolute values, the directions of the variables were deemed redundant for testing the hypothesis.Furthermore, by combining both medial-lateral and anterior-posterior sway the total CoP area also included oblique movements which otherwise may have been diminished in the statistical comparison.
Data retrieved for statistical analyses included CoP area, the median pupil diameter in millimetres, and the median vertical vergence amplitude during each time period; median values were chosen due to the uneven distribution of the oculomotor and postural responses between participants.Data analysis on VIVV parameters were carried out on smoothed data sets (Adjacent-Averaging, 100 Points of Window).The effects of the smoothing can be seen in Fig. 1, presenting the eye movement-, pupillary, and postural response to the roll stimulation.The figure also exemplifies the distribution of blinks and their effect on the eye movement response.As can be seen in the figure, these effects were mild and evenly distributed between the time periods and were therefore considered unlikely to introduce blink induced confounders.

Statistical analysis
The present study explored physiological responses to changes in optokinetic conditions.For this reason, statistical analyses were carried out on the differences between viewing conditions for each variable.For example, the VIVV amplitude (i.e., average eye position) for the initial 20 s were set as the baseline value and consequently set as 0. The baseline value was subtracted from the amplitude gathered during the following 10 s, yielding either a positive or a negative value that represented the amplitude shift.This value was in turn subtracted from the amplitude collected during the following 10 s, during which the optokinetic stimulation had shifted direction.The same procedure was followed for VIVV, pupil-size, and CoP.This allowed us to analyse and present changes for each variable over time.These changes were also used to establish a correlation matrix, investigating how well each parameter could be used to predict the response pattern.These correlations were carried out using absolute values, as neither pupil-size nor sway area could present negative values.However, as irregularities in the iris may lead to asymmetrical pupillary shifts within the same participant and lead to the eye-tracker recording false VIVV (Walsh andCharman 1988, Wyatt 1995), a full correlation analysis was performed on the raw data as well; the absence of a correlation between these responses and the VIVV amplitude was regarded as indicative that no false eye movements had been recorded.
As data did not meet criteria for normal distribution a Generalized Linear Model (GLM) with a multinomial distribution and cumulative logic link function was implemented to assess how each variable reacted to changes in optokinetic direction over time.An initial investigation was carried out with all VIVV, pupil-size, and CoP responses collated into separate models, using optokinetic direction as a co-variate and time (baseline, first motion, second motion, terminal rest) as a withinsubject factor.This was done to assess the data from a broad perspective.In addition, each variable was separated into secondary GLM analyses evaluating the effects of time on each variable and optokinetic condition.This was done to evaluate the specific effects of visual motion on each variable.Initial analyses were carried out maintaining data points as vectors.As optokinetic motion beyond roll-plane rotations were not expected to produce any binocular benefit for VIVV, an additional GLM was employed where the oculomotor amplitudes were normalized in terms of which direction the vertical vergence was first observed (see results).This allowed us to evaluate if VIVV responded to shifts in optokinetic directions, as outlined the previous segment relating to eye-tracking analysis.The statistical model compares variables across all time periods of each trial, indicating if any significant changes have occurred over time rather than between any two adjacent periods.This approach takes into account baseline values, both first and second directional values, and the responses observed once the stimulus returned to rest.All analyses were done with alpha set to 0.05.

Results
The present study investigated Visually Induced Vertical Vergence, pupil-size, and postural sway responses to a series of optokinetic stimulations in the roll, yaw, and pitch rotational planes, as well as while viewing a radial optic flow stimulation, and how the direction of the stimulus affected these reflexes.As vertical vergence carries no fusional purpose for yaw, pitch, and radial optic flow, the directional changes for these conditions were normalized to the clockwise direction of roll.It is consequently important to note that the results presented in this manuscript outlines how VIVV responds to changes in optokinetic directions.Pupil-size and postural sway were not normalized as they are not associated with any particular response pattern for any optokinetic motion.Extreme outliers were identified using a Grubb's test set at the 95 % confidence interval.Values beyond this value were re-evaluated for possible gaze instability.In total, two traces were excluded for roll, and six traces for intensified roll, leaving 62 and 58 traces on which analyses were carried out respectively.No outliers were identified for pitch, yaw, or radial optic flow stimulations, leaving 64 traces for the statistical protocol.No data were removed from pupil-size or postural sway values as it could not be ruled out that deviating responses were triggered by the visual stimulation.To account for non-normal distributed data and missing data points, a Generalized Linear Model was implemented.
The first step was to ascertain that the methodology allowed for detecting VIVV during roll plane stimulations.This was done during the low-intensity roll protocol, and a significant VIVV response over time was identified (X 2 (3, N=62) = 11.019,p = 0.012)).Similarly, the postural sway was significantly affected over time (X 2 (3, N=64) = 24.080,p < 0.001)), while pupil-size remained unaffected.This meant that both VIVV and the balance responses were responsive to changes in motion direction (Fig. 2).We then performed the same type of analysis for responses expressed during the radial optic flow protocol, known to produce significant vection (Bubka et al., 2008).The radial optic flow stimulation yielded significant VIVV, pupil-size, and postural sway responses over time (X 2 (3, N=64) = 12.314, p = 0.006); X 2 (3, N=64) = 10.101,p = 0.018); X 2 (3, N=64) = 10.959,p = 0.012), respectively; Fig. 2).representative traces for VIVV evoked by optokinetic roll rotations and radial optic flow can be seen in Fig. 3, illustrating amplitudal trends based on the direction of the stimulation.

Fig. 3.
Representative traces of the visually induced vertical vergence (VIVV) response in response to optokinetic rotations in the roll plane (30 deg/s), and as induced by radial optic flow.Trace data has been smoothened through adjacent-averaging at 400 points of window in relation to the analysed data.The direction has been normalized to clockwise roll meaning that negative amplitudes are presented first.Graphical data has been adjusted horizontally to better allow comparisons between optokinetic patterns.This means that the baseline has been adjusted so that it will roughly align with the graphical meridian at the change of the optokinetic direction at 40 s.The first dashed vertical line signifies the start of the optokinetic stimulation, the second line signifies the inverting of its direction, and the third line signifies the end of the stimulation.The horizontal dashed line serves as a reference line for VIVV amplitudes, here drawn at zero.
In order to evaluate if the VIVV responses could be predictive of the balance responses in the shape of pupil-size and postural sway, correlation analyses were performed.These showed significant and moderate correlations between VIVV and pupil-size changes in all dimensions (roll (r = 0.550, p < 0.001); yaw (r = 0.572, p < 0.001), pitch (r = 0.529, p < 0.001), radial optic flow (r = 0.623, p < 0.001), and intensified roll (r = 0.592, p < 0.001)).The same was observed for VIVV and postural sway (roll (r = 0.666, p < 0.001), yaw (r = 0.509, p < 0.001), pitch (r = 0.544, p < 0.001), radial optic flow (r = 0.506, p < 0.001), and intensified roll (r = 0.545, p < 0.001)).This means that VIVV was moderately predictive of stress and balance in all dimensions.A correlation analysis including pupil-size and VIVV in vectors was performed to preclude the inclusion of false VIVV due to irregular pupillary oscillations.No significant correlation was observed for pitch, radial optic flow, or any roll condition, and the significant correlation identified in the yaw plane may be viewed as negligible (r = 0.27, p = 0.032).This meant that the VIVV recorded was not due to false pupillary displacements.

Discussion
The present study set out to investigate how visually induced vertical vergence responds to different optokinetic motion directions, thereby elucidating its relationship to visual motion.Results showed that VIVV was responsive to changes in optokinetic directions in all tested visual motion patterns.This illustrates that VIVV reflects a sensorimotor visual motion processing reflex not necessarily purposed to maintaining binocularity.Its comprehensive correlation with pupillary size and postural sway further indicates that the response may be viewed as a visually induced righting response similar to a dorsal light reflex.
As shown in Fig. 4, VIVV, pupil-size, and postural sway were all affected by changes in optokinetic direction, and while VIVV have consistently been shown to generally quite small amplitudes (Wibble and Pansell 2019, Wibble et al., 2020a, 2020b, 2020c, Wibble and Pansell 2020, Frattini and Wibble 2021, Wibble et al., 2023), the present study found that measuring the median value over the course of 20 s was sufficient to ascertain its presence and detect directional changes when using the present VR setup.This established that VIVV, as well as the two balance variables, may be reliably provoked by visual stimulations,  Trace data has been smoothened through adjacent-averaging at 400 points of window in relation to the analysed data.The direction has been normalized to clockwise roll meaning that negative amplitudes are presented first.Graphical data has been adjusted horizontally to better allow comparisons between optokinetic patterns.This means that the baseline has been adjusted so that it will roughly align with the graphical meridian at the change of the optokinetic direction at 40 s.The first dashed vertical line signifies the start of the optokinetic stimulation, the second line signifies the inverting of its direction, and the third line signifies the end of the stimulation.The horizontal dashed line serves as a reference line for VIVV amplitudes, here drawn at zero.allowing more nuanced analysis of how the responses differ between optokinetic patterns.One may note that the experimental setup implemented in the present study fails to meet with the conditions an individual is expected to encounter during natural circumstances.Mobile eye-and headtracking is available (Franchak et al., 2021), and an improved methodology could aim to test how VIVV responds to motion in the natural environment while also assessing postural stability, for example through accelerometers on hip, ankle, or trunk locations.The optokinetic drum employed in this study is indisputably an artificial testing protocol, albeit its motion presented in a more ecological setting in the VR.While the motion content itself is highly artificial, lacking real-life counterpart, the visual data serves to replicate real-life scenario in terms of activating neural regions processing visual field sizes (Schor and Narayan 1981) and velocities (Cohen et al., 1977).The protocol implemented in the present study therefore serves to activate neural ocular motor pathways reflecting a physiological motion processing response.
Beyond the stimulation parameters, eye movement patterns also differ depending on the task carried out by the subject, with a searching task may generate more visual spread compared to simply walking (Franchak et al., 2021).Much like the optokinetic drum, the use of a visual fixation point is a highly experimental condition that rarely occurs in natural viewing.The procedure is commonly used in experimental designs to reduce unwanted eye movements, and by making it head-fixed the mechanism of visual suppression was expected to mitigate confounders caused by head movements and the VOR.However, it also introduced limitations in interpreting the naturalistic performance of the observed eye movements.The fixation point can also be expected to influence the degree of postural sway.Near targets tend to be associated with less postural instability compared to distant viewing, and dedicated visual inspection tasks produce greater sway than searching tasks (Stoffregen et al., 2000).One may therefore expect that the fixation point, and the instructions to stand as still as possible likely influenced participant's postural sway, and that the highly experimental conditions further enhanced this effect.A more natural instruction could therefore have been to ask subjects to remain as stable as possible, which may be considered more physiologically beneficial.Additionally, postural sway was reduced to only one variable in the present setup, precluding fractal dynamics with the aim to focus on magnitudes of instability for correlation analyses.Future studies could therefore benefit from focusing on evaluating these responses in a more natural setting.Altogether, the methodology used in the present study may be considered commonplace in visual-and postural research, as it allows for greater control of testing conditions while reducing unwanted noise despite its lack of ecological counterparts.The results of the present study may therefore be viewed as physiologically plausible, although the affect on human vision and postural control deserves further investigation.
The first step in the present study was to determine that the roll plane stimulation could produce detectable eye-movement responses.As seen in Fig. 2, this was established for the 30 deg/s rotation.However, the roll stimulus carried out at 60 deg/s did not lead to any significant changes over time.It is noteworthy that this trial led to a considerable range in the standard error.The OKR may fluctuate greatly in mean and median amplitudes at increased motion velocities as the resetting quick-phase, which moves the eye in the opposite direction to the pursuing slowphase, may pass its initial baseline and effectively generate negative oculomotor gain relative to the visual scene (Wibble et al., 2020a(Wibble et al., , 2020b(Wibble et al., , 2020c) ) or the subject's head position (Wibble et al., 2020a(Wibble et al., , 2020b(Wibble et al., , 2020c)).Greater quick-phase amplitudes relative to the slowphase counterparts will therefore lead to decreased median amplitudes, which may have contributed to the large standard error presented in the present study.This phenomenon can be seen in Fig. 5, where the negative trend of the second half of the optokinetic stimulation is offset by the resetting eye-movements.A manual examination of each slowphase may have precluded such an effect, albeit at the cost of being quite labour intensive.The present study therefore suggests that experimental protocols may benefit from lower optokinetic velocities to allow for automated amplitude calculations without the need of advanced analysis protocols.Concerning the assessment of the vergence magnitude, the HTC Vive operates according to a Cartesian coordinate system, retrieving positional data of the eye in three planes.The subsequent analysis of this data therefore supposes that the visual, and ocular motor, systems adhere to a Cartesian system.Eye positions may generally be described using one of five coordinate systems.Four of these implement a Cartesian model (Fick, Helmholtz, Harmes, and Hess) (Schor 2000), while one uses a Polar coordinate system (Listing) (Simonsz and Den Tonkelaar 1990).Listing's approach reduces the degrees of freedom from three to two, excluding torsional data to approximate the physiological position of the eye.From a neurophysiological perspective, both Cartesian and polar coordinate systems are used to decode visual information (Gallant et al., 1993).The HTC Vive is consequently well-suited for the present methodology.
It is important to note that this study normalized the direction of each VIVV for non-roll trials.Yaw, pitch, and radial optic flow stimulations were not expected to produce any distinct VIVV responses as there is no clear physiological reason for their induction as compared to roll plane rotations which contributes to increasing the fusional range (Schor et al., 2002).To investigate if these conditions resulted in any VIVV response, it was deemed that they would have to be sensitive to direction changes, indicating a reflexive responsiveness to the viewed motion.By normalizing the order at which negative or positive VIVV amplitude differences were recorded we were able to investigate if these changes corresponded to directional changes in the optokinetic pattern.
As seen in Fig. 3, this was the case for all optokinetic rotations, and as the same response pattern was evoked during radial optic flow stimulations (Fig. 2) these results suggest that VIVV may act as an oculomotor biomarker for visual motion processing.The re-organization of the data naturally introduces additional risks for type one errors to the analyses, although the limited number of such adjustments for each optokinetic stimulus mitigated this risk.Ultimately, the re-organization was carried out to investigate if VIVV responds to changes in optokinetic directions and cannot be used to infer if VIVV can be expected to have a directional preference in response to visual motion beyond roll-plane rotations.An additional possible limitation when analysing small amplitude eye movements is false displacements due to asymmetric pupillary contractions (Walsh andCharman 1988, Wyatt 1995).However, the present findings cannot be explained by such a confounder, as any such effect would have been increasingly manifest with increased pupillary sizes; as can be seen in Fig. 4, the pupil size increases with a change in optokinetic direction while the VIVV instead decreases.Furthermore, the correlation analysis implementing directional data for the VIVV showed no relevant level of correlation with the pupil size.
These results indicate that VIVV can be observed to several dimensions of optokinetic motion without serving any clear physiological purpose, as it fails to preserve binocularity.This further supports the hypothesis that VIVV relies primarily on non-visual feedback pathways.Not only does VIVV not aid in increasing the fusional range, but it may even be viewed as detrimental to binocular vision by creating visual disparity in the visual field; this effect is however likely negligible in healthy individuals considering the modest amplitudes observed.The present findings, incongruous with the physiological framework in which VIVV may be expected, consequently invites speculation on what mechanisms this oculomotor response represents.While vertical vergence has traditionally been believed to be triggered by head rotations (Brandt and Dieterich 1991), when viewing visual disparities (Van Rijn and Collewijn 1994), or as a necessary by-product of ocular torsion (Guyton 1988), we have in a series of studies shown that VIVV may be triggered by viewing a rotating visual scene, and that this response is distinctly different from that of visually induced ocular torsion; the vergence velocity, relative to torsion, is increased with heightened optokinetic accelerations, and decreased in the presence of additional visual information density, and unlike ocular torsion VIVV is supressed during binocularity (Wibble and Pansell 2019, Wibble et al., 2020a, 2020b, 2020c).This shift in ratio means that vertical vergence cannot be considered a by-product, or artefact, of the torsional response.Moreover, VIVV appears to be subdued by visual input (Wibble and Pansell 2020) and is readily enhanced in individuals suffering from visually induced dizziness in post-concussion syndrome (Wibble et al., 2023).Considering the associating with vestibular activity and symptoms one may hypothesise that VIVV reflects a visual activation of primarily vestibular pathways (Wibble and Pansell 2020).It has previously been put forward that Dissociate Vertical Divergence (DVD), more commonly called Dissociate Vertical Deviation (Guyton 2000), manifests as a vertical divergence of the eyes in individuals who grew up with amblyopia, serves as a remainder of an otherwise vestigial dorsal light reflex (DLR) (Brodsky 1999).The DLR allows certain animals, primarily fish and insects, to reflexively reorient in relation to the sun rather than gravity, albeit through visual input rather than vestibular signaling (Hughes 1966, Stavn 1970).The DLR relies on subcortical pathways relaying visual input to brainstem areas important for postural control, notably through its enhancement of vestibular signals (Ullén et al., 1996).As VIVV was broadly correlated with postural instability and increased stress we therefore put forward that VIVV may represent an ocular motor manifestation of this vestigial DLR, and a visual activation of the vestibular network in healthy adults.It is important to note that this hypothesis is speculative, and one could also suggest that VIVV could represent a visual activation of the vestibular system without necessarily representing any distinct vestigial reflex, serving an entirely different ocular motor function.The visual system's capacity for mitigating vertical disparities is lower than those presented in the horizontal plane (Hess 2018), and a reflexive ocular motor response in the vertical dimension may serve to control for this fact, although the present results would suggest that such a function appears quite poorly calibrated.One may also consider that the observed changes in VIVV, pupil-size, and sway area were all caused independently of one another despite the robust level of correlation, and future studies may do well to test different viewing conditions and fixation distances to test this hypothesis.One may also note that there was a general trend for median VIVV values to continue to increase even after the motion had stopped, as suggested by Figs. 2, 4, and 6.As indicated by the confidence interval, this was not the case for all participants.However, we hypothesise that the prevalence of this response was due to the velocity storage mechanism (VSM), through which motion information is build up in the vestibular system to prolong the neural capacity to detect movement to mitigate sensory adaptation effects (Raphan et al., 1979, Maioli 1988).This mechanism would prolong both the OKR and the subsequent decay of the eye position from the end of the optokinetic viewing back to baseline.As the VSM also relies on the intensity of the motion, it could also explain the fluctuations observed in the 60 deg/s roll stimulation.
The present study expanded on the relationship between VIVV and typical balance responses by highlighting its thorough correlation with changes in pupil-size and postural sway.These changes were also affected over time for several motion stimulations, likely reflecting the impact sudden directional changes may have on the viewer.One may note that only radial optic flow and the intensified roll stimulation yielded both changes in pupil-size and postural sway while also being the only trials not producing significant VIVV.From a neurophysiological perspective, ocular motor and postural responses originate in different parts of the vestibular nuclei (Cullen 2019), and different response patterns may therefore be expected depending on the nature of the motion.For example, postural sway was primarily in the anterolateral direction, and as such one may expect radial optic flow to have a greater impact on this balance response due to its optokinetic direction.Intensified roll may have produced similar responses due to its moving at a higher velocity.The correlation between eye movements and postural control may carry particular clinical utility when assessing visually induced dizziness or visual motion hypersensitivity (Wibble and Pansell 2023).While these findings elucidate how human physiological balance responses behave during various optokinetic motion conditions, further studies comparing these variables to a range of motion patterns and velocities are warranted to evaluate the clinical and physiological significance of these reflexes.
In conclusion, the present study showed that visually induced vertical vergence can be triggered by optokinetic motion in any of the three rotational planes, yaw, pitch, and roll, as well as radial optic flow.This VIVV response was quantifiable using a commercial VR and eye-tracker implementing a rudimentary analysis protocol.The vergence responses were correlated with typical balance responses in the form of stress and postural sway as indicated by changes in pupil-size and sway area respectively.Altogether these findings suggest that VIVV does not rely on visual feedback pathways, and likely reflect a visual activation of the vestibular network.This physiological purpose of this reflexive ocular motor response in humans remains unclear.Vertical vergence has been put forward as a human representation of the dorsal light reflex, and one may by extension hypothesise that VIVV could serve as a reflection of this basic response.While vertical vergence is a well-described eyemovement response, we put forward that VIVV may be considered a distinct ocular motor reflex due to its failing to support binocularity.While the present study was limited by not including any variables related to perception, and further studies are warranted to test its physiological benefit, we suggest that VIVV may hold utility as a biomarker for human motion-processing.

Fig. 1 .
Fig. 1.Representative traces of the visually induced vertical vergence (VIVV), the pupillary response, and the postural sway as indicated by the centre-of-pressure (CoP) for one subject exposed to the roll stimulation.The optokinetic start (at 20 s) and end (at 60 s) have been highlighting with vertical lines in each graph, as has the time of the directional change (at 40 s) of the visual motion.For VIVV, the raw unfiltered data can be seen in black, with the blinks represented by near instantaneous peaks.The red trace represents the smoothed trace using adjacent averaging at 100 points-of-window, and which was used for the analyses.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4 .
Fig. 4. The effects of motion direction on visually induced vertical vergence (VIVV), pupil-size, and postural sway over time and across all optokinetic trials (60 deg/ s).VIVV is given in amplitudes in degrees, pupil-size in millimetres, and postural sway as an indexed value of the centre-of-pressure (CoP).All values are given as the difference from on period to the previous one, yielding a value of zero for all baseline values.The subsequent time periods indicate the start of the optokinetic stimulation where the visual scene moves in one direction for twenty seconds before suddenly switching to the inverse second direction in which it continues for twenty seconds before coming to rest.Negative values indicate inversed VIVV.Pupil-size and postural sway only increased between optokinetic changes.Values are given as medians with the 95 % confidence interval.*p < 0.05; **p < 0.01; ***p < 0.001.

Fig. 5 .
Fig. 5. Representative traces of the visually induced vertical vergence (VIVV) response in response to optokinetic rotations in the roll, yaw, and pitch planes (60 deg/ s).Trace data has been smoothened through adjacent-averaging at 400 points of window in relation to the analysed data.The direction has been normalized to clockwise roll meaning that negative amplitudes are presented first.Graphical data has been adjusted horizontally to better allow comparisons between optokinetic patterns.This means that the baseline has been adjusted so that it will roughly align with the graphical meridian at the change of the optokinetic direction at 40 s.The first dashed vertical line signifies the start of the optokinetic stimulation, the second line signifies the inverting of its direction, and the third line signifies the end of the stimulation.The horizontal dashed line serves as a reference line for VIVV amplitudes, here drawn at zero.

Fig. 6 .
Fig. 6.The effects of motion direction on visually induced vertical vergence (VIVV) as recording during full-scene rotation in the roll, yaw, and pitch planes (60 deg/ s).All VIVV values are given as amplitudes in degrees.All values are given as the difference from on period to the previous one, yielding a value of zero for all baseline values.The subsequent time periods indicate the start of the optokinetic stimulation where the visual scene moves in one direction for twenty seconds before suddenly switching to the inverse second direction in which it continues for twenty seconds before coming to rest.Negative values indicate inversed VIVV.Pupil-size and postural sway only increased between optokinetic changes.Values are given as medians with the 95 % confidence interval.*p < 0.05; **p < 0.01; ***p < 0.001.