Vergence
More actions
Vergence is the angle of a person's two eyes relative to each other, and the eye movement that changes that angle. When attention shifts between near and far objects, the eyes converge (rotate inward) or diverge (rotate toward parallel) so that the fixated object stays centered in both eyes and its image falls on the fovea of each.[1] The closer the target, the larger the vergence angle.[2]
In natural viewing, vergence is tied to accommodation, the focusing of the eye's lens: a change in one drives a change in the other.[3] Stereoscopic head-mounted displays break this link, because the eyes verge to the simulated depth of an object while focus stays fixed at the display's single image distance. The resulting vergence-accommodation conflict is a known cause of visual discomfort and fatigue in VR and AR, and measuring vergence with eye tracking is one way to drive varifocal displays that correct it.[4][5]
Definition
Moving the eyes between two points in space usually requires two kinds of movement at once. Conjunctive movements (version), typically saccades, move both eyes in the same direction and change where gaze points. Disjunctive movements (vergence) move the eyes in opposite directions and change the convergence angle between them. Caziot and Bremmer describe version as the average orientation of the two eyes and vergence as the convergence angle between them; in natural behavior the two almost always occur together, and the special cases that need only one of them (targets at the same distance in different directions, or in the same direction at different distances) are mostly found in laboratory experiments.[6]
Vision scientists distinguish the vergence distance, the distance of the fixated object, from the vergence response, the distance to the point where the two lines of sight actually intersect.[3] The two need not match exactly. For a given object depth there is a small range of vergence, Panum's fusional area, within which the two retinal images still fuse into one; a vergence error larger than about 15 to 30 arcminutes causes fusion to fail, leading to suppression of one eye's image or double vision (diplopia).[5]
Vergence eye movements
Vergence movements are generally slower than saccades. In a 2014 study of combined saccade and vergence movements in 12 healthy optometry students, Lang and colleagues found that the saccade component started later but ran faster, in both mean and peak velocity, than the vergence component. The same paper summarizes earlier work showing that vergence combined with a saccade is faster than isolated vergence, while a saccade combined with vergence is markedly slower than an isolated saccade.[7] When an unexpected stimulus appears at a new depth, Dunn's review of the literature gives a vergence latency of 150 to 200 ms and an accommodation latency of about 300 ms before the eyes respond. Vergence then settles fully on the new depth, while accommodation tends to lag and stop between the old and new focal distances.[5]
The start of the two movement types appears to be controlled separately. Caziot and Bremmer (2025) found that saccade latencies depended on the eccentricity and hemifield of a target but not on its disparity, while vergence latencies depended on the sign and size of the target's disparity but not on its eccentricity. They found no correlation between saccadic and vergence latencies and concluded that the two initiation mechanisms operate independently.[6]
Vergence problems are also a clinical topic. Convergence insufficiency is an increasingly common vision disorder, and the standard rehabilitation exercise, the push-up technique, has the patient maintain fusion while fixating the tip of a pencil that is moved along the midline toward the nose.[7]
Coupling with accommodation
Accommodation and vergence responses are normally coupled. Accommodative changes evoke vergence changes (accommodative vergence), and vergence changes evoke accommodative changes (vergence accommodation). In the real world the coupling helps because the focal distance and the vergence distance of a fixated object are almost always the same. One benefit is speed: accommodation is faster when both blur and binocular disparity signal the same change in distance, and vergence is likewise faster when blur and disparity agree than when only disparity changes.[3]
The range of vergence and accommodation that a viewer can reach at the same time without excessive error in either is called the zone of clear single binocular vision. A smaller region inside it, Percival's zone of comfort, holds the combinations a viewer can sustain without discomfort; Hoffman and colleagues describe it as about one third the width of the zone of clear single binocular vision.[3] Earlier work they cite reported changes in the cross-link ratios between accommodation and vergence (AC/A and CA/C) after prolonged viewing of a 3D display.[3]
Applications in VR and AR
Vergence-accommodation conflict
A conventional stereoscopic display presents all images on one surface. The vergence stimulus changes as the viewer looks around the simulated scene, but the focal distance stays fixed, so the viewer has to uncouple vergence from accommodation.[3] In a head-mounted display the eyes change their vergence angle to fixate objects at different stereoscopic depths, while the lenses of the eyes must stay accommodated to the fixed distance of the virtual image. Padmanaban and colleagues list double vision, reduced visual clarity, visual discomfort and fatigue as symptoms of this conflict, and note that the missing focus cue also weakens depth perception.[4]
Hoffman, Girshick, Akeley and Banks tested the effect directly with a display that could present correct or nearly correct focus cues. When focus cues matched the vergence stimulus, viewers needed less time to identify a stereoscopic stimulus, stereoacuity in a time-limited task improved, distortions in perceived depth were reduced, and viewers reported less fatigue and discomfort.[3] UploadVR summarized the same conflict for VR users in 2023: the eyes converge or diverge toward the virtual object "but can't actually focus (accommodate) to the virtual distance to the object", which causes eye strain and can make close virtual objects look blurry.[8]
Vergence-driven varifocal displays
A varifocal display has a single image plane whose focal distance can be moved to follow the user's focus, which requires the system to know where the user is focusing.[5] Measuring focus directly with an autorefractor or wavefront sensor needs complex on-axis optics, and Dunn notes that the fastest commercial devices sample fewer than 10 times per second. The indirect method uses the coupling of vergence and accommodation instead: a binocular eye tracker measures the direction of each eye, the intersection of the two gaze lines gives a 3D fixation point, and its distance is taken as the focal depth. Eye trackers commonly sample above 60 Hz and can reach 1000 Hz. Because the calculation depends on two separate gaze measurements, any error in vergence compounds into the depth estimate; only horizontal gaze error and interpupillary distance error affect it.[5]
Dunn calculated how accurate the eye tracker must be for a median adult. In VR, keeping the display within the zone of comfort (plus or minus 0.8 diopters of long-term error) requires gaze tracking error below 1.444 degrees. In optical see-through AR, where a virtual object can be compared with a real one at the same depth, the virtual focus should stay within the eye's depth of field (taken as plus or minus 0.3 diopters), which requires error below 0.541 degrees. He also describes a hybrid alternative that intersects the gaze direction with known scene geometry to find a candidate depth and checks it against the vergence-based fixation point, because scene depth alone can pick the wrong surface near depth edges.[5]
Padmanaban and colleagues (Stanford University and Dartmouth College) described a wearable prototype of this kind in 2017, built on the Samsung Gear VR platform with a binocular eye tracker and a stepper motor that moved the screen relative to the magnifying lenses in real time. The system used the eye tracker to find where the user was looking in the VR scene, calculated the distance of that location, and moved the virtual image to that distance in each frame so that vergence and accommodation distances could be matched.[4] Facebook's Half Dome prototype, shown in 2018, similarly used eye tracking to move the displays forward and back mechanically; UploadVR reported that it solved the vergence-accommodation conflict but that the mechanical approach raised serious reliability problems that made it unsuitable for shipping products.[8] Meta's Butterscotch Varifocal prototype, shown at SIGGRAPH 2023, combined the varifocal optics of Half Dome with near-retinal resolution; Meta said it supported accommodation from 25 cm to infinity.[9]
Vergence as a measure of depth
Because vergence angle changes with fixation distance, eye-tracked vergence can be used to study how deep a user perceives a virtual object to be. In a 2023 preprint, Arefin, Swan, Cohen Hoffing and Thurman measured vergence angle in 13 participants fixating targets at different depths, using real objects, AR objects and fully virtual objects. A larger vergence angle was associated with closer targets in all three conditions; once individual differences were controlled for, the difference between environments was statistically significant but small, and the authors report that vergence gave a more veridical depth estimate than participants' subjective depth judgments.[2]
See also
References
- ↑ "Eye Movements". Tobii XR Devzone. Tobii. https://developer.tobii.com/xr/learn/eye-behavior/eye-movements/. Retrieved 2026-09-27.
- ↑ 2.0 2.1 Mohammed Safayet Arefin, J. Edward Swan II, Russell Cohen Hoffing, Steven Thurman (2023-11-12). "Mapping Eye Vergence Angle to the Depth of Real and Virtual Objects as an Objective Measure of Depth Perception". arXiv preprint 2311.09242. https://arxiv.org/abs/2311.09242. Retrieved 2026-09-27.
- ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 David M. Hoffman, Ahna R. Girshick, Kurt Akeley, Martin S. Banks (2008-03-28). "Vergence-accommodation conflicts hinder visual performance and cause visual fatigue". Journal of Vision, vol. 8, no. 3, article 33. doi:10.1167/8.3.33. https://pmc.ncbi.nlm.nih.gov/articles/PMC2879326/. Retrieved 2026-09-27.
- ↑ 4.0 4.1 4.2 Nitish Padmanaban, Robert Konrad, Tal Stramer, Emily A. Cooper, Gordon Wetzstein (2017). "Optimizing virtual reality for all users through gaze-contingent and adaptive focus displays". Proceedings of the National Academy of Sciences, vol. 114, no. 9, pp. 2183-2188. doi:10.1073/pnas.1617251114. https://pmc.ncbi.nlm.nih.gov/articles/PMC5338519/. Retrieved 2026-09-27.
- ↑ 5.0 5.1 5.2 5.3 5.4 5.5 David Dunn (2019). "Required Accuracy of Gaze Tracking for Varifocal Displays". 2019 IEEE Conference on Virtual Reality and 3D User Interfaces (VR), pp. 1838-1842. University of North Carolina at Chapel Hill. https://telepresence.web.unc.edu/wp-content/uploads/sites/11620/2019/03/Dunn_2019_RequiredAccuracyOfGazeTrackingForVarifocalDisplays.pdf. Retrieved 2026-09-27.
- ↑ 6.0 6.1 Baptiste Caziot, Frank Bremmer (2025-08-11). "Dissociate triggering of conjunctive and disjunctive eye movements". Scientific Reports, vol. 15, article 29355. doi:10.1038/s41598-025-12031-5. https://www.nature.com/articles/s41598-025-12031-5. Retrieved 2026-09-27.
- ↑ 7.0 7.1 Alexandre Lang, Chrystal Gaertner, Elham Ghassemi, Qing Yang, Christophe Orssaud, Zoi Kapoula (2014-06-02). "Saccade-vergence properties remain more stable over short-time repetition under overlap than under gap task: a preliminary study". Frontiers in Human Neuroscience, vol. 8, article 372. doi:10.3389/fnhum.2014.00372. https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2014.00372/full. Retrieved 2026-09-27.
- ↑ 8.0 8.1 David Heaney (2023-06-05). "Meta To Show 'Near Retinal' Resolution Varifocal Headset Prototype At SIGGRAPH". UploadVR. https://www.uploadvr.com/meta-retinal-resolution-varifocal-prototype/. Retrieved 2026-09-27.
- ↑ David Heaney (2023-07-31). "Meta Will Demo Retinal Varifocal & 'Light Field' Passthrough Prototype Headsets Next Week". UploadVR. https://www.uploadvr.com/meta-showing-butterscotch-varifocal-retinal-flamera-light-field-passthrough/. Retrieved 2026-09-27.