Diplopia

Diplopia, commonly called double vision, is the perception of two images of a single object. Clinicians separate two forms. Monocular diplopia persists when one eye is closed or covered, while binocular diplopia occurs only with both eyes open and goes away when either eye is covered. Misaligned eyes are a common cause of binocular diplopia, while monocular diplopia is usually refractive in nature.

In virtual reality (VR) and augmented reality (AR), diplopia is a design constraint. A binocular head-mounted display shows a separate image to each eye and depends on the visual system to fuse the two into one percept. Stereoscopic content with more binocular disparity than the eyes can fuse, vertical misalignment between the left and right display channels, and mismatches between the user's eyes and the headset optics can all produce double images, or the eye strain of working to avoid them. Apple tells Apple Vision Pro users to stop using the headset immediately if they notice blurred or double vision, and clinical researchers have built VR headset tests that measure the eye misalignment of patients with diplopia.

Reviewed 11 October 2026. Checked every cited claim against the Cleveland Clinic and AAO pages, the Peli 1999 chapter, the Gibaldi et al. 2025 paper, the cited journal abstracts and full texts, Apple's WWDC23 session and safety page, the Microsoft patent and the heise article. About review dates.

Definition and types

Cleveland Clinic defines diplopia as "the medical term for double vision or seeing double". According to Dictionary.com, the word is New Latin, dating to 1805-15, and is built from the combining form diplo-, meaning "double".

The first step in a clinical examination is to establish which form is present. In a 2022 article for the American Academy of Ophthalmology, Kimberly Gokoffski wrote that double vision from ocular misalignment is present only when both eyes are open, and that patients with binocular diplopia describe two completely separated, crisp images. Patients with monocular diplopia instead describe a crisp image with an overlapping shadowed image. Refractive diplopia often resolves when the patient looks through a pinhole or blinks.

Form	Present with one eye covered?	Typical description	Examples of causes
Monocular	Yes	A crisp image with an overlapping, shadowed image	Dry eye, refractive error, cataract, epiretinal membrane
Binocular	No	Two separate, crisp images	Ocular misalignment from cranial nerve palsies, myasthenia gravis, giant cell arteritis or decompensated strabismus such as intermittent exotropia

Gokoffski notes that binocular diplopia can be a warning sign because it may reflect an underlying intracranial process. Cleveland Clinic also describes binocular diplopia as usually the more serious form, and lists causes of double vision that include astigmatism, cataracts, dry eye, migraines, cranial nerve problems and head injuries.

Double vision is a common reason for medical visits. A study of US national survey data from 2003 to 2012 by Lindsey De Lott and colleagues, published in JAMA Ophthalmology in 2017, estimated 804,647 ambulatory visits and 49,790 emergency department visits for diplopia per year. Ophthalmologists conducted 70.4 percent of the ambulatory visits. About 16 percent of the emergency department visits ended in a diagnosis of stroke or transient ischemic attack, while none of the ten most frequent ambulatory diagnoses was life threatening.

Binocular fusion and physiological diplopia

Normal binocular vision depends on fusion, the merging of the images from the two eyes into one percept. An ophthalmology tutorial by Rahul Bhola for the University of Iowa's EyeRounds describes diplopia as "the hallmark of retinal disparity": it occurs when a single object point stimulates non-corresponding points on the two retinas. The horopter is the set of object points imaged on corresponding retinal points at a given fixation distance. Around it lies Panum's fusional area, a band within which disparate images are still seen as single. The tutorial gives its width as about 6-10 arc minutes near the fovea, widening to 30-40 arc minutes at 12 degrees from the fovea, and states that it is roughly 20 times wider for fuzzy, slowly moving stimuli than for sharp, rapidly moving ones. Outside Panum's area, physiological diplopia occurs.

The same tutorial separates diplopia from confusion. Confusion arises when corresponding retinal areas receive dissimilar images; diplopia arises when non-corresponding areas receive similar ones. The brain can suppress one eye's input to cope with both, and Bhola summarizes that suppression is "foveal in order to tackle confusion and extrafoveal in order to avoid diplopia". Fusion has a sensory part, which interprets two similar images as one, and a motor part, which aligns the eyes so that sensory fusion can be kept. The tutorial lists normal fusional vergence amplitudes of 35 to 40 prism diopters for horizontal convergence and 5 to 7 prism diopters for divergence, but only about 3 prism diopters for vertical vergence in each direction.

The fusion limit is not a fixed angle. In a 1984 paper in Vision Research, Clifton Schor, Ivan Wood and Jane Ogawa reported that sensory fusion, previously thought to have a maximum extent at the fovea of 1 degree, is at least 600 percent larger when stimulated by coarse (low spatial frequency) detail. They found a constant phase disparity limit of 90 degrees, and at low spatial frequencies the fusion limit equaled the upper disparity limit for stereoscopic depth perception, which they took to suggest a common mechanism for the two.

History

In a 2003 paper in Arabic Sciences and Philosophy, Dominique Raynaud traced the modern concept of retinal correspondence to Christiaan Huygens (1704) and to an experiment attributed to Christoph Scheiner (1619), and argued that the experiment had already been conceptualized by Ptolemy and by Ibn al-Haytham. According to Raynaud, Ibn al-Haytham's experiments established the notion of corresponding points and the cases of homonymous and crossed diplopia, and anticipated the later discovery of Panum's area.

Alignment between the two eyes' images became an engineering question with binocular instruments. Gibaldi and colleagues note that vertical image alignment was studied for hand-held binoculars and binocular microscopes (Jacobs, 1943) and for head-up displays in aircraft cockpits (Gold and Hyman, 1970; Gold, 1971). Eli Peli's 1999 review of head-mounted display optometry lists alignment tolerances recommended for binoculars in 1948 and in the US military handbook MIL-HDBK-141 (1962). In 1986 Self reviewed the tolerances available for binocular helmet-mounted displays and found them mostly in military reports, often stated without any account of how they had been derived. Gibaldi and colleagues called Self's review "seminal" and opened their 2025 study with its premise that "zero optical image differences and zero alignment errors are not possible with binocular devices".

Diplopia in VR and AR

Disparity and fusion limits in stereo content

A stereoscopic headset creates depth by offsetting objects laterally between the left and right images. Philip Turnbull and John Phillips note that, to prevent double vision as gaze moves between objects at different simulated depths, users make vergence eye movements that reduce the retinal disparity of the object they look at. Content outside the range of disparities that the eyes can fuse is seen double.

Peli's chapter estimated these ranges for head-mounted displays from clinical vergence norms (Morgan's norms). For a virtual screen at 1 m, the break point (loss of fusion, or double vision) fell at 6.8 degrees of uncrossed and 11.5 degrees of crossed disparity. Peli stressed that these are break values and not suitable for continuous viewing, and derived a smaller comfort zone of about 2 degrees uncrossed and 2.3 degrees crossed. He also noted that in the fovea Panum's area is about 15 arc minutes horizontally, so disparities of that size can be shown without triggering vergence, and that a disparity of up to 2 degrees shown too briefly for an eye movement (about 0.25 s) can still give a sense of depth even though no fusion takes place and a double image is seen. Peli described work by Nagata (1996) in which, for two of three subjects, large disparities at the fixated target were better tolerated when the surrounding images were blurred than when they were sharp, and suggested that blurring non-fixated detail could reduce strain on fusion in headsets.

The fixed focal distance of most headsets adds a second problem, the vergence-accommodation conflict. In a 2008 Journal of Vision study, David Hoffman, Ahna Girshick, Kurt Akeley and Martin Banks wrote that the uncoupling of vergence and accommodation required by 3D displays "frequently reduces one's ability to fuse the binocular stimulus" and causes discomfort and fatigue. Using a prototype display with nearly correct focus cues, they found that viewers identified stereoscopic stimuli faster and reported less fatigue. A follow-up study by Shibata, Kim, Hoffman and Banks in 2011 found that clinical measures of phoria and of the zone of clear single binocular vision predicted how susceptible a person was to discomfort with stereo displays.

Platform guidance addresses diplopia directly. In the session "Design considerations for vision and motion" at the Worldwide Developers Conference in 2023, Manda Paul and Herman Damveld of Apple said that viewers "might experience discomforts such as double vision or eye fatigue" when depth cues are missing, conflicting or misleading. They gave repeating patterns as an example: if each eye locks onto a different element of the pattern, the fused image is perceived at the wrong depth, and Apple recommended using smaller portions of the pattern or breaking it up "to avoid double vision". The session also told developers who render their own stereo content to keep the disparity correct for each eye, and to place content that requires long fixation, such as text for reading, farther than arm's length.

Display alignment tolerances

Each display channel of a binocular headset or pair of AR glasses must be aligned with the other. Peli wrote that vertical alignment is "probably the most important tolerance" for head-mounted displays, and noted that vertical phoria changes are considered more important clinically than lateral ones.

Source	Context	Vertical misalignment figure
ANSI Z80 spectacle standard (cited by Peli)	Spectacles, constant use	0.25 prism diopter (8.6 arcmin)
MIL-HDBK-141 (1962) and Johnson (1948) (cited by Peli)	Binoculars	17 and 34 arcmin
Peli (1999)	Recommendation for head-mounted displays	0.75 prism diopter
Self (1986), as summarized by Gibaldi et al.	Helmet-mounted display literature review	Tolerated values ranged from 3.4 to 34.5 arcmin across studies
Speranza and Wilcox (2002), as summarized by Gibaldi et al.	About 35 minutes of 3D film viewing	15 to 20 arcmin tolerated before discomfort
Tyler et al. (2012), as summarized by Gibaldi et al.	Stereoscopic images	No discomfort below 5 arcmin; moderate discomfort up to 12 arcmin
Gibaldi et al. (2025)	Simulated see-through AR, 26 participants	Median 15 arcmin (range 9-22) before stereopsis was impaired

According to Gibaldi and colleagues, Self attributed the wide spread of published values to two factors: scene content (with a complex background, tolerable differences were about ten times smaller than with a uniform background) and to the criterion used: comfort-based limits are tighter than limits based on fusion and diplopia, because the eyes can usually still fuse a vertical offset after it has become uncomfortable. Peli also cited a personal communication from Sony reporting that a vertical misalignment of peripheral targets as small as 0.25 degrees caused discomfort and that more than 0.5 degrees caused diplopia, while noting that only the peripheral targets were misaligned in that test. Because a headset can be knocked out of alignment, he suggested including vertical alignment targets so users could check it.

Unequal image sizes in the two eyes (aniseikonia) can also defeat fusion. Peli cited clinical literature in which aniseikonia of more than 5 percent is not compatible with binocular vision and usually results in diplopia or suppression. Noting that there were no data on the level of aniseikonia in head-mounted displays that causes discomfort or difficulty in fusion, he recommended limiting it to less than 1 percent based on clinical experience.

Gibaldi and colleagues, a group of Magic Leap researchers whose co-authors included Jenny Read of Newcastle University, studied binocular vertical misalignment in optical see-through AR. In such devices the real world forms aligned images on the retinas while the virtual content may be offset. They simulated this on a stereoscopic monitor with shutter glasses and found that the visual system compensates for the offset in equal parts through sensory fusion and through vertical vergence eye movements. Tolerance varied widely between participants, and the authors suggested per-user calibration and highly accurate display alignment. In VR or video passthrough, by contrast, a display offset applies to the whole field of view, which the authors compare to laboratory studies where vertical disparities of 1.7 to 2.8 degrees or more were tolerated before stereovision broke down or diplopia arose.

Software correction of the offset has also been patented. A Microsoft patent granted in March 2025, "Vertical misalignment correction in binocular display systems" (US 12,242,063 B2, filed March 2023), notes that users who cannot fuse misaligned images can experience "blurred or double vision (diplopia)". It describes detecting the misalignment with cameras or eye-tracking sensors and shifting the mapping of each rendered image up or down by display pixel rows, correcting more quickly when the user looks at a small object, for which the fusional range is smaller. In a 2025 article on the Meta Orion prototype glasses, heise reported that their optical elements must be aligned to within a tenth of the width of a human hair and that the torsion-resistant magnesium frame helps keep that alignment stable; according to Meta, the glasses also detect small frame deformations, such as expansion and contraction with temperature, and correct the optical alignment digitally.

Interpupillary distance and lens decentration

Headset optics are designed around a nominal eye separation. Peli explained that a user whose interpupillary distance is smaller than the distance between the headset's optical centers looks through the inner parts of both lenses, which acts like a pair of prisms in front of the eyes. Earlier authors had expected this prismatic effect to cause eyestrain and headaches, but in a study of 53 subjects Regan and Price (1996) found no such overall effect. Peli calculated that only the lens power beyond what is needed to place the virtual screen at infinity contributes to the prism. For a virtual screen at 2 m and a full 10 mm difference between the user's IPD and the optics, the effect is only 0.5 prism diopter, the tolerance permitted for spectacles under ANSI Z80.1-1972.

Peli also discussed headsets whose convergence demand does not match the accommodative demand set by the virtual image distance, a situation equivalent to placing a prism in front of the eye. If the mismatch is large enough, he wrote, vision on first looking through such a system may be blurred or double. For smaller offsets, vision may stay clear but continued use can cause eyestrain. After a few minutes the visual system starts to adapt by shifting its baseline vergence (prism adaptation); people with strong binocular vision readapt within minutes after removing the headset, but some users with less functional binocular systems may have blur, double vision or eyestrain.

Studies of headset users

In 1993 Mark Mon-Williams, John Wann and Simon Rushton examined 20 adults aged 19 to 29 before and after they wore a commercial head-mounted display for 10 minutes while cycling through a virtual world, with the twin screens set to the average interpupillary distance of the group. They reported clear signs of induced binocular stress in a number of the subjects. Reviewing this and other work in 1999, Peli wrote that there were no reports of double vision persisting after use of head-mounted displays, although reports of blur, eyestrain and headaches were abundant.

Later studies found mixed short-term effects. Turnbull and Phillips tested 19 young adults after 40-minute sessions in real and virtual environments on an Oculus Rift DK2 and found no evidence that the headset's optics harmed binocular status in the short term. In a 2021 randomized trial, Hyeon Jeong Yoon and colleagues had 58 volunteers aged 20 to 39 play games for 2 hours on a Samsung Gear VR (Innovator Edition) with a Galaxy S6 phone and, on a different day, on a smartphone. Near point of convergence, accommodation, exophoria and stereopsis changed significantly after VR use but not after smartphone use, and headache, dizziness and nausea were worse after VR. Self-reported double vision did not differ between the two conditions (median score 0 in both).

Safety documentation for current headsets names double vision among symptoms that should end a session. Apple's safety information for Apple Vision Pro says to stop using the device immediately on experiencing discomfort such as eyestrain, eye pain "or a change in vision such as blurred or double vision", and notes that people with an eye or vision condition may find that use aggravates their symptoms.

Measuring diplopia with VR headsets

Because a headset controls exactly what each eye sees, it can also serve as a test instrument. In 2017 N. Nesaratnam, P. Thomas and A. Vivian compared a VR headset test with the traditional Lees screen in three patients, with a fourth nerve palsy, comitant esotropia and restrictive thyroid eye disease. The patterns of deviation agreed, and the authors described the work as the first use of a VR headset to assess ocular misalignment.

A team at Goethe University Frankfurt later built a VR version of the Harms tangent screen test, which measures subjective eye deviation in patients with diplopia, on an Oculus Go with software written in Unity. In 85 adult patients, horizontal and vertical deviations agreed very well with the conventional test (intraclass correlations 0.93 to 0.98), and torsional deviations less well (0.79 to 0.93). The VR test took an average of 6 minutes, compared with 15 minutes for the conventional one, and did not require orthoptic training to administer.

See also

• Binocular vision
• Binocular disparity
• Stereopsis
• Vergence
• Vergence-accommodation conflict
• Interpupillary distance
• Visual fatigue
• Eye