Persistence of vision

Persistence of vision is a term with two related but distinct meanings, both of which matter for virtual reality and augmented reality displays. In its older, perceptual sense it is the long-standing idea that the eye briefly retains an image, or afterimage, on the retina after the light that produced it has gone, an effect historically invoked to explain why a rapid sequence of still pictures is seen as continuous motion. Modern vision science treats that historical explanation of motion perception as incomplete and largely superseded. In its second, engineering sense, which is the one that dominates discussion of head-mounted displays, "persistence" refers to how long each pixel stays illuminated within a single frame. This display persistence is central to image clarity in VR: long, or "full", persistence smears the image during head motion, so modern headsets use low-persistence displays that light each pixel only briefly per frame.

The historical perceptual concept

The perceptual notion has a long history. In 1768 Patrick d'Arcy recognised the effect in the luminous ring seen when a glowing object is whirled quickly, and using a purpose-built machine he measured a duration of about 0.13 seconds for a burning coal to be seen as a complete circle of light. The explanation most often quoted comes from Peter Mark Roget, who presented his findings on the appearance of moving wheel spokes viewed through apertures to the Royal Society on 9 December 1824. Roget proposed that "an impression made by a pencil of rays on the retina, if sufficiently vivid, will remain for a certain time after the cause has ceased." Joseph Plateau investigated the phenomenon further and built early animation devices around it.

For most of the twentieth century, persistence of vision was the textbook explanation for why optical toys such as the phenakistiscope and zoetrope, and later film projection, produce the impression of smooth movement: each frame was said to linger on the retina long enough to blend into the next. This is also where the common belief that the eye "sees at" a fixed number of frames per second comes from.

Why it is considered an incomplete explanation

Contemporary perception researchers regard retinal afterimage persistence as an inadequate account of how motion pictures work. As one summary of the literature puts it, "psychologists and physiologists have rejected the relevance of the theory of retinal persistence" as the mechanism behind apparent motion, even though the idea persists in many film texts. The core objection is that afterimages are static; simply overlaying one lingering still image on the next would produce a blur or a collage, not a clear sense of movement.

The better-accepted explanations come from the study of apparent (stroboscopic) motion. In 1912 Max Wertheimer reported that when two stationary figures are shown in rapid succession, observers perceive motion between them without seeing anything in the intervening space, an effect he named the phi phenomenon. A closely related illusion, beta movement, occurs when a series of still images shows an object in slightly shifted positions, creating the impression that a single object is moving. These are now understood as products of how the visual system processes change over time rather than as a passive smear on the retina; in short, the eye is not a camera and does not record discrete frames. Joseph and Barbara Anderson set out this critique for film studies in papers published in 1978 and 1998, arguing that persistence of vision is inadequate to explain cinematic motion.

Relation to flicker fusion

A separate phenomenon often confused with persistence of vision is the flicker fusion threshold (critical flicker frequency), the frequency at which a flickering light stops appearing to flicker and looks steady to an observer. For moving images this threshold is usually taken to be between about 48 and 60 Hz under typical viewing conditions, though it can be considerably higher, and it varies with brightness, the region of the retina stimulated, and the individual. Flicker fusion is what allows a strobed or interrupted light source to look continuous. It was traditionally lumped under the "persistence of vision" label, but vision science now treats flicker fusion, positive afterimages, and motion blur as distinct effects rather than a single mechanism. The distinction matters for VR because a low-persistence display, which deliberately strobes each pixel on and off, must run at or above the viewer's flicker fusion threshold to avoid visible flicker.

Display persistence in VR and AR

In the context of head-mounted displays, persistence has a precise technical meaning: it is the length of time, within each frame, that a given pixel actually emits light. A full-persistence display keeps each pixel lit for essentially the whole frame interval, the behaviour of an ordinary sample-and-hold panel such as a typical LCD or OLED monitor. A low-persistence display instead illuminates each pixel for only a small fraction of the frame and leaves it dark for the remainder.

This is a much bigger problem in VR than on a desktop monitor because the display is strapped to the head and the eyes are constantly tracking the virtual scene. When the eye smoothly follows a point in the world while the headset moves, a pixel that stays lit for the whole frame is dragged across the retina, smearing into a streak. The amount of smear is set not by the fraction of the frame the pixel is lit but by the absolute time it is lit multiplied by the speed of the eye relative to the display. Because that streak grows with persistence, a full-persistence headset looks blurry whenever the user turns to look around, even if the rendered image itself is perfectly sharp.

Persistence, judder, and refresh rate

The VR display artifact that persistence drives is judder, described on this wiki as a mixture of smearing and strobing produced by a head-mounted display with low refresh rate or high persistence. Reducing persistence attacks the smearing half of judder directly: if each pixel is lit only briefly, it does not have time to travel far across the retina, so the streak collapses to a sharp point. The trade-off is that the briefly flashed image can make strobing more visible, and the cure for strobing is a higher refresh rate.

Persistence and refresh rate are therefore linked but not the same thing. At a high enough refresh rate even a full-persistence display becomes sharp, because a full frame is itself very short: this wiki notes that truly eliminating the smearing of judder would take a refresh rate of about 1000 Hz, at which a full frame lasts only 1 ms. Such refresh rates are impractical with today's panels, so low persistence is used as a way to obtain the motion clarity of a much faster display without actually running at that frequency. UploadVR notes that low persistence "can deliver the same blur reduction as a 500 Hz panel would," which is why it became standard across modern headsets despite their comparatively modest refresh rates.

Origins in VR research

The application of low persistence to consumer VR grew out of work in the early 2010s by two groups in parallel: Valve, through the research of Michael Abrash, and Oculus, which reached it with input from the display-motion site Blur Busters. Abrash, who later became chief scientist at Oculus, laid out the problem in talks and writing on judder, explaining that judder is fundamentally an artifact of persistence: during each frame pixels remain illuminated for a considerable period, and the smear part of judder results from each lit pixel moving across the retina because of eye motion relative to the display. He also described the other half of judder, strobing, as the image jumping back to its correct position at the start of each displayed frame, and noted that lowering persistence to remove smear can leave residual strobing that only a higher frame rate fully resolves.

Oculus first publicly demonstrated low persistence in its Crystal Cove prototype at CES in January 2014. The feature reached developers in the Oculus Rift Development Kit 2 (DK2), announced at GDC on 19 March 2014 and shipped from July 2014. The DK2 used a low-persistence OLED panel at 960 by 1080 pixels per eye with a 75 Hz refresh rate, and exposed selectable persistence settings of 2 ms, 3 ms, and full. Reviewers reported that the result was a sharp image during head movement, in contrast to the pronounced motion blur of the original DK1, attributing the improvement mostly to low persistence.

Low persistence has been standard in headsets since. On the original consumer Oculus Rift, CTO John Carmack stated that pixels were lit for about 2 ms per frame, and its successor the Rift S used an even lower persistence figure, which Oculus said let it deliver less motion blur than the original despite dropping the refresh rate from 90 Hz to 80 Hz.

The two meanings compared

Aspect	Historical perceptual sense	VR display ("persistence") sense
What it describes	A claimed afterimage that lingers on the retina after the stimulus stops	The time each display pixel actually emits light within a frame
Typical figure	An afterimage on the order of about 0.13 s as measured by d'Arcy	Often around 2 ms per frame on a low-persistence VR display
Current scientific status	Regarded as an incomplete or superseded explanation of motion perception; apparent motion (phi/beta) and flicker fusion are preferred	A well-defined, measurable display property and an active engineering target
Relevance to VR	Background to why sequences of frames read as motion at all	Directly determines motion clarity, judder, and perceived sharpness in a headset

See also

• Judder
• Refresh rate
• Head-mounted display
• Oculus Rift DK2
• Michael Abrash
• Virtual reality
• Augmented reality