Display Refresh and Flicker Delivery

Display refresh and flicker delivery concerns a failure mode in screen-delivered visual entrainment: a display cannot produce an arbitrary flicker frequency, and requesting one silently produces a different frequency instead.

No encyclopaedia covers this, and neither does most of the device literature. It is documented here because the Institute found it in its own shipped software on 7 September 2026 and fixed it, and because the failure converts the safest frequency in the entrainment catalogue into the most dangerous one.

The mechanism

A display changes what it shows only at a frame boundary. Software that paints a flicker per animation frame therefore cannot place a transition anywhere else — the requested waveform is quantised to the refresh rate. Simulating the common panels for a 40 Hz request:

{| class="wikitable"

! Requested !! Display !! Actually delivered !! Pattern

| 40 Hz || 60 Hz || ~20 Hz || <code># . # # . # # . # # . #</code>

| 40 Hz || 120 Hz || ~39.8 Hz || <code># # . # # . # # . # # .</code>

| 40 Hz || 144 Hz || ~39.8 Hz || <code># # . . # # . . # . . #</code> — irregular duty

| 40 Hz || 90 Hz || ~39.8 Hz || <code># # . # . # . # # . # .</code> — irregular duty

On a 60 Hz panel — the commonest display in the world — a 40 Hz request resolves to a repeating three-frame on-off-on pattern. That is a 20 Hz stimulus, and 15–25 Hz is the most seizure-provocative photic band in the clinical literature (Fisher et al., Epilepsia 2005). The request for the safest catalogue frequency becomes, silently, a stimulus at the peak of the provocative range.

Which refresh rates can render 40 Hz

Only an integer multiple of the target: 80, 120, 160 or 240 Hz.

Not 60. Not 90. And not 144 — which gives 3.6 frames per cycle, so the most common "high refresh" gaming monitor cannot do it either. This is the result that surprises people.

Note also that at 120 Hz a three-frame period cannot be 50% duty — it is 33% or 67%.

Why it is easy to ship

There is no <code>screen.refreshRate</code> in the web platform. The <code>Screen</code> interface exposes <code>availHeight</code>, <code>availWidth</code>, <code>colorDepth</code>, <code>height</code>, <code>isExtended</code>, <code>orientation</code>, <code>pixelDepth</code> and <code>width</code> — nothing about timing. So a browser-delivered protocol must measure the refresh rate at runtime and then act on the measurement.

The Institute's own page did the first half and not the second: it measured the rate, displayed it to the participant, and never read it back. The honesty was in the prose and not in the code — which is the failure mode to expect wherever a stimulus is delivered by a display rather than by dedicated hardware.

WebXR is the exception: <code>XRSession.frameRate</code>, <code>supportedFrameRates</code> and <code>updateTargetFrameRate()</code> are the only web surfaces that both report and request a display rate.

The check

Log the timestamps the painter actually paints on over a few seconds, take the median inter-frame interval, then:

ratio = measured_refresh / target_hz

clean if abs(ratio - round(ratio)) < 0.02 and round(ratio) >= 2

If it fails, the honest options are to refuse the visual channel, to run audio only, or to move to hardware. Substituting a nearby rate is not an option, because the nearby rate is the hazard.

Scope

This affects screen-delivered stimuli: browsers, phones, tablets, apps, at-home and remote study arms, and head-mounted displays. It does not affect an LED array driven by a microcontroller, which is what most laboratory rigs use — laboratory work is generally safe from it and clinical at-home arms generally are not.

Phone panels are commonly 60 Hz or LTPO-variable (1–120 Hz, shifting with content and battery state), which is the worst case: the rate measured at the start of a session may not be the rate ten minutes in.

Head-mounted displays

120 Hz divides 40 exactly. 90 Hz and 72 Hz do not. An effect authored and verified at 120 Hz is not the same effect in a 90 Hz session, and nothing in a typical pipeline warns anyone.

The enclosure compounds it. ITU-R BT.1702-3 Annex 5 names a dark room, a bright or large screen, and close viewing as the three aggravating factors for photosensitive seizure risk — a head-mounted display is all three at once, with no periphery to look away into and the device fastened on. A full-field effect also fails both limbs of WCAG 2.3.1 by construction.