Large, high-contrast flashes are the highest risk

What it is

Three flashes per second in a 16-pixel corner dot is not the same risk as a half-screen black–white strobe. General flash area threshold: what drives a discharge is the high-contrast area that enters the visual field, not “whether something is moving.” The product criterion makes “large” a measurable window: when content is viewed at 1024×768, concurrent flashes that occupy more than a quarter of any 341×256-pixel rectangle have crossed the general-flash area line.

High contrast is not “it feels punchy.” A general flash needs a relative-luminance change of 10% or more of maximum, with the darker side below 0.80. Opposing transitions between saturated reds are counted as red flashes on their own; a small luminance swing does not clear them.

Why it happens

The stretch of visual cortex driven in sync grows with the stimulus’s visual angle and contrast. A tiny spinner perturbs a small patch; high-contrast alternation that fills central and near-peripheral field pulls a large population onto one rhythm. Full-screen video, a looping background at 100% viewport height, a whole page flashing white under an inverted theme—those are the high-risk shapes. A 24×24 loading dot can blink and still miss the quarter-area line.

Red flash is a separate gate because saturated alternation on the red-cone pathway can be dangerous even when the luminance swing is modest. A grey-luminance-only check misses red–deep-red and red–black material. Area and contrast decide, after frequency is already in question, whether this flash counts as having crossed the general or red threshold. How to count frequency is the previous layer; this layer answers how large and how violent the flash has to be to count.

Studying it

Once flash-pairs have been found in one-second windows, measure area and contrast—do not re-argue “is it fast”:

1. For each opposing pair of frames, mark the pixel set that meets the luminance or red-flash definition and take the union area.
2. Slide a 341×256 analysis window across the viewport (the operationalisation of about 10° of field at 1024×768 viewing). If concurrent flashes occupy more than a quarter of that window, the piece fails on area. Any single window over the line fails the content.
3. Score red flash separately: both states meet a high saturated-red ratio, and the red-versus-green-blue difference crosses the red-flash definition.
4. The same PEAT / Harding analyser reports frequency and area together. In a manual pass, do not stop at “there is a flash at this time code”; open the spatial mask and see whether the flash is local ornament or a large slab of picture.

Independent variables: flashing pixels as a fraction of the analysis window, relative-luminance delta, saturated-red alternation, control-sized versus full-screen layer. Dependent variables: existence of an over-threshold window, number of frames it lasts.

Replay at a real viewing distance and in full-screen; do not estimate coverage on a design artboard.

Where it stops holding

If the user shrinks the window, the fraction changes; conformance is computed at the specified reference viewing condition, not “someone used a small window.” Magnification can turn a small flash into a full field—an acknowledged harsher case, not a reason to relax the main threshold, only extra caution against even small flashes. Flashes under the area line can still steal attention, but they no longer take this medical area gate. High-contrast static or slow-moving gratings are a different photosensitivity risk; tools sometimes report them too—do not merge that number with flash area.

Applying it

• Ban high-contrast on/off of full-screen or large regions: errors get a stable fill plus text, not a page that strobes red; hero video must not contain strobe cuts.
• Keep decorative motion inside small controls and drop the contrast; review saturated red alternation even when the area looks small.
• Verify: run a flash analysis with a spatial mask on the finished piece. Record every over-threshold 341×256 window’s position and coverage, whether relative-luminance change is ≥10%, and whether the dark side is <0.80. Capture full-screen layers, canvas, and embedded players on their own. Changing only frequency, or sliding the flash into the centre of the field while it still fills the analysis window, is not a pass.

Related

• Same group: J4.10.1 Flash rate has a hard safety ceiling · J4.10.3 The harm is medical, not a matter of taste
• Nearby: A1.08 Flicker fusion and display refresh · A1.02.4 Peripheral flicker captures attention · D1.09 Flashing alerts
• Search terms: general flash area threshold · red flash · 341×256 · photosensitive epilepsy