Discriminability is set by luminance contrast, not hue difference

What it is

Whether two visual elements can be told apart quickly and reliably depends mainly on their luminance contrast — the difference in brightness — not on how far apart their hues are. Two colours can differ strongly in hue and still sit at nearly the same brightness, a state called isoluminance; at an isoluminant boundary, fine detail becomes blurry and hard to localize even though the colours are obviously "different."

This is easy to misread as "different colours should be easy to see." Hue difference answers "are these two things different?" Luminance contrast answers "can I read the edge, the fine print, the detail on top of it?" The two answers can diverge: red and green at matched brightness are instantly reported as different colours, yet render as fine text or thin lines they become noticeably harder to read.

Why it happens

The visual system does not use the same channel for "where is something" and "what colour is it." Fine spatial detail and edge localization are carried mainly by a luminance-driven pathway, while the pathway carrying hue information has much lower spatial resolution — the chromatic contrast sensitivity function cuts off at far lower spatial frequencies than the luminance CSF. Hue differences are reliably perceived only at coarse spatial scales; the finer the detail, the faster the chromatic channel's resolving power drops off.

That is why an isoluminant boundary looks soft and its edge seems to drift: the only signal available at that boundary is the low-resolution chromatic channel, and the luminance channel that would normally draw a crisp contour has nothing to work with. Fine text, thin lines and small icons all depend on precise edge localization, so when they rely on hue alone at matched luminance, legibility drops noticeably even if the hue contrast itself is strong.

Studying it

The standard way to determine whether two colours are isoluminant for an observer is heterochromatic flicker photometry (HFP): two colours are flickered rapidly at the same location while the observer adjusts one colour's luminance until the perceived flicker is minimized — the point of minimum flicker is taken as equiluminance. The minimum motion technique works on the same logic using perceived motion instead of flicker.

Typical independent variables: hue angle difference, luminance difference (or luminance held equal by design), spatial frequency of the test pattern. Typical dependent variables: contrast detection threshold, edge localization accuracy, detail-recognition accuracy.

In interface research this is often used as a worst-case test: flatten a colour scheme's luminance and check whether information is still legible from hue alone. If it is not, the scheme is at risk in real rendering conditions, where device gamut and ambient light can shift the luminance relationship between colours.

Methodological caution: the isoluminance point shows substantial individual variation — age, lens yellowing, and personal spectral sensitivity all shift it — so a colour pair measured as isoluminant for one observer is not guaranteed to be isoluminant for another, and lab values should not be treated as a fixed constant for all users.

Where it stops holding

• Large colour fields are not bound by this. Luminance contrast mainly constrains fine detail and edge localization; distinguishing two large colour regions from each other does not depend on sharp edges, so hue alone is usually enough.
• This is about detail legibility, not category discrimination. Two highly saturated but isoluminant colours are still instantly reported as "different" — what slows down is locating an edge, reading fine print, or resolving a pattern printed on them.
• Aging makes it worse. Lens yellowing with age further attenuates blue-yellow chromatic signal, so older users read isoluminant boundaries worse than younger ones.
• The isoluminance point drifts across devices. A pair measured as equiluminant on a calibrated lab monitor is not guaranteed to hold on a device with a different gamut or white point.

Applying it

• Check the luminance difference, not just hue difference, between any two elements that need to be told apart — text on background, icon on background, border on background. A quick proxy: convert the screenshot to grayscale and see whether the distinction survives.
• For thin lines, small text, and densely packed small icons, use luminance contrast as the primary way to build hierarchy; treat hue difference as a supporting cue, not the main one.
• Large background regions and decorative colour blocks can rely on hue alone; forcing a luminance gap there is not necessary.
• How to check: convert the colour scheme to grayscale, or run it through a colour-vision simulator that strips the hue channel, and see whether key information — text, icon edges, status distinctions — remains legible. Anything that disappears in grayscale is an isoluminance risk.

Related

• Same group: A1.04.2 Contrast sensitivity varies with spatial frequency; thin strokes need higher contrast · A1.04.3 Contrast sensitivity declines with age and further in low light
• Nearby: J2.03 Information must not depend on colour alone · F3.03 Colour and lightness hierarchy · J2.10 Non-text contrast
• Search terms: luminance contrast · isoluminance · chromatic contrast sensitivity · heterochromatic flicker photometry