Color perception test: method

What it measures

(1) The hue at which the visitor switches from calling a colour 'green' to 'blue' (a colour-category boundary). (2) Fine hue discrimination, as errors in ordering colours of equal lightness and chroma on the visitor's screen.

Procedure

All colours are OKLCH with lightness 0.72 and chroma 0.12 (inside the sRGB gamut at every hue), so only the hue changes; they are shown on a neutral mid-grey. Part 1: one patch at a time from the hue range 140° (green end) to 250° (blue end); the visitor answers Green or Blue (buttons or the ← and → keys); the patch stays until the answer, then 350 ms of grey. Two staircases start at 150° and 240°; on each trial one of the unfinished staircases is chosen at random. A 'green' answer moves that staircase towards blue, a 'blue' answer towards green; the step starts at 16° and halves at each reversal down to 2°. 12 trials per staircase (24 in total). Part 2: 52 caps evenly spaced round the hue circle (about 6.9° apart, starting at hue 0°) form 4 rows; each row has a fixed cap at both ends and 12 shuffled caps between them. The visitor drags caps, or taps two to swap them, until the row changes smoothly. No time limit, no practice.

Parameters

Scoring

Boundary: a logistic function P(blue | hue) = 1 / (1 + exp(−(hue − b) / s)) is fitted to all 24 answers by maximum likelihood over a grid (b in 0.5° steps, s from 1.5° to 40°); the boundary is b, the 50% point, in degrees of OKLCH hue, limited to 140°–250°. If every answer was the same, the boundary is set to the range end and marked as beyond it. No right or wrong value. Ordering: caps are numbered round the circle; each movable cap scores |n − n_left| + |n − n_right|, and the error score is the sum of (cap score − 2) over all caps (0 = perfect order, as in the 100-hue test), per row and in total; lower is better. Also reported: the number of misplaced caps and the hue region with the most error.

Valid-run rules

Each staircase must have exactly 12 answers, and every shown hue must match the hue the staircase rules give for the answers so far (checked on the server to ±0.05°); each row must contain exactly its own caps. Otherwise the run is rejected.

Differences from the classic task

The Farnsworth–Munsell 100-hue test uses 85 physical caps in 4 trays under a standard illuminant and has clinical norms. This version uses 52 screen colours of our own (48 movable) at constant OKLCH lightness and chroma, on uncalibrated displays, so its scores cannot be compared with clinical norms. The blue–green boundary task is not part of the 100-hue test.

Limitations

Screens differ in colour reproduction, and night modes, blue-light filters, brightness and room lighting shift colours; results are comparable only on the same screen and settings. The boundary also depends on the colour words of the page language. Not a colour-vision screening or a diagnosis.

Languages and review status

Language-independent colour stimuli, instructions in all site languages. The answer labels 'Green' and 'Blue' are translated, so the naming boundary reflects the colour terms of the page language (some languages use one word for both).