Optimal step sizes for colour and luminance staircasing

We investigated the best step size for measuring luminous and isoluminant coloured (red or blue/tritan) detection thresholds. Stimuli were generated on a colour TV monitor as 1°, 600 msec spots at one of four eccentric (1.6°) locations. Thresholds were expressed as contrasts (%) in cardinal colour space (RG, L or BY) for incremental probes from a white background (CIE 1931 x = 0.30, y = 0.31; L=33 cd/m2). One observer (aged 28) was extensively tested, and trends were confirmed on three others (aged 25-33). Staircase step sizes varied from 0.2 to 4.8 deci-log units (dLog) with starting points chosen to he within 2-3 steps of threshold. A logarithmic staircase comprising 13 reversals was performed at each step size, with threshold being taken as the geometric mean of the last 12 reversals of four randomly interleaved staircases. Both threshold and variance increased with larger step sizes. Thresholds show a V-shaped relationship with step size; minimum thresholds were found with 1.25 dLog steps. Variance showed a broken-line relationship with step size, increasing for step sizes >1.0 dLog. No differences were found among the three coloured stimuli. We conclude that small steps (0.4-1.0 dLog) enhance precision at the expense of efficiency (time), whereas large steps (1.5-3 dLog) improve efficiency but reduce precision. In a clinical setting, 2-3 dLog step sizes should be adequate for all colours.