Abstract
Recently it has been demonstrated that colour constancy increases significantly if a surface moves
across a textured background (Werner, 2004), as compared to an equivalent static condition (typical
Mondrian presentation). This was found to be selective for object motion, suggesting the involvement of
higher order motion mechanisms. Different motion processes can be distinguished on grounds of their
temporal characteristics. In order to evaluate the involved motion processes, the temporal tuning of the
effect of motion on colour constancy was determined. The background pattern consisted of a static heterochromatic
checkerboard pattern (18 x 18 deg, Lmean = 19.3 cd/m 2) in front of which a testfield (1.8 1.8 deg, Ltest = 19.3 cd/m 2) moved horizontally with constant speed. The chromaticity of the testfield
was identical to the mean chromaticity and was under standard condition achromatic (u* = 0.197, v* =
0.468). Colour constancy was tested for a simulated homogenous illumination change across the entire
test-pattern from D65 towards a perceptually green illumination (u* = 0.166, v* = 0.472) resulting in
a chromaticity shift of testfield and background along an equiluminant L/M axis (ΔE ∗ uv = 22.88).
Colour constancy was measured using a hue cancellation method for the achromatic appearance of the
testfield after a 5 s adaptation period to the new illumination, whereby the testfield moved continuously
with one of 8 different speeds (0.3 to 14 deg sec-1). Three subjects with normal colour vision (Cambridge
Colour Test) and normal visual acuity participated in the experiments. Colour constancy was found to
increase nearly linearly with speed, reached a maximum value around 2 deg sec-1 and then dropped off
sharply at higher speeds. The corner value for higher velocities (velocity at which the motion induced
increase of colour constancy has dropped to 1
2 of its maximum value) is equivalent to a temporal frequency
of 2 Hz. It is therefore consistent with the temporal properties of the relatively slow third order
motion/feature tracking system, which can be linked to tracking the position of objects moving across a scene (Lu & Sperling, 2001; Seiffert & Cavanagh, 1999).
