The authors have previously hypothesised that colour vision has evolve
d not only to encode colour per se but also, perhaps principally, to e
nhance luminance-based visual processing, so that far colour informati
on to be fully effective, luminance as well as chromatic variations sh
ould be present in visual targets. Results of previous experiments, te
sting detection of spatial gratings and detection and perceived bright
ness of Mach bands support the hypothesis. Further experiments are rep
orted in which the hypothesis was tested by using a higher-level task
of pattern recognition. Subjects had to discriminate between luminance
(isochromatic), isoluminant (chromatic), or combined colour/luminance
ellipses and circles. It was found that the ability to discriminate b
etween a circle and an ellipse was greatly enhanced when both colour a
nd luminance variations were present as compared with the pure luminan
ce or colour presentations. Summation-square analysis shows linear col
our-luminance summation which can be modeled by a single-analyser mode
l.