A computer-controlled test of colour vision is described, in which luminance noise and masking contours are used to ensure that the subject's responses depend on chromatic signals. The test avoids the need--common to most computer-controlled tests--to define equiluminance for the individual subject before the colour test itself can be administered. The test achieves a good separation of protan and deutan subjects and reveals the large range of chromatic sensibilities among anomalous trichromats. As a population, dichromats had higher thresholds on the tritan axis of the test than did normals. In an extension of the test, full discrimination ellipses were measured for normal and colour-deficient observers. The nature of anomalous trichromacy is discussed and the possibility is raised that hybrid genes, resulting from genetic recombination, may code for incorrectly labelled or functionally impaired molecules.