The hypothesis that dichromatic behavior on a clinical anomaloscope can be explained by the complement and arrangement of the long- (L) and middle-wavelength (M) pigment genes was tested. It was predicted that dichromacy is associated with an X-chromosome pigment gene array capable of producing only a single functional pigment type. The simplest case of this is when deletion has left only a single X-chromosome pigment gene. The production of a single L or M pigment type can also result from rearrangements in which multiple genes remain. Often, only the two genes at the 5' end of the array are expressed; thus, dichromacy is also predicted to occur if one of these is defective or encodes a defective pigment, or if both of them encode pigments with identical spectral sensitivities. Subjects were 128 males who accepted the full range of admixtures of the two primary lights as matching the comparison light on a Neitz or Nagel anomaloscope. Strikingly, examination of the L and M pigment genes revealed a potential cause for a color-vision defect in all 128 dichromats. This indicates that the major component of color-vision deficiency could be attributed to alterations of the pigment genes or their regulatory regions in all cases, and the variety of gene arrangements associated with dichromacy is cataloged here. However, a fraction of the dichromats (17 out of 128; 13%) had genes predicted to encode pigments that would result in two populations of cones with different spectral sensitivities. Nine of the 17 were predicted to have two pigments with slightly different spectral peaks (usually < or = 2.5 nm) and eight had genes which specified pigments identical in peak absorption, but different in amino acid positions previously associated with optical density differences. In other subjects, reported previously, the same small spectral differences were associated with anomalous trichromacy rather than dichromacy. It appears that when the spectral difference specified by the genes is very small, the amount of residual red-green color vision measured varies; some individuals test as dichromats, others test as anomalous trichromats. The discrepancy is probably partly attributable to testing method differences and partly to a difference in performance not perception, but it seems there must also be cases in which other factors, for example, cone ratio, contribute to a person's ability to extract a color signal from a small spectral difference.