Relative proportions of long-wavelength-sensitive (L) to middle-wavelength-sensitive (M) cones were estimated by use of the flicker-photometric electroretinogram (ERG). It has been demonstrated that a major source of error in estimates of cone proportions from spectral luminosity functions is the known variation in the lambda(max) of the photopigments [Vision Res. 38, 1961 (1998)]. To correct for these errors, estimates of cone proportions were derived by use of individualized L-cone spectral sensitivity curves deduced from photopigment gene sequences from each subject. For some individuals this correction made a large difference in the estimated cone proportions compared with the value obtained when a fixed standard L cone was assumed. The largest discrepancy occurred in a man estimated to have 62% L cones (L:M ratio 1.6:1) when a standard L pigment was assumed but a value of 80% L cones (L:M ratio 4:1) when his individualized L-cone spectrum was used. From repeated measurements made with the ERG, it was determined that individual estimates of the relative L-to-M cone contributions, expressed as %L cones, are usually reliable within approximately 2%. The average L:M ratio for 15 male subjects was estimated at 2:1 (67% L cones). Previously, a large range of individual variability was reported for L:M ratios obtained from photometry. An unresolved issue concerns how much of the range might be attributed to error. Here efforts have been taken to markedly reduce measurement error. Nonetheless, a large range of individual differences persists. Estimated L:M ratios for individuals ranged from 0.6:1 to 12:1 (40% L to 92% L).