Firstly, we tested whether perceived unique hues are linearly related to the cone signals (see Model). We fitted null planes for red, green, yellow and blue. The figure depicts the deviations from a linear model, expressed in perceptual errors. The average error is relatively small (<5) hence supporting a linear model.

Secondly, to test whether different observer have different colour appearance mechanisms, we fitted the linear model for each observer individually (INDIVIDUAL in the figure) or we assumed that all observers are the same and fitted all data with a single set of parameters (POOLED). In both cases, we get good fits (error < 5), which suggests that the variability between observers is relatively small.

In summary, when assessed with this hue selection task, colour-normal human observers seem to have very similar colour appearance mechanisms even though they may have different relative (L:M)  cone numbers. Where this invariance comes from is still a matter of debate.

^ Back to top