Independence of mechanisms tuned along cardinal and non-cardinal axes of color space: evidence from factor analysis

AbstractMany previous studies employing paradigms such as adaptation, masking and summation-near-threshold have demonstrated the existence of separate mechanisms underlying the detection of the three cardinal axes of color space: L+M, L−M and S−(L+M). In addition, some studies have demonstrated the existence of higher-order mechanisms tuned to non-cardinal axes (which are made up of combinations of the cardinal axes). In order to address the issue of separate and independent color mechanisms further, here we applied factor analysis to contrast threshold data obtained from 41 subjects for nine different axes in color space (the three cardinal axes and the six non-cardinal axes midway between). In line with previous studies, the results of a three-factor analysis performed on contrast thresholds for the cardinal axes revealed independence across the three. However, in some of our factor analyses (for example, when a two-factor analysis was performed on the cardinal axes), intercorrelation was observed between L−M and S−(L+M) stimuli. With regard to higher-order mechanisms, our factor analyses revealed mechanisms selective for non-cardinal axes within the (L−M)/(L+M) and (S−(L+M))/(L+M) color planes, but not the (L−M)/(S−(L+M)) color plane. To ensure that the intercorrelation observed between L−M and S−(L+M) cardinal axes was not due to the particular stimulus parameters or testing measures employed, in three of our subjects we performed a “summation-near-threshold” experiment using experimental conditions nearly identical to those in the factor analysis experiments. In accordance with previous findings [Vision Research 39 (1999) 733], L−M and S−(L+M) stimuli were found to be separable in this analysis. This seeming discrepancy between the results of our factor analysis and those obtained from paradigms such as summation-near-threshold can be resolved by proposing that the mechanisms underlying detection of L−M and S−(L+M) stimuli are separable (as defined by the ability to isolate activity within each mechanism using select stimuli), yet nonetheless intercorrelated. Such intercorrelation could arise if these two mechanisms are limited by the same source of variability and/or subject to the same gain control