Table 1. Differences in the fly-subjective brightness and chromaticity of the background colors

Animals were trained to discriminate and generalize between colors presented alone or as background for black T-shaped patterns. Colors had either nonoverlapping (1), partially overlapping (2), or completely overlapping spectra (3/4). Brightness differences were estimated through the receptor signals of R1–R6 receptors. All colors presented in sequence differed strongly in brightness, except those with the partially overlapping spectra (2). Chromaticity was computed for different input of R7 and R8 receptor signals. The second column shows values characterizing the color differences using input from all four spectral R7 and R8 receptor types. Note the larger the value, the better colors would be distinguished by the flies. The third and fifth columns (Rh3/Rh5 and Rh4/Rh6) refer to hypothetic single-opponent mechanisms based on the input from the R7/R8 pair in two ommatidial subtypes. The fifth column (Rh4–Rh6) shows the results of calculating the chromatic input from the three predominantly excited receptor types. The behavioral data are closest to the color differences as predicted from the opponency signal of the central photoreceptors of the Rh4/Rh6 ommatidial type and coherently from the Rh4–Rh6 input being dominated by the signals from Rh4/Rh6 ommatidia. For instance, flies generalized between spectra of similar shape (3–6) and did not generalize between spectra with dissimilar shape (1–2), and also have larger chromatic differences than spectral pairs in 3/4 and 5/6.