Semantic-differential data of color emotions for color pairs were collected and examined with factor analysis (FA) and independent component analysis (ICA). The axial orientations of factors cannot be identified by FA, whereas they can be determined by ICA with the intrinsic statistical properties of data such as kurtosis. Three factors or components were extracted by FA and ICA. The factors extracted by FA with the varimax method were consistent with the primary factors: evaluation, activity, and potency. When ICA extracted components with positive kurtosis, the components did not match any of the primary factors. However, the independent components extracted by ICA for negative kurtosis were consistent with the primary factors, and similar to the factors obtained by FA. The results of ICA suggest that the evaluation, activity, and potency factors are independent dimensions in psychological space of color emotions for color pairs. The distributions of the factor scores in FA and the scores of the independent components obtained by ICA for negative kurtosis were not Gaussian, but they had negative kurtosis. These support the use ofICA instead of FA. (c) 2012 Wiley Periodicals, Inc. Col Res Appl, 38, 297-308, 2013