Lightness and Saturation Confound

What it is

Karen B. Schloss, Christoph Witzel and Lai ran two experiments and a supplementary one: the first reanalysed colour and emotion ratings for the thirty-seven Berkeley Color Project colours with the terms happy, sad and angry, with forty-eight participants, controlling lightness and chroma statistically, and the second had twenty participants judge thirty-two colours with lightness and chroma controlled colorimetrically. In the authors' words, when they controlled lightness and chroma statistically in the first experiment and colorimetrically in the second, yellow hues were no happier than blue hues and in some cases blue hues were actually happier, and their reading is that previous evidence of yellow being happier than blue is likely due to variations in lightness between coloured stimuli rather than variations in hue.

Domicele Jonauskaité and Christine Mohr's review reaches the same conclusion from the other direction, across a hundred and thirty-two peer-reviewed articles and 42,266 participants in sixty-four countries, finding that all basic colour categories have systematic correspondences with valence, arousal and power and with discrete affective terms, that most of those correspondences are many to many, and that the systematic effects are driven by lightness, saturation and hue. Alex O. Holcombe's commentary adds the adjacent defect, that almost every published colour claim rests on testing one hue, or sometimes a few, and that he found no paper stating its conclusions in terms of the specific hue or hues tested.

In effect

It disposes of the one negative association blue is popularly assumed to carry, because blue is not sad in these data once lightness and chroma are held constant, and that removes a standing objection to using it by controlling a variable rather than by arguing. It also tells you which lever is real, since a designer choosing between a pale desaturated blue and a deep saturated one is changing more, on this evidence, than a designer choosing between blue and green at matched lightness and chroma. The honest other half of the same result belongs with it, that yellows retain the greatest potential to be happiest, not because of hue but because they can simultaneously maximise lightness and chroma, which is a real property of the colour space rather than a confound.

What it does not say

It does not say hue never matters, because Schloss and colleagues predicted and found that redness and anger genuinely does track hue and they draw that contrast deliberately, so the claim is about which associations survive the control rather than about hue being inert. It does not generalise beyond rated associations to experienced emotion, a limit Jonauskaité and Mohr state themselves, since most correspondences were tested at the level of association and it remains to be seen whether they translate to the impact of colour on experienced emotions and specific contexts, and nothing in this material is a behavioural outcome. The samples are small where the control is tightest, at forty-eight participants in the statistical experiment and twenty in the colorimetric one, which is enough to make the confound visible and not enough to map it. And the confound cannot currently be measured properly because of Holcombe's problem, as separating hue from lightness and saturation across a colour category needs several hues per category, and on his own costing a five-hue study at the field's recommended power would need six hundred and fifty participants in a laboratory, which has not been run.

---