An additive approach to predict harmony for three-colour combinations is proposed in this article. It is hypothesised that a three-colour combination can be seen as a combination of three colour pairs, each generating a harmonious/disharmonious feeling that can he quantified by a two-colour harmony model the authors previously derived; the average of these three harmony values can then determine the overall harmony. To establish whether this hypothesis was valid, two psychophysical experiments were conducted in the United Kingdom and the United States. Experiment 1 used 6545 three-colour wheels as the stimuli, presented individually on a calibrated cathode ray tube display. Under the same viewing conditions, Experiment 2 used 111 interior images as the stimuli. In each experiment, 20 British and 31 American participated as the observers. An additional test was undertaken, with 64 observers taking part, to address the issue of large sample size as encountered in Experiment 1, using 90 colour wheels selected randomly from those used in Experiment 1. The experimental results show close agreement between the observers' response and the harmony value predicted by the proposed method, with a correlation coefficient of 0.71 for the 6545 colour wheels, 0.93 for the 111 interior images and 0.88 for the additional 90 colour wheels. The results support the additive approach as a simple but robust method for predicting harmony in any three-colour combinations, which may also apply to combinations generated by any number of colours. (C) 2010 Wiley Periodicals, Inc. Col Res Appl, 36, 355 372, 2011; Published online 23 June 2010 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/col.20624