What determines which colour combinations will be attractive to a person and which will not? Is colour attractiveness only a subjective human experience, or can we predict it based on physical colour parameters? One of the pioneers of the attraction of colour theories was Michel Eugène Chevreul (1786–1889). He distinguished two types of colour harmony – analog colour and contrast – and tried to describe what harmonics are based on physical colour parameters. This was later done by other scientists. Later, semantic evaluation of colours was introduced and factor analysis attempted to identify emotions caused by colours or combinations of colours. The aim of this research is to test whether there is a consistent pattern of judgments of colour combinations under controlled conditions and, if so, to what extent they are influenced by the objective physical characteristics of those combinations. Subjects. The study involved 40 students (20 men, 20 women). All subjects had normal colour vision and were not related to fine art. Research tools. The study used 8 colours: 4 opponent (green, red, yellow and blue) and 4 additional (orange, lettuce, blue and purple). The 28 colour combinations (made up of two different colours) were composed of those 8 colours and printed onto cardboard card where each colour had area of 80 mm x 80 mm. Questionnaire of 40 adjectives consisting of 20 pairs of antonyms were used for semantic colour assessment. Procedure. The investigation was conducted in a dark room. Initially, all 28 cards with colour combinations were placed randomly on a desk lit by a 40 cm high fluorescent lamp (4000K correlated colour temperature). The subject was asked to select one of the cards with the most preferable colour combination, to write its code on the questionnaire and to mark all the epithets in the questionnaire which suits this colour combination. The same procedure was applied to the all other cards. One experiment lasted 35–50 minutes. Results and conclusions. Independent component analysis distinguished 4 dimensions describing colours: pleasure, energy, purple color and strength. Logistic regression analysis was run on colour factor loadings to discriminate colour combinations into two groups: liked and disliked colour combinations. It shows that that colour combination could be predicted as being liked or disliked with 85% probability. Adding physical colour parameters to the regression increases prognostic probability to 92 %. Also a relationship between subjective factors and physical characteristics of colour combinations was found. Pleasure correlates with hue contrast and strength with saturation contrast. It can be argued that the reliability of colour combinations is determined by both subjective and physical factors.