In our metameric experiment, the colour of a filtered incandescent lamp was matched with the additive mixture of three LEDs in a Lummer-Brodhun-type visual photometer. Two sets of primaries were used, one had their dominant wavelengths at 467, 533, and 600 nm; the other set had dominant wavelengths at 478, 552, and 635 nm. These values correspond approximately to the characteristic wavelengths of the Prime and Non-Prime Colour spectral regions defined by W. A. Thornton.(1) Both the light of the incandescent lamp and that of the LED clusters were seen monocularly in a centrally divided bipartite field at a visual angle of 2degrees. The luminance of the matching fields was in the order of 20 cd/m(2) to provide sufficient gamut for the LED mixture. Ten young observers with normal colour vision participated in the experiment. The emission spectra of the viewing fields were measured with an array-type spectro-radiometer, and two sets of colour-matching functions were used to calculate the chromaticity of the matching stimuli: the CIE 1931 standard colorimetric observer and the Judd-Vos modification of the colour-matching functions. We found that the Judd-Vos modification of the CIE 1931 standard observer represents more accurately the real observers in the evaluation of our results. No systematic differences between the use of the two sets of LEDs were detected in contradiction to Thornton'sfindings. (C) 2004 Wiley Periodicals, Inc.