Tooth colour is often measured with a small window for illumination and measurement. This causes edge loss of the light, resulting in systematic errors in colour coordinates. This paper gives a quantification of the edge losses, and explains their cause. We measured reflectance spectra for 27 Formalin fixated extracted incisors using a small-window reflectance spectrophotometer equipped with external diaphragms of 3, 4, and 5 mm diameter, and using a spectroradiometer. We calculated the colour coordinates L*a*b* from these spectra. Finally, 16 randomly chosen teeth were illuminated with a pencil beam (lambda = 543 nm, and lambda = 633 nm) while the emerging light was measured as a function of distance from the illuminated spot using a CCD detector. These data were used to calculate small-window edge losses, and thus to predict the small-window reflectance factors relative to spectroradiometrically determined reflectance, at both 543 nm and 633 nm. In all instruments the same spot on the tooth was illuminated and measured, and the teeth were always wet. Colour coordinates for the small-window colour measurements deviate significantly from those determined using the spectroradiometer. These deviations can be explained from the wavelength-dependant edge loss that arises in small-window colour measurement.