This contribution tries to expel the Dunwich Horror of ＇＇undimensioned＇＇ tristimulus values from colorimetry. I intend to show that tristimulus values have units, and that we may attribute a common dimension to these units. Colorimetric units are independent of physical units. They are sets of primaries, determined by directions and relative lengths in color space but without definition of an absolute length, because in colorimetry quantities generally refer to ＇＇reference stimuli.＇＇ Color-matching functions can be defined as tristimulus values of monochromatic stimuli. To define them as Mr. Brill does in his Definition 1 is possible and consistent, but it does not correspond to the use of these functions in colorimetric practice. And it leads to inconsistencies if color-matching functions are interpreted as having the units watt/watt.