In the ordinary spectrophotometer and in Laurent's “half-shade “polarimeter, two neighbouring patches of light of the same colour but of different intensities are presented to the eye of the observer, who by an appropriate means reduces the intensity of the brighter until in his judgment it is brought down to the same intensity as the other. The accuracy of such a measurement must depend on two factors. The first factor is the accuracy with which the observer's eye can judge of the equality of the two patches of light, and the second factor is the accuracy with which the instrumental reading indicates the intensity of the comparison beam, i.e., of the beam whose brightness is reduced till it becomes equal to that of the other. Now it is found that in ordinary cases the error of the eye's judgment in such measurements amounts to about 4% or 5%, while the measurement of the instrumental regulation of the light can be made much more accurately.