During the past four years I have been conducting surveys of the colour vision of students in the University of Glasgow. The first survey was made by a colour-perception spectrometer very similar to Dr Edridge-Green's instrument, and embraced 79 observers. The second survey was made by Dr Edridge-Green's bead test, and embraced 100 observers. The third survey, carried out in collaboration with Miss Margaret A. Dunlop, was made by an original method, called here for short the microscope test, and embraced 1000 observers. At present there are two other surveys under progress. The object of these surveys is to find a numerical method of specifying goodness of colour vision; to see, by the application of statistical methods, whether the colour blind fall naturally into groups or are merely outliers of a homogeneous population; to find whether colour blindness is a Mendelian characteristic for men and merely an extreme case of normal variation for women; and to throw light on the subject of colour vision generally. Consequently, the normal have been investigated with as much care as the colour blind. But in the course of the four years I have made the acquaintance of many trained observers with abnormal colour vision, and have been possessed with an ever-growing desire to know exactly, irrespective of all theory, what was the matter with their colour vision. In spite of the vast literature on the subject, the tests generally have been of a very superficial nature, and unsatisfactory to the man with mathematical instincts. As these abnormal cases were beginning to leave the University, I addressed myself last spring to the problem of finding a method of testing which would describe their condition independent of theory, and, indeed, independent of words. This paper describes how the problem was solved, and gives data for twenty-three cases of colour blindness, four of normal colour vision, and one case of exceptionally good colour vision.