In the tests which are now to be reported, three portions of the spectrum were used: the red end, the blue-violet end, and a middle region, chiefly green. The red light was obtained by the use of a filter which was made by placing two plates of ruby glass in one of the glass boxes, filling the box with filtered water and then sealing it to prevent evaporation. The blue-violet was obtained by the use of a filter box which contained a 5 per cent solution of copper ammonium sulphate. The green, which, however, was not monochromatic, was obtained by the use of a filter box which contained a saturated solution of nickel nitrate. These three sets of filters were examined spectroscopically both before the experiments had been made and after their completion.[1] The red filters, of which I had two for shifting the lights, transmitted only red light. The blue-violet filters, two also, at first appeared to transmit only portions of the blue and violet of the spectrum, but my later examination revealed a trace of green. It is important to note, however, that the red and the blue-violet filters were mutually exclusive in the portions of the spectrum which they transmitted. Of all the filters used the green finally proved the least satisfactory. I detected some yellow and blue in addition to green in my first examination, and later I discovered a trace of red. Apparently the transmitting power of the solutions changed slightly during the course of the experiments. On this account certain solutions are undesirable for experiments on color vision, for one must be certain of the constancy of the condition of stimulation. It is to be understood, of course, that each of the three filters transmitted, so far as the eye is concerned, only the color named. I consider the red filter perfectly satisfactory, the blue- violet very good, and the green poor. Henceforth, in testing color vision in animals, I shall make use of colored glasses as filters, if it is in any way possible to obtain or have manufactured blue, green, and yellow glasses which are as satisfactory as the ruby. [Footnote 1: A Janssen-Hoffman spectroscope was used.] The apparatus needs no further description, as its other important features were identical with those of the reflected light experiment box. The use of artificial light for the illumination of the electric-boxes made it necessary to conduct all of the following tests in a dark-room. The method of experimentation was practically the same as that already described. A mouse which had been placed in _A_ by the experimenter was permitted to enter _B_ and thence to return to _A_ by entering one of the electric-boxes, the red or blue or green one, as the case might be. Mistakes in choice were punished by an electric shock. One further point in the method demands description and discussion before the results of the tests are considered, namely, the manner of regulating and measuring the brightness of the lights.