78 Proceedings of the Eoyal Society of Edinburgh. [Sess. 
ignored for more than half a century, is nevertheless the key to the study 
of colour vision. I have made it the basis of my new method. 
Fig. 5 is a plan of the apparatus employed. C is a collimator mounted 
horizontally close up to a window ; G a piece of ground glass fastened by 
an elastic band in front of its slit S, which was a very wide one. P is 
a glass prism cemented to the window pane : the line of vision of the 
collimator was deflected by this prism 15 degrees up into the western 
B G 
Fig. 5. 
sky. Consequently, it easily cleared the roof of the infirmary at the other 
side of the recreation field. T is a telescope fitted with a ground-glass 
screen E instead of an eyepiece. Under ordinary conditions a rectangular 
white image of the slit would appear on this screen, but in front of the 
telescope objective was fixed a biprism B, with its refracting edge parallel 
to the slit, which had the effect of changing the image into two. These 
images of the slit were about 9 mm. high, mm. broad, and 8 mm. apart. 
In front of the biprism was a brass plate with two equal rectangular 
123456789 
Fig. 6. 
openings, one in front of each face of the biprism, and in front of these 
openings colour filters M and N could be screwed up and down. The 
positions of these filters, and consequently the proportions of the areas of 
the openings which they covered, could be read on scales at the side. The 
object glasses of the telescope and collimator were of 5 cm. aperture, and 
57 cm. focal length. 
Altogether nine pairs of filters were used, the two filters of each pair 
being alike. Fig. 6 represents the whole set of nine. Let us suppose that 
filter 1 is placed at N (fig . 5), with its red half covering the opening in the 
brass plate. Then the corresponding image at E appears a pure red. If 
the filter is screwed up, so that its blue half gradually comes on to the 
