Theory of Colour Vision. 409 



according to the great majority of observers the extreme 

 violet approaches red. It seems consequently that Sir Win. 

 Abney's colour vision is not quite normal in the extreme 

 violet. I have therefore adopted Dieterici's observations as 

 the basis of my work. 



When we have three sets of variables the sum of which is 

 always the same, such as the R, G, B in Table I., they can be 

 represented on a diagram as the perpendiculars from a point 

 on the sides of an equilateral triangle. Maxwell and Konig 

 and Dieterici represent their results in diagrams of this 

 kind, although each of their perpendiculars is in a different 

 unit, Konig and Dieterici's diagram being very well known; 

 there is graph paper on the market for diagrams of this kind, 

 divided into little equilateral triangles. I decided, however, 

 after using seven or eight sheets of it, that trilinear co- 

 ordinates are not to be recommended for colour diagrams. 

 We can get on very much better with ordinary squared 

 paper. 



Fig. 2 represents Dieterici's results, the R being plotted as 

 abscissa and the G as ordinate. The figure can be regarded 

 as a trilinear coordinate diagram, the fundamental triangle 

 being RGB. If, for example, the point P is considered, its 

 perpendicular distances from GB, BR, and RG are respec- 

 tively — 22, 51, and 50'2 units. Owing to the triangle 

 being right-angled isosceles the 50*2 must be multiplied by 

 y/2 giving 71 as a result. The amount of blue present in 

 the colour represented by P may thus be obtained by taking 

 the perpendicular distance from RGand multiplying by ^/2. 

 But it is simpler to add the R and G and subtract the result 

 from 100. 



Fig. 2 is a Newton's colour diagram. It has the property 

 that if any two colours are represented by points and their 

 intensities by the masses of particles placed at these points, 

 then the colour of the mixture is represented by the centroid 

 of the two masses. The curve representing the colours of 

 the spectrum is something like a parabola, although the 

 seven points next the red end lie on a straight line ; I have 

 drawn in a parabola in a broken line, in order to show how 

 much the observations deviate from it. All actual colours 

 are represented by points inside the spectral curve. Sun- 

 light contains 28 R, 69 G, and is represented by the point W 

 in the diagram ; it is thus very close to the yellowish-green, 

 the brightest colour in the spectrum. 



If other fundamental colours are chosen instead of 

 Maxwell's R, G, B, what effect hns this on the shape of the 

 spectral curve ? It can easily be shown that the straight 



