Theory of Colour Vision. 405 



Konig and Dietericl and Abney take as their fundamental 

 colours the extreme red and extreme violet o£ the spectrum 

 and an imaginary green, more saturated than any actual 

 green. Maxwell takes as fundamental colours a red at 

 630*2 fjLfju, a green at 528*1 fifi, and a blue at 456*9 //,//,, all 

 actual spectral colours, chosen at points where the hue varies 

 slowly with the wave-length. I think Maxwell's procedure 

 is undoubtedly the better in this respect ; results in the first 

 case should always be stated in terms of real colours. Also 

 the extreme ends of the spectrum are unsuitable as standards, 

 since they are faint and not visible to some colour-blind 

 persons, and, according to Rayleigh, many persons with 

 colour-vision otherwise normal are anomalous in the extreme 

 violet. Konig and Dieterici and Abney give their results in 

 the form of tables compiled from u smoothed " curves ; 

 Maxwell gives actual observations for sixteen w r ave-lengths 

 selected so as to be equally spaced over a prismatic spectrum. 

 I consider Maxwell's procedure the better in this point also ; 

 in view of the inaccuracy of the results 16 wave-lengths are 

 quite sufficient to take in the spectrum. I have therefore 

 expressed the other observer's results in terms of Maxwell's 

 three standards for Maxwell's 16 wave-lengths ; incidentally 

 I have found Maxwell's and Konig and Dieterici's data 

 incomplete, and I have had to help their observations out by 

 Abney 's. 



Maxwell determines his wave-lengths by what is now 

 called Edser and Butler's method, and expresses them in 

 Fraunhofer's measure. Fraunhofer's wave-length determi- 

 nations were made in Parisian inches, and are quite accurate 

 enough still for colour-vision work. I translated Maxwell's 

 wave-lengths into [Xfju by multiplying by "2707, then keeping 

 the three fundamental wave-lengths unaltered I adjusted 

 the others until the first and second differences increased 

 continuously, and adopted the values so obtained ; the wave- 

 lengths he gives for scale-numbers 20 and 48 are quite con- 

 siderably out. Maxwell expresses the luminosity of his 

 fundamental colours in terms of slit width, and adds his 

 units in figs. 6 and 7 as if they were all of the same value, 

 thus producing a very striking hollow in the top of his 

 luminosity curves. But slit widths in the red, green, and 

 blue have not the same value, owing both to the different 

 intensities of the spectrum at these points and to the different 

 degree of dispersion it undergoes in the prism. According 

 to Abney (1900 paper, p. 285), the luminosities of a prismatic 

 solar spectrum (taken graphically from fig. 10) should be 



