RECURRENT VISION 



99 



equal to the sum of its parts. Most physical measurements conform 

 to these axioms, but it cannot be assumed that the same agreement 

 holds good for measurements involving qualitative differences, such 

 as those of colour. The reliability of a method of heterochromatic 

 photometry must be judged according to the following criteria. First, 

 the shape of the luminosity curve must not be altered by change in 

 the reference standard. Second, the sum of the measurements of the 

 brightness of the parts of the spectrum must be equal to the brightness 

 of the recombined spectrum. 



10 



52 



56 



58 



60 



62 



64 



Fig. 38. Photopic luminosity curves taken by the flicker and equality of brightness 

 methods. Abscissae, wave-lengths from 520 p.^ to 640 MM 5 ordinates, arbitrary 

 scale. (Ives.) 



So far as the position of maxima and the general form of the curves 

 are concerned all the methods agree fairly well. Comparison of the 

 different methods shows that the areas of their curves differ most for 

 the visual acuity method as compared with the other three. These 

 latter do not agree well. Of them, the equality of brightness method 

 gives greatest variations, the critical frequency method smaller, and 

 the nicker method least. The crucial test has been applied to the flicker 

 method by Whitman 1 , Tufts 2 and Ives. Of these the last named is by 

 far the most accurate. The method consists essentially in measuring 

 the luminosities of the parts of the spectrum against a standard which 



Phys. Rev. m. 241, 1896. 



2 Ibid. xxv. 433 ; 1907. 



72 



