360 Mr. H. E. Ives on the 



Let us consider the measurement of spectral colours 

 against one taken as the standard of illumination, or against 

 a white. On the critical frequency-log. I diagram, the 

 standard and the colours to be measured are each represented 

 by straight lines. At a given illumination a quantity of 

 any coloured radiation may be chosen such that flicker will 

 disappear with it at the same speed as with a chosen standard. 

 If the rise and fall or the contour of the flicker sensation is 

 independent of the speed or changes in the same manner 

 for all colours with change of speed, then this coloured light 

 may be alternated with the standard, and the speed may be 

 reduced to the point where colour flicker is just about to 

 appear — forming a flicker photometer. The speed not 

 altering the equality of the two contours, the relative 

 quantities of the standard light and the coloured will be the 

 same as was required to make flicker disappear with each 

 separately at the critical frequency for the standard. If 

 this is the state of affairs the same relationship between 

 critical frequency and slit widths exists for both standard 

 and colour in the flicker method as with the method of 

 critical frequency. That is 



F = K s logI+^ s , 



F=KxlogSx+px. 



We may then substitute for the critical frequency in question 

 its value in terms of the standard illumination, and obtain 

 as a result 



K s logI + C = KxlogSx, 



that is, the logarithms of the slit-widths will plot as a 

 straight line against the logarithms of the illuminations. 



On the second assumption the relative quantities of the 

 standard and the coloured radiation would be read off 1 the 

 diagram (fig. 4) for the speed at which the flicker photo- 

 meter is run. The possibility is, however, practically ruled 

 out by the fact that, as shown by reference to the data of 

 the first paper, the flicker photometer speeds for an illumi- 

 nation showing a strong reversed Purkinje effect are 8 to 

 10 cycles per second ; therefore, from the data in the 

 present paper they are well down in the Purkinje region. 

 Speed cannot therefore be the dominant factor. 



Considerable time and care have been spent in testing the 

 first possibility. The data in the previous paper were plotted 



log Sx 

 •575 /* 



in the form of f ^g^ ~ against log I. The resultant lines 



