8 Fred J. Bates 



These apparent rotations, however, are not entirely due to the 

 simple cutting down of the amplitudes of certain wave-lengths 

 by the absorbing substance. Superimposed upon this is the 

 Purkinjie effect, or change in the luminosity curve for different 

 intensity. This effect is ordinarily negligible, but under certain 

 conditions it may become considerable. This, we are led to sus- 

 pect by a study of Konig's^ luminosity curves for gas lights at 

 different intensities. Similar curves for sunlight are apparently 

 lacking, so that no calculation bearing upon this point could be 

 made. With the wave-lengths from 573 n-jx to 587 m/x as the source 

 of light and a rotation of 4°, the setting of the analyzer differed 

 by 0.02° when the light intensity was reduced 44 times by the 

 rotating sector. The angle between the principal axes of the 

 nicols in the polarizing system was several minutes of arc. 



In fig. 2, R is the luminosity curve resulting from passing sun- 

 light through a i cm. thickness of alcoholic fuchsin solution, 

 concentration 0.000024. It is obtained by modifying the ordi- 

 nates of the luminosity curve 5" by the transmission curve F. 

 Whenever the difference in the rotation of two substances is be- 

 ing measured, or either the optical center of gravity of the light 

 source is bemg determined for absolute measurement with the 

 half-shade system, or the position of the observing telescope is 

 being calibrated in wave-lengths for either the clear or the black 

 space methods, the errors due to the use of two luminosity curves 

 of which R and 6" are a specific example always enter the mea- 

 surements unless proper compensation be made. These curves 

 may be almost identical for transparent and vary greatly for ab- 

 sorbing substances. In the former case the errors are ordinarily 

 negligible ; in the latter they may become very large. From the 

 theory it is evident that the angle will be a maximum at those 

 points in the spectrum where the comparative slope of the two 

 curv^es is such as to give the greatest difference in the intensity 

 of the longest and shortest wave-lengths reaching the eye. Con- 

 sider curves 6" and R with the half-shade system. At 600 /x/x 

 the error would be approximately the same as that calculated in 

 equation (VI), (0.006°), since the conditions are about the same. 



^Beiirdge zur Psychologic und Physiologic der Sinncsorgane. 



262 



