354 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



It is difficult to see how non-parallelism of slit and ruling could in 

 this case introduce a shift. Keller seems to have overlooked the fact 



Figure 1. 



AA', direction of grating rul- 

 ing; EE', direction of slit; LL', 

 direction of resultant line. 



I 



F 



Figure 2. 



AA', two exposures of an arc line 

 superimposed horizontally, but not 

 vertically ; FF', spark line. 



that this triple method was employed, for no mention is made of it in 

 his paper. However, despite the fact that it was not apparent how 



