

PROCEl-lUMiS OF Till". AMERICAN ACADKMV. 



different plates varies considerably. This is probably due to the fact 

 tii.it it was difficult to set the very end of the spark image accurately 

 u I m. 11 tin' -lit. As ahown in the previous paper, the part of the image 

 employed influences the character of the line and the value of the 



shift. ' 



During the progress of the work it was suggested to us that the use 

 of the tip of the spark line as that part of the line upon which to set 

 the thread of the microscope in measuring was perhaps objectionable 

 owing to the fact that there might be a shift due to diffraction result- 

 ing from reducing the virtual aperture of the grating by strips of 

 black paper set only roughly perpendicular to the ruling, the measure- 

 ment being made by a mm. scale. Three exposures on one plate were 

 therefore made, — all of the arc, and the first and third superimposed 

 as usual. No shift was shown when the slit was either parallel or 

 oriented, as indicated in the table on page 3o9. 



At the end of the series of experiments the water rheostat was cut 

 out of the transformer circuit, and in its place was inserted a choke 

 coil of closed magnetic circuit of U form with adjustable armature. 

 When adjusted roughly to show maximum power as measured by the 

 wattmeter, with a spark-length as indicated in Table III, the shift was 

 increased to 0.032 t. m. in the mean for X 3900.68 and 0.033 t. m. for 

 X 3913.58. 



TABLE III. 



Conditions same .is in Table I, except spark length = mm. in plate 125 and 15 mm. 

 in plates 126 to 128. Time of exposures for spark — 60 seconds. 



