288 PROCEEDINGS OF THE AMERICAN ACADEMY. 



on each of three lines : Ti, AA 3898. 645, 3901.114, and 3904.926 (Row- 

 land's table of solar wave-lengths), as they appeared on a focus plate. 

 The first line only was reversed : all were somewhat too dense for very 

 accurate settings. Assuming the values of the wave-lengths of the first 

 and third lines as given, that of the second was calculated. The first 

 determination gave 3901.112, the second 3901.114; the mean 3901.113, 

 but 0.001 tenth-meter different from the wave-length as given above. 



Unless otherwise stated the developer used was the Wallace solution. 



The spark lines which were studied most thoroughly, namely Ti, 

 AA 3900 and 3913 ; Zn, AA 4680, 4722, and 4810, generally appear en- 

 hanced, diffuse, and unsymmetrically broadened. Many iron lines also 

 show unsymmetrical broadening to some extent. 



Figures 1, 2, and 3 of the accompanying plate are illustrative of the 

 lines studied. The scratches, appearing at the tips of the spark and 

 between the arc and spark segments, were drawn by pricking two holes 

 in the film of a positive of the original negative at points which seemed 

 to lie in the "centre of gravity" of the very tips of the spark lines. 

 These two points were then joined by a ruler and the film carefully 

 scratched away along its edge in the four regions above mentioned. 

 A negative of the positive was then made. From the three negatives 

 prepared in this manner the positive reproductions were developed. 



The arc line lies considerably to the left, or violet, in all cases but that 

 of Figure 2, which is a " self-induction line." It will be noticed that 

 the double arc exposure as sketched in Figure C appears only in Figure 

 3 of the plate. As stated previously, this non-superposition method 

 was not used at the very first of the investigation. 



The scale of the enlargements is about five to one. 



Let us discuss first the results as given by the titanium plates in 

 Table III: — 



1. The members of the set of six plates, set 1 (see next to last column 

 of the table), agree in showing that with large capacity shifts exist (about 

 0.023 t.-m.), even in the absence of a secondary gap. 



2. Set 2 (5 plates) shows that, with a secondary gap of 4 mm. and 

 lessened capacity, considerable shifts exist (about 0.037 t.-m.). The 

 effect of a secondary gap is to increase the potential of the discharge, 

 and this gives the same result as an increase of capacity. 



3. A coudition (set 3) given by an increase of secondary gap and 

 further reduction of capacity still presents a shift (about 0.033 t.-m.). 

 It may also be seen that plates Nos. 8, 16, and 25 (sets 1, 2, and 3), on 

 which the arc spectrum was photographed outside of the spark spectrum, 



