564 HENKY A. ROWLAND 



which the camera was displaced. A straight line was then passed through 

 all the marks as nearly as may be, and the correction taken off. This 

 correction could thus be obtained to T^TF division of Angstrom, and 

 amounted to only a few hundredths of a division at most. Possibly T 1 - 

 division of Angstrom was the greatest correction required for length. 



In this way each plate represents the average of all the wave-length 

 determinations throughout its extent, and will not admit of any correc- 

 tion save a linear one, should such ever be required in working over the 

 table again. 



In every plate having a solar and metallic spectrum upon it, there is 

 often indeed always a slight displacement. This is due either to 

 some slight displacement of the apparatus in changing from one spectrum 

 to the other, or to the fact that the solar and the electric light pass 

 through the slit and fall on the grating differently. In all cases an at- 

 tempt was made to eliminate it by exposing on the solar spectrum, both 

 before and after the axe, but there still remained a displacement of 

 TTTF to yf-g- division of Angstrom, which was determined and corrected 

 for by measuring the difference between the metallic and coinciding solar 

 lines, selecting a great number of them, if possible. 



The changes from sun to arc light are much more extensive than from 

 one order of solar spectrum to another. In two cases I have tested the 

 latter and found no displacement, and have no fear that it exists in 

 the others. 



In working up the plates, I have started at the plates whose centre is 

 at wave-length 4600, and proceeded either way from that point. For 

 this purpose I have used the plates originally obtained for metallic 

 spectra, generally using the lines due to the impurities. The method, 

 I believe, is obvious from the table. For a long region no substandard^ 

 are necessary, but are used whenever they become so. 



[The tables are omitted.! 



