OF ARTS AND SCIENCES. ' 261 



in the various spectra on the same plate. A portion of the difference 

 may be due to the impossibility of setting the lines exactly at right 

 angles to the micrometer screw, and in some cases, as in Plate 56, to 

 a possible movement of the plate between the measures of the two 

 spectra. This cannot account for all the difference, however, since in 

 several cases it is obvious to the eye. The difference in wave-length 

 would vary in different parts of the spectrum, but an idea of its mag- 

 nitude may be inferred by the rule that one turn of the screw near the 

 II line corresponds to a change of 80 in the wave-length. The two 

 spectra are thus disjDlaced l.y several units, an amount which the 

 residuals show is quite beyond the accidental errors of measurement. 

 The adoption of a new zero for each spectrum thus appears to have 

 been entirely justified. 



From the last two columns of Table X. we see that the plates are 

 sensitive to rays of light of wave-lengths 3750 to 4800, that is, from 

 M nearly to F. When the light is intense, the spectra extend beyond 

 these limits. F is photographed in Plate 73, and in Plate 21a line 

 marked '' very doubtful " was observed at 3708. It was not included 

 in Tables III. and lY., as it probably only indicates the beginning of 

 the spectrum. 



To secure entire independence in the results, the measures were 

 completed before the reductions were begun. The lines in each plate 

 were measured without comparison with any map, and no search was 

 made for lines which appeared to be wanting. When two similar 

 spectra were photographed side by side, as in Plate 21, care was taken 

 to cover one when measuring the other. Under these circumstances, 

 the agreement in the measures of several plates is strong evidence of 

 the identity of the spectra. 



