560 HEXEY A. ROWLAND 



o 



amounting to less than one- or two-tenths of a division of Angstrom, and 

 so corrected the whole series to the standards. Sometimes two, or even 

 three, overlapping spectra were measured at once. 



To make Table X, the following process was used: 



1st. From all the observations at my disposal, I determined a few 

 more lines around the main standards, and put them in the second col- 

 umn, marked St., so that I should have a greater number of points to 

 draw my curve through. 



2d. I then put down a few observations which were made by meas- 

 uring overlapping spectra. 



3d. Then the main eye observations were put down as follows: 



p extending from 4071 to 7040, 3d spectrum, 14,436 grating. 

 q " 4999 to 7035, " " " " " 



" " 4859 to 7040, " " " " " 

 c " " 4859 to 6079, (fragmentary^. 



" 6855 to 6909, 2d spectrum, 14,436 grating. 



5162 to 7201, " " " " " 



5742 to 7628, 1st spectrum, 14,436 grating. 



1 " " 6065 to 7671, " " '< " " 

 C' ' " 6855 to 7714, plane grating. 



R' '< " 5139 to 5296, 2d spectrum, 14,436 grating. 



t " " 6499 to 6929, " " " " " 



a " " 6278 to 6322, " " " " " 



E " " 4048 to 4824, " " " " " 



4th. The series of photographs containing coincident spectra, mostly 

 on plates so short as to make the spectra nearly normal, were now in- 

 troduced. The plates were numbered from 1 to 20, ISTos. 7 and 19 being 

 rejected because imperfect. 



This series of plates was obtained by photographing a narrow strip 

 of one spectrum between two strips of another, the overlapping spectra 

 being separated by absorption. In order to eliminate any change in 

 the apparatus during the exposure, the latter was divided into three 

 parts, the first and third being given to the same spectrum. 



This series of plates gives me a continuous series of photographs from 

 wave-length 7200 to the extremity of the ultra violet spectrum, each 

 part being interwoven with one or two other parts of the spectrum. 

 Thus, wave-length 3900 comes from 5200 and 5850 with only a slight 

 difference in values. There is scarcely any difference in any wave- 

 length as derived from any portion of the spectrum; thus proving the 

 accuracy of the whole table. The description of the plates is as follows: 



