246 PROCEEDINGS OF THE AMERICAN ACADEMY 



discussion of the results. The measures were made with the microme- 

 ter described iu the Annals of Harvard College Observatory, VIII. 42. 

 The photograph was moved under a microscope by a micrometer screw 

 having a pitch of t?'^ of an inch. By turning the screw, the lines were 

 brought to coincide with the end of a hair in the field of view of the 

 microscope, and the jjosition was determined by the divided head of the 

 screw. As is usual when examining photographs, very low powers 

 were essential, or niucli of tiie finer detail would be lost. The best 

 results were attained wiili an eyepiece and objective, each equivalent 

 to a lens of about 2 indies focus, giving when combined a magnifying 

 power of about fifteen diameters. The luicertainty in the position of 

 the lines of the photograph greatly exceeded the errors of setting and 

 the irregularities of the screw. To establish this point a measurement 

 was made of the divisions of a glass scale constructed by Professor 

 Rogers, in which the errors were wholly inappreciable with so low a 

 magnifying power. Ten divisions were compared, each of which was 

 T7^5 of an inch, and equalled -jJ^ of a revolution of the screw. The 

 results gave an average deviation of .0015 of a revolution, mainly due 

 to a slight eccentricity of the head of the screw. This eccentricity 

 amounted to about .002 of a revolution. Twelve divisions, each equal 

 to t}^ of an inch (one turn of the screw), gave an average deviation 

 of .002 of a revolution. These errors might be still further reduced 

 if required, by taking additional precautions ; but they are entirely 

 insensible compared with the uncertainty in the position of the lines in 

 the photograph. They correspond in fact to only about .02 of a ten- 

 millionth of a millimeter in the wa^e-length, Throuirhout the work, 

 all the settings were made by turning the screw in the direction in 

 which the readings increase. 



The reduc'tion of the measures to wave-lengths was greatly aided 

 by the fact that but little change was made in the spectroscopic appa- 

 ratus throughout the entire investigation, and that especial care was 

 taken to .«ecure stiffness in its construction. A sinirle curve could 

 thus be used for all the i-eductions, at least jDrovisionally. On the 

 other hand, it does not seem safe to assume that, when two spectra are 

 photograjjlied side by .side on a plate, the lines in each will have the 

 same position. 



In some of the j)hotographs, especially in I'latcs Go and {]'), there is 

 a perceptible want of coincidence between lines wliich are doubtless in 

 reality of identical origin. No such deviation could be caused by any 

 fiexui-e of tlie telescope, l>ut it might be due to flexui-e of ilie spectro- 

 scope when directed towards objects at different altitudes, or to change 



