16 PROCEEDINGS OP THE AMERICAN ACADEMY 



lines carefully noted according to their wave-lengths. In order to 

 eliminate any personal error, they are examined by both observers 

 separately, and their results afterwards compared. 



To eliminate errors arising from suspected impurities of materials, 

 as also from the impurities known to exist in the carbons employed, 

 we took what we called " comparison photographs." For these, we 

 placed in the carbon cup a portion of the substances known or sus- 

 pected to be present as impurities in our metal, and then photographed 

 the spectrum thus given on the upper half of the plate ; a piece of 

 the metal under experiment was then placed in the lamp, and the 

 spectrum photographed on the lower part of the plate. Any lines 

 due to impurities would then extend entirely across the plate, while 

 those of the pure metal would extend only half-way. In addition to 

 this precaution we consulted all accessible tables and plates as to the 

 position of known lines of metallic spectra, and also compared together 

 all our photographs of the same region. If all of these tests left any 

 doubt as to the origin of a given line, it was at once subjected to 

 special investigation until all doubt was removed. 



The dispersion given by the apparatus in the order of spectrum in 

 which we worked is such that a single wave-length occupies on the 

 negative a space of 1.12 mm. This makes the distance between the 

 lines Dj and D 2 6.7 mm., while the length of spectrum from A to H 

 is about 4.1 m. With so great dispersion it would hardly be possible 

 to mistake the position of a line by any very considerable amount, or 

 to confound neighboring lines belonging to different metals. 



For reasons readily apparent, it was found so difficult to photograph 

 under high dispersive power those parts of the spectrum not lying 

 between wave-length 3600 and wave-length 5000, that our photo- 

 graphic work was done chiefly within those limits. It was, however, 

 supplemented in many cases by eye observations in other portions of 

 the spectrum. 



We are convinced that there is much in the whole matter of coinci- 

 dences of metallic and solar lines that needs re-examination ; that 

 something more than the mere coincidence of two or three lines out of 

 many is necessary to establish even the probability of the presence of 

 a metal in the sun. With the best instruments the violet portion 

 of the solar spectrum is found to be so thickly set with fine lines, that, 

 if a metallic line were projected upon it at random, in many places 

 the chances for a coincidence would be even, and coincidences could 

 not fail to occur in case of such metals as cerium and vanadium 

 which give hundreds of lines in the arc. 



