374 PROCEEDINGS OF THE AMERICAN ACADEMY. 



they remaiued invisible during a period occupied by a shifting of several 

 fringe-widths, but by observing six or seven successive disappearances, 

 values were obtained which did not differ by more than two or three 

 per cent. The average of a number of readings showed that the sodium 

 vapor shifted the two sets of fringes in the ratio of 10.5 to 11, that is, 

 fringes formed by waves of length 587487 are shifted 10.5 fringe-widths, 

 while those formed by waves of length 587513 are shifted 11 fringe- 

 widths. These numbers do not look very formidable until we translate 

 them into prismatic deviations. A prism having the same dispersion 

 giving a total deviation of only 11 degrees would seperate the two com- 

 ponents of the Zeeman triplet half a degree, or if we could construct a 

 sodium vapor prism giving the same deviation as a 60 degree glass prism, 

 two lines in the spectrum twenty-three times as close together as the D 

 lines would appear separated by a distance greater than the distance 

 between the red and bluish-green of the spectrum formed by the glass 

 prism. 



The same method was applied to a determination of the dispersion 

 between the yellow mercury lines, for the purpose of obtaining a check 

 on the curve obtained by comparing the shifts of the helium fringes with 

 fringes formed by the light from the monochromatic illuminator. A 

 prism was placed in front of the observing telescope to separate the 

 yellow fringes from the more intense green ones. On setting them 

 in motion by forming sodium vapor in the path of the light they were 

 found to disappear periodically in the same manner. It was impossible 

 to get as accurate data as were obtained with the Zeeman lines, since 

 there were moments of complete invisibility, but by counting steadily 

 it was possible to tide over the few seconds when the field was a blank 

 and get a fair estimate of the relative shifts. A number of observa- 

 tions were made, the mean of which gave 5:6 as the ratio of the 

 shifts of wave-lengths 5770 and 5790. These values fitted almost 

 exactly on the dispersion curve obtained with the monochromatic illu- 

 minator and the helium tube. Obviously a more accurate method would 

 have been to provide a device by which the light of one of the yellow 

 lines could have been cut off as desired. An attempt was made to do this 

 with the monochromatic illuminator, but the loss of light was too great. 



Some very curious results were obtained by illuminating the inter- 

 ferometer with sodium light. The fringes disappeared at the moment 

 when the thermo-element indicated a temperature of 180. It was at first 

 thought that the disappearance was due to absorption of the light by 



