372 PROCEEDINGS OF THE AMERICAN ACADEMY. 



are invisible at the beginning and end of the experiment, only coming 

 into view vphen a considerable quantity of sodium vapor is present. The 

 same thing is true to a less extent with helium light, the fringes disap- 

 pearing after a shift of about two hundred fringe-widths has taken place. 

 By increasing the length of the air path, either by turniiig the screw of the 

 instrument, or introducing one or more thick plates of plane-parallel 

 glass, they again come into view, and the counting can be continued. 

 The latter method was preferred, as the glass plate could be dropped into 

 place in a moment without interrupting the count. This trick was only 

 necessary in making the absolute determinations with helium light, as in 

 the comparison work the vapor never attained such a density as to cause 

 the helium fringes to disap[)ear. When working close to the helium line 

 with the monochromatic illuminator it was necessary to work with very 

 narrow slits, otherwise the enormous dispersion of the vapor obliterated 

 the fringes almost immediately, in spite of all precautions. It was im- 

 possible to approach closer to the absorption band than the helium line, 

 owing to this trouble, and ratios of 1 : 2 are about as large as we feel 

 much confidence in, though (5n one or two occasions we set the monochro- 

 matic illuminator on the D3 line, and obtained equal shifts in the two 

 systems. It was impossible to observe a shift of more than one or two 

 fringes in this case, before the system disappeared. The most accurate 

 determinations are doubtless those where the ratio is between 1 : 3 

 and 1 : 15. 



In the blue-green region the ratio is in the neighborhood of 1 : 40, 

 and the values are not as accurate, as an accidental displacement of half 

 a fringe-width, which may easily occur as the result of a slight change in 

 the temperature of the instrument or the air, exercises a much greater 

 effect on the results. On tiiis account the determinations made by the 

 prismatic method are to be considered more reliable in this region of the 

 spectrum. 



DeTICRMINATION OB' THE DiSPEUSION IN THE IMMEDIATE 



Vicinity of the D Lines. 



The extraordinarily high value of the dispersion close to the absorption 

 band makes it impossible to sift out from the spectrum light sufficiently 

 homogeneous to yield interference fringes, when even a very small 

 amount of sodium vapor is introduced into the path of the light. It 

 was found difficult to get satisfactory values with the light from the 

 monochromatic illuminator much closer to Dg than the distance se[)a- 

 ratiii"' D., from D^, the fringes becoming blurred as soon as the smallest 



