MOUNT WILSON SOLAR OBSERVATORY. 



241 



the part of the measurer toward increased separation. This seems to 

 be an effect of contrast, the maximmn being located nearer the free 

 edge of the hne, for which the contrast is greater. 



For close pairs to the red of X4000, component intensities 3 and 4, 

 with mean separations of 0.276, 0.145, and 0.75 A, the Rowland separa- 

 tions exceed the Mount Wilson values systematically by +0.003, 

 +0.008, and +0.013 A, respectively. That these differences are 

 errors in the Rowland values is made probable by the agreement 

 between the results of the diverse methods used at Mount Wilson, the 

 concordance between the spectrograms of different orders, and the 

 established tendency toward over-separation with any departure 

 from best conditions. 



A FURTHER TEST OF THE ANOMALOUS DISPERSION THEORY. 



A mutual influence of closely adjacent Fraunhofer lines which dis- 

 places them in opposite directions, a quasi-repulsion, is a deduction 

 from the anomalous dispersion theory. An observational confiniiation 

 of such an influence would be a direct proof of the efficiency of anom- 

 alous dispersion in producing solar phenomena; its value as a criterion 

 is evident. For solar lines with companions to the violet the sun-arc 

 displacements should be greater than the mean for lines of the same 

 pressure group, and less for solar lines with companions to the red; 

 further, the separation of the adjacent lines should be greater in the 

 solar spectrum than in arc spectra. 



Mr. St. John has tested these points, with the results for the sun-arc 

 displacements shown in the accompanying table, while for 45 pairs the 

 mean separation in the solar spectrum, within the limits of error, is 

 identical with that in arc spectra. In neither case, therefore, do the 

 results support the anomalous dispei-sion theory. The behavior of 

 lines with companions is like that of similar free-standing hnes. 

 Whether the separation in the solar spectrum is greater or less than in 

 arc spectra depends upon the configuration of the pair. For 8 of the 

 45 pairs it should be larger and for 15 smaller than in arc spectra. 

 Actually the differences are 0.005 and 0.0035 A, respectively, in the 

 required directions. 



Comparison between the Rowland and International wave-lengths 

 of the iron hnes shows apparent displacements of the sign required 

 by the hypothesis of nmtual influence (Albrecht). As the previous 



