460 
ASTRONOMY: C. E. ST. JOHN 
of error, remove any direct evidence that anomalous dispersion contri- 
butes to the production of the Fraunhofer Hues. 
As the Mount Wilson data for sun-arc displacements and for the 
comparative separation between close pairs of lines insolarand terrestrial 
spectra do not show the effect of mutual influence, it seems necessary 
to assume either that the Mount Wilson data are affected by systematic 
errors just sufficient to annihilate the effect of anomalous dispersion, 
or that the Rowland wave-lengths for lines in close pairs are systematic- 
ally in error; a slight over-separation of such pairs would introduce an 
effect of the sign indicated by the theory. The results of an investi- 
gation upon ''The Accuracy of the Measured Separations of Close Solar 
Pairs; Systematic Errors in the Rowland Table for such Lines" are given 
in a Contribution from this Observatory."^ For pairs consisting of lines 
of intensities 3 and 4, with mean separations of 0.274, 0.145 and 
0.075 A, the Rowlam reparations exceed the Mount Wilson values by 
0.003, 0.008 and 0.0. ^ A, respectively. As errors of this sign and 
magnitude would account for the deviations found by Albrecht, an 
exhaustive examination of the cases included in Albrecht's list was 
undertaken; details of this investigation will appear in a Contribution 
from this Observatory. 
The wave-lengths of the 104 lines used by Albrecht have been re- 
ferred to those of neighboring free-standing lines. The measurements 
have been made by two observers upon spectrograms of a scale and 
dispersion that previous experience had shown were best adapted to 
each case. The errors found in the Rowland values are systematic and 
of a sign and magnitude corresponding to the Albrecht deviations. 
This correspondence is marked, large and small values of the one being 
associated with large and small values of the other. The coefficient 
of correlation between the two sets of observations is -f 0.55 ±0.05, 
indicating that the correlation is a practical certainty. No explanation 
of this correlation seems possible other than that the errors in the Row- 
land wave-lengths and the Albrecht deviations are two phases of the 
same phenomenon, that in fact the deivations observed by Albrecht 
are a measurement of the Rowland errors. 
Summary and Conclusion. — 1. The violet and red components of 
close pairs of solar lines show the same displacement as isolated lines 
when compared with the spectrum of the arc. 
2. The mean separation of close pairs in the solar spectrum is the 
same as that determined from terrestrial sources whether the component 
lines are due to the same or different elements. 
3. The Rowland wave-lengths for close pairs of solar Hues are sys- 
