ASTRONOMY: C. E. ST. JOHN 
229 
Mount Wilson by 0.011 A, 0.007 A, and 0.004 A, respectively. These 
systematic differences furnish new ground for not adopting any form of 
operator for transforming the wave-lengths in Rowland's Preliminary 
Table to the international standard, and indicate some possible pitfalls 
incident to statistical methods of comparison. 
Though the Rowland wave-lengths cannot be transformed so as to 
reproduce rigorously the wave-lengths expressed in the international 
system, it does not follow that the Preliminary Table of Solar Wave- 
lengths has outHved its usefulness. It will be some years before com- 
plete tables of solar wave-lengths, based upon the new standards, will 
be available, and even then the Rowland table will serve indefinitely 
as a reference for other things than wave-lengths. It has been assumed 
that the accidental errors in it are considerably greater than =±=0.01 A, 
but recent comparisons between certain lines in Rowland's table and 
their wave-lengths as measured here upon plates of a very high dis- 
persion, 1 mm. = 0.3 angstrom, using as standards neighboring free- 
standing lines, show that this is an over-estimate. For 54 lines in pairs 
with separations between 0.25 and 0.50 angstrom, the mean variation 
from Rowland's values is =±=0.003 angstrom. As more than 200 lines 
were used as standards, it appears that, for the types of Hues involved, 
the accidental errors in the Rowland wave-lengths are much less than 
0.01 angstrom. It is probable, as Frost and Adams remark,^ that 
errors of this magnitude occur but rarely and mainly then for lines 
whose measurement is inherently difficult, such as the very weak and 
the strong shaded lines. When it was first noticed that the differences 
between the Rowland wave-lengths in the Preliminary Table and those 
found by the newer arc determinations were not constant over even a 
short spectral region, it appears to have been assumed that the large 
variations were due to errors in the Rowland values, but data similar 
to those reported in Tables I, II, and III show that the apparent dis- 
crepancies represent real differences in the behavior of lines and tend to 
increase rather than destroy confidence in the accuracy of the relative 
wave-lengths of lines not presenting special difficulties of measurement 
and sufficiently separated from others to be within the power of spectro- 
graphs in common use. 
' Ch. Fabry and A. Perot, Astroph. J., 15, 272 (1902). 
2 J. Hartmann, Astroph. J., 18, 167 (1908). 
3 Sebastian Albrecht, Astroph. J., 41, 333 (1915). 
^ Charles E. St. John, Mt. Wilson Contr., No. 93, p. 35; Astroph. J., 41, 63 (1915). 
5 Charles E. St. John and L. W. Ware, Mt. Wilson Contr., No. 61, pp. 31-32; Astroph. /., 
36, 45-46 (1912). 
^ E. B. Frost and W. S. Adams. Publications of Yerkes Observatory, vol. 2, p. 155. 
