OF SPECTRA AT THE SOLAR ECLIPSE OF JANUARY 22, 1898. 
385 
H and F about twenty-six lines were used, which are, however, not so well defined or 
so satisfactory as standards as the others. 
In constructing a scale of wave-lengths for the reduction of the bright-line spectra 
(Nos. 3 and 4), the best defined known lines between D and FI in these spectra were 
also selected as standards, and used together with the twenty-six cusp spectrum lines. 
But in the ultra-violet the scale depends entirely upon the standards of the cusp 
S23ectrum, the whole of the lines in this region in Nos. 3 and 4 (including the 
hydrogen lines) being treated as unknown. A careful comparison made between 
measures of known lines in Nos. 3 and 4 spectra and the mean measures of the two 
cusp spectra shows that it is safe to assume that all the spectra obtained before mid- 
eclipse may be considered to be identical as regards dispersion. 
A sufticient number of standards well distributed throughout the entire spectrum 
being thus obtained, the approximate relation between scale reading and wave-length 
at all points in the spectrum was next determined by graphical methods, and from the 
interjDolation curve obtained it was easy to compute the scale reading corresponding 
to a number of definite points in the spectrum separated by equal intervals of wave¬ 
length. 
The scale reading corresponding to each 50 tenth-metres of wave-length was in this 
way computed, taking in each case the mean result given by three or four standard 
lines situated near ; and the values thus obtained were finally slightly corrected 
by smoothing difierences until fifth differences were made to increase in a regular 
progression. 
The smoothed values in no case differ from the values derived from the measures Ijy 
quantities greater than the uncei'tainties of the measures themselves. The mean 
difference in the best defined portion of the spectrum X 4100 to X 3400 is ‘OOIG millim.. 
and the greatest difference 4)05 millim., the latter corresponding to T5 tenth-metre 
at X 4100 and '07 tenth-metre at X 3400. 
In the visible spectrum the definition under the microscope is very poor in all the 
negatives except No. 4, and the measures in consequence are not very consistent, l)ut 
for the sake of completeness every line has been measured, and an accurate scale of 
wave-lengths constructed for the entire range of spectrum photographed. 
From the accordance between the different sets of measures of the bright lines in 
No. 3 spectrum the probable errors of the wave-length determinations given in 
Table I. (p, 404 et seq.) would appear to he about '15 tenth-metre at X 4000-, decreasing 
to '05 tenth-metre at X 3400. 
The Brlght-Ii)ie Spectra. 
Spectrum No. 3 (Plate 11, fig. 1).—This negative was exposed at the moment when 
the dark lines of the Fra\mhofer spectrum were seen to disappear and innumeraljle 
bright lines flashed out across tlie strip of continuous spectrum which still remained 
visible. 
3 D 
VOL. cxcvii.—A. 
