ON THE TOTAL SOLAR ECLIPSE OF AUGUST 20, 188G. 
Assuming the wave-lengths of F and of H to be given, we can, from the observed 
positirais, calculate by some interpolation formula the wave-lengths of the intervening 
reference lines. The formula I used is the usual one 
where and X 3 are the two known wave-lengths, X.^ that to be determined. D is the 
distance measured on the plate between X^ and X 3 , a tliat between X^ and X^. The 
wave-lengths thus found are approximate only; but, by comparing the calculated 
O 
values with those found in Angstrom’s map and given above, we can construct a tal)le 
of corrections for the reference lines. A curve was drawn having the calulatecl wave¬ 
lengths as abscissae and the corrections as ordinates. This curve was found to be 
quite regular in shape. The same interpolation formula was now used to calculate 
approximate wave-lengths for the corona lines, and the corrections were read off from 
the curve. I have found this combination of interpolation by calculation and by a 
graphical method to be very convenient. In order to show the accuracy which may 
be obtained in this way I give the following example :— 
I left out originally the reference line 46G6'5 altogether. There was, therefore, a 
very large gap between the lines F and Hy. The curve had to be drawn, more or 
less, as a straight line between the points corresponding to the two lines. Treating the 
reference line, then, first like an ordinary corona line, that is, finding its approximate 
wave-length by calculation and the correction from the curve, I found 4667'0 as its 
wave-length, being very near the correct value. The curvm was improved by making 
it still further agree with this line, and, as there is no gap of equal magnitude in other 
parts of the spectrum, I concluded that now the errors due to intej-polation are 
negligible. It is more difficult to give an idea of the possible error due to faulty 
measurement. Most of the important lines were measured several times, and, from 
the way in which the wave-lengths agree, I should say that an error of 1‘5 near F and 
1 near H in X* metres will not often occur in the stronger lines. The position of 
the weaker lines is, of course, more uncertain. 
As the strong corona line 4232'8 had to be used as the starting point for the 
measurements of the second plate, considerable trouble was taken to find its position. 
Perhaps the best idea of the accuracy which can be reached in such measurements 
can be got from a comparison of the individual measurements. The following then 
are in centimetres, the distances from FI of the corona line and of the calcium line 
4226, as measured on the corona plate and on the reference spectrum respectively. 
The measurements were taken at different times during the last two years. Those 
marked■“ were taken by my assistant, Mr. Arthur Stanton. 
