460 
MR. J. EVERSHED ON THE SOLAR ECLIPSE OF 1900, MAY 28. 
portion of the screw, but with the negative reversed end for end. A comparison of 
the direct and reversed measures revealed systematic differences amounting to as 
much as ’02 millim. in a run of 100 millims. 
Although this error of run would have very little effect on the resulting wave¬ 
lengths, which depend ultimately on short measured distances from known lines, it 
was considered more satisfactory to measure the photographs in three sections of 
about 70 millims. each, selecting a portion of the screw which gave consistent results 
over this range. In order to reduce the accidental errors of setting, and to detect 
blunders, each section was measured twice, one set of measures with the red end to 
the right, and the other with the red end to the left. 
This method involved some extra labour in combining the measures, and in joining- 
up the three sections into one consistent whole by means of the lines which 
overlapped between the sections. However, the definition of some of the images is 
so good that any amount of trouble taken in getting satisfactory measures seemed to 
be justified. 
The relation between wave-length and measured distances at all points in the 
spectrum was determined approximately by graphical methods, using 42 well-known 
lines, including lines of hydrogen, calcium, titanium and iron, &c. A large number 
of the finer lines were then identified with certainty, and in the final reduction the 
broad over-exposed hydrogen and calcium lines were rejected as standards, and about 
65 more suitable lines were selected which are well distributed throughout the 
spectrum, using in the ultra-violet region lines which I considered thoroughly well 
identified in the spectra obtained in 1898.^' 
From the standard lines the position in millimetres of each 50 tenth-metre of wave¬ 
length was computed, taking the mean value given by four or five of the nearest 
standards in each case. A table of differences was then made giving the inter¬ 
mediate values by interpolation and the value in millimetres of one tenth-metre at 
every 25 units. 
The wave-lengths of all the lines, including the standards, were computed from 
this table, using second differences. 
Each of the three spectra measured was reduced independently, using the same 
standard lines, but computing a separate table for each. A direct comparison of the 
three sets of measures showed that they were very nearly identical, and one table 
might have served for all. But there appear small systematic differences, due in part 
the fact that the measures were made at different distances from the centre of the 
arcs, and probably in part also to slight irregular contraction of the photographic 
films in drying. 
It was therefore considered more satisfactory to treat each sjDectrum entirely 
independently, combining in the end the wave-length values obtained to arrive at the 
most probable values measured on all three spectra. 
* ‘ Phil. Trans.,’ A, vol. 197. 
