86 
PROFESSOR W. 1ST. HARTLEY AND MR. W. E. ADEREY 
comparable one with the other, he had to make a correction of — 1 '0 tenth-metre to 
those obtained by the latter means. An explanation of this will be found in M. 
Cornu’s paper. 
Third .—The error arising from the adjustment of the grating to its position for 
photographing the spectrum. This does not exceed 0"5 tenth-metre, and, it should be 
noted, appears in all the measurements of the plate. This may be eliminated by 
taking the mean of the measurements of the cadmium lines. 
Fourth .—The error in the measurements arising from change of temperature 
affecting the dividing engine. This is within ± 0'2 for well defined lines, and it is- 
always corrected by the cadmium lines. The errors of observation in the measuring 
of lines are also, for well-defined lines, at most i 0'2. 
The first and second errors are of opposite sign for the spectra, to the right and left 
of the regular reflection of the slit, and are therefore easily eliminated by taking the 
mean of the spectra on the two sides. 
The wave-lengths given in the accompanying tables are all comparable with one 
another. If an error occurs in any of the values for the cadmium lines, it will be 
common to all the lines in that part of the spectrum in which the particular line 
occurs, and will be easily eliminated. If the wave-lengths assigned to the cadmium 
lines are correct, there are only two errors to which the wave-lengths in the tables 
that have been calculated from grating spectra are liable. These are that incurred in 
the measurement of the lines by the microscope and dividing engine, and that due to 
the interpolation curve. The errors of measurement mentioned above are, for well- 
defined lines, in no case greater than f o^ oth of an inch ; this, in terms of wave-length, 
equals 0’2, 0T7, and 0"12 tenth-metres for the portions 6-12, 11-18, and 17-26, 
respectively. The error of the interpolation curve is not more, we believe, than 0"1 
tenth-metre. The maximum for well-defined lines probably does not exceed f 0'3 
tenth-metre. In the case of faint lines, the general error is larger, but it seldom rises 
to more than 0'5. Thus in the accompanying table of air lines, the wave-lengths of 
which were all determined directly by measuring diffraction spectra, two sets of 
numbers taken from different photographs are given, and the numbers for the following- 
six, which are all very feeble and diffuse lines, and therefore difficult to measure, are 
the only ones differing by more than the general error :— 
Aluminium plate. 
Copper plate. 
Portion 6-12. 
Portion 6-12. 
4402-0 
4403-1 
4215*9 
4217-1 
4025-9 
4024-7 
3851-0 
3850-0 
3842-2 
3841-2 
3325-3 
3324-1 
