174 L. Bell—Absolute Wave-length of Light. 
these experiments a sort of compromise was effected. A small 
thermometer was attached to the thin metallic slip that held 
the edge of the grating, and shielded by cotton from air cur- 
rents which of course would affect it much more than they 
would the grating. The thermometer was a small Fahrenheit 
graduated to quarter degrees and quite sensitive. It was care- 
fully compared, throughout the range of temperatures em- 
ployed, with thermometer Baudin 7312, which served as a 
standard in all the measurements regular and linear, and during 
part of the time was placed directly over the grating to give a 
check on the attached thermometer. This expedient was finally 
abandoned as unlikely to help the matter much. 
The corrections for temperature were deduced from the 
assumed coefficient of expansion of glass, which was taken as 
0:0000085. This was reduced to angular correction for the 
approximate value of g and applied directly to the observed 
angles. Since the temperatures at which observations were 
made varied little from 20° C. and were quite equally distribu- 
ted on both sides of that figure any error in the assumed coefli- 
cient would hardly affect the average result, but would appear, 
if at all, as a slight increase in the probable error. 
760™™ (reduced) was taken as standard pressure and the 
values for the days of observation were taken from the U.S. 
Signal Service observations for the hours of 11 A. M. and 3 P. M. 
on those days. ‘The average for the measurements made with 
grating I was 761™™, and for those with grating II 760™, so 
that corrections for pressure were uncalled for. 
The effect of the velocity of the apparatus through space is 
a subject concerning which there has been much discussion. 
Angstrém deduced acorrection, but van der Willigen in‘ quite 
a lengthy discussion of the whole matter came to the conclu- 
sion that there was no error due to the above cause. .Since 
that period the question has been raised from time to time, but 
no decisive investigations on the subject have yet been pub- 
lished. At present however it seems to be tolerably well set- 
tled that no correction is needed, as the error, if there be any, 
is of an order of magnitude entirely negligible, and in the 
present paper none has been applied. 
The angular measurements, after all corrections were applied, 
may thus be regarded as determined with a high degree of 
accuracy—most probably to less than one part in half a million. 
MEASUREMENT OF THE GRATINGS. 
The exact determination of the grating space is by far the 
most difficult portion of a research on absolute wave length, 
and has been uniformly the: most fruitful source of errors. 
