180 L. Bell—Absolute Wave-length of Light. 
definition, but can be detected by cutting down the aperture of 
the grating till the spectrum from the abnormal portion is 
relatively bright and distinct enough to be seen. The effective 
grating space, producing the spectra on which measurements 
are made is, of course, that of the normal portion only. Both 
the gratings used in these experiments were affected by the 
above error, No. I, very slightly, No. II, somewhat more ‘seri- 
ously. Not only the discrepancies between different gratings, 
but those between different orders of spectra in the same grat- 
ing are due to this cause. [or while in one order, where the 
effect due to the abnormal portion is imperceptible, the spec- 
trum as measured is produced by the effective grating space 
alone, in another order there may be produced a slight shading 
- off of the lines so that their apparent centers may correspond 
approximately to the average grating space. In any case, it is 
quite clear that a combination of the results from different 
orders of spectra will-not eliminate the error. 
The remedy lies either in stopping out the imperfect portion 
of the grating, or measuring it and introducing g correction. 
As the work of angular measurements was nearly finished before 
the study of the gratings was begun owing to a delay in getting 
apparatus, the latter course was adopted in these experiments. 
Each grating was examined in detail, and the relation of the 
grating spaces in the various portions of it carefully deter- 
mined. From these data a simple graphical method gave the 
correction to be applied to the wave length. In each grating 
the fault was confined to a small portion, and as the order of 
spectrum employed in each was selected on account of its good 
definition and freedom from anything lke haziness or shading 
of the lines, it seems safe to assume that the abnormal portion 
produced no visible effect and that consequently the correction — 
above mentioned counteracts the error quite effectually.. In 
erating I the correction was one part in 300,000, and in grating 
II one part in 60,000. Applying these to the wave lengths we 
have for grating I, 
Wiaveilength (055 siesta epee 5896-11 
CO ER EC EVOTTG eh ee ON ee cae ttt prea a ee 02 
Gorrectediwslit eas) ee ee 589609 
And for grating II, 
Wavelength. 420 3) ein aes 5895°95 
@orrections 2 4630 eae SECO) 
Corrected -w. lose 2k ee es 5896°05 
Combining these and giving to Grating I the greater weight 
