489 
of the double refraction in strained glass. 
weighted mean was obtained, thus y® — yW was given double 
weight and y®—y m triple weight, since the one represents two 
and the other three fringe intervals. This gave a good determina- 
tion of a mean fringe interval 2/ 0 (w+1) — y 0 {n) in each case. The 
result was then divided by the calculated wave-length. 
In calculating the latter a small correction has sometimes had 
to be put in on account of the fact that the spectrum lines were 
not, in all cases, perfectly straight, owing to spherical aberration 
in the lenses. In most cases this correction was too small to be 
applied. When sensible it was applied, so that the wave-lengths 
given in the table below are corrected. 
The results obtained from the first two days’ observations are 
shown in the table (page 488). 
These show a steady increase of G from the red to the violet. 
In other words Wertheim’s law does not hold, but the relative 
retardation in centimetres in air decreases as the wave-length 
increases. 
The total variation is between 6 and 7 per cent, and could not, 
it seems, be accounted for by errors of observation. The fringes 
as seen by means of a powerful arc light were sharp and well 
defined, easily measurable to tenths of a millimetre division. The 
mean value of y 0 {n+1) — y 0 [n) should be accurate to the second place, 
though even an error of one unit there would not account for the 
difference. 
The experiments were repeated a large number of times, with 
a view to eliminating the accidental errors. The results are ex- 
hibited in the table (page 490). 
The sets taken on Feb. 24 and 25, those on Feb. 28 (II), of 
March 5 and that of March 6 are from measures taken on the 
fringes in the upper part of the beam (under tension). The 
observations of March 1, 3, 4, 8, 10 and (I) of March 5 are made 
on the fringes visible in the lower part of the beam (under com- 
pression). 
The latter do not indicate such a large variation of C as do the 
other observations. Those on March 1 and March 4 almost con- 
firm Wertheim’s law. But those of March 3 show a slight varia- 
tion, tending to make C increase as \ diminishes ; those of March 5 
and 8 show a strong variation and those of March 10 a moderate 
variation in this sense. 
All the results corresponding to measurements taken upon 
fringes in the upper half of the beam decidedly confirm this 
variation. 
