616 
MR. R. T. GLAZEBROOK ON THE REFRACTION OF PLANE 
Table IV. gives the values of <j>, D 0 -j-T, 1ST K , <p', 0 Q and the differences between 0 O 
and N e . 
Tbe last two columns give respectively the differences between tbe consecutive 
values of 0 Q and consecutive values of N E . 
The differences of 0 Q correspond to the rotation in the last column of Table II., and 
the differences of N E to the constant experimental value of that rotation, viz., 4° 6' 20". 
The last two columns, therefore, enable us to compare these experiments with those 
recorded in Table II. 
A)~M is given, i being the angle of the prism because it occurred in this form in 
the formulae, and is just as easily found at once as D 0 . 
Table V. gives the similar values for the case in which the extraordinary wave only 
is transmitted. 
The zero from which N 0 is measured is of course that from which N E has been 
measured. 
As has already been said, each of the numbers in the columns D-\-i and N is the 
mean of five observations. 
The observations of the deviation rarely differed by as much as 20", so that the 
mean is probably accurate to 5". At the lower angles of incidence that is in the 
neighbourhood of the position of minimum deviation this would produce an error of 
about 1' in the value of 0. When the angle of incidence is about 45° the error 
introduced into the value of 0 by an error of 5" in D is practically inappreciable. 
Thus the theoretical values of 0 are probably very accurate, the possible error being 
greatest when the angle of incidence is small, and then probably it is considerably less 
than 2'. The differences in the observed values of N were greater ; except in the case 
of the first two sets recorded, when owing to the high angle of incidence very little 
light got through the prism and the field was very dark—the greatest difference 
between two observed experimental values of the same quantity was 5'. The mean 
error of the observations is rather under 1' 30", so that the recorded values of N () 
and N e are probably accurate within this limit. 
For the first two values of N 0 the extreme differences among the observations were 
as much as 10', but a greater number of observations was taken. The mean error for 
them w r as 3'. In the case of the first two observations of N E the difficulty was not 
so great; the extreme differences were 4', and the mean error about 1' 30". 
We proceed now to discuss the results of the table. 
If the agreement between theory and experiment were complete, the differences 
recorded in the third column from the end in each table would be zero throughout. 
In each case as there recorded, they increase for a time as the angle of incidence 
decreases and then decrease again. We of course must remember that our numbers 
do not give absolutely the difference between theory and experiment; there is a un¬ 
known constant to be considered which we have arbitrarily determined so as to make 
• the difference in the first line of Table IV, zero, It may quite well be that we are 
