330 Mr. J. ©. M‘Connel’s Notes on 
So far the matter is tolerably simple, but if the Nicol 
produce deviation of the light complications are introduced. 
Fig. 7. 
B 
The figure represents a flat-ended Nicol B traversed by a 
ray of light PL MP,. The argument, however, is perfectly 
general. A small error is due to the Nicol C having to deal 
first with light parallel to MP, and afterwards with light 
parallel to PP’. This may be eliminated by taking two 
readings, the Nicol B in the interval having been turned 
through two right angles or thereabouts about LM. The 
mean thus found gives us the position of the plane of polari- 
zation of P L, which is most nearly parallel to the plane of 
polarization of MP,. We know nothing about the plane of 
polarization of light which issues parallel to MP’. Clearly _ 
therefore, when our Nicol is mounted on the spectrometer- 
table, we ought not to turn it exactly through 180°, but we 
ought to turn it till the incident light is related to the face 
M in the same way as P, M was at first. 
We have then this practical rule. Place the auxiliary 
Nicol on the spectrometer in such a position that the deviation 
Jies wholly in the plane of rotation. To get the second 
reading, turn the table in the direction of the deviation through 
an angle equal to 150° less the deviation. Correct the mean 
of the two readings by half the rotation-angle as before. 
Thus we are able to get an accurate value of the zero- 
reading in spite of the dissymmetry of the Nicol and the 
deviation of the light, provided only the two faces of the 
Nicol are good planes. We have been compelled, however, 
to limit ourselves to first approximations. On this point it is 
to be noticed that the outstanding error is due to two causes, 
the inaccuracy of adjustment of the auxiliary Nicol, and the 
dissymmetry of the Nicol itself. The latter we may assume 
to be very small; while the former has very slight eftect. 
It only comes in as a secondary cause, for if the dissymmetry 
were removed the outstanding error would be zero. Hence it 
is not necessary in general to take elaborate precautions about 
the setting of the Nicol in the determination either of the 
zero-reading or of the rotation-angle. If great accuracy be 
