168 TRANSACTIONS OF THE [ApPR. 1, 
probable accuracy of the results which are to be given later, 
several sets of observations were made upon the unpolarized 
light of a gas flame. The following illustrate about the average 
course of the readings. The zero ‘of the instrument not being 
known, the positions of equality on each side of the positions of 
extinction were determined : 
Left. Right. 
50.5 39.2 
50.0 39.1 
50.0 38.7 
50.4 38.8 
51.3 40.0 
50.0 40.0 > 
51.0 40.0 
51.1 39.9 
50.2 39.2 
49.5 SR a is 
50.4 Bia 
Since the light from a gas flame is unpolarized, the value of w 
should have been 45°. The difference is not large, but is slightly 
greater than the maximum error ascribed to the instrument by 
Cornu. The above is about an average set of readings. The 
extreme difference is 1.8°, a difference perhaps slightly greater 
than that usually found. 
A second slight error may sometimes arise in the use of this 
instrument from the fact that the two images produced by the 
double prism do not correspond to exactly the same points on 
the luminous surface. Hence, in order that the results may be 
correct, it is necessary that the adjoining portions of the incan- 
descent surface be exactly alike. In none of the experiments 
here recorded were the portions of the luminous surface produe- 
ing the two images more than 3 mm. apart. Care was always: 
taken to direct the instrument toward a portion of the surface 
which appeared to be entirely uniform. This error may, I 
think, be safely disregarded in all of the following cases except 
one, which will be mentioned later. 
A third remark which should be made upon the accuracy of 
the instrument is that observations for large amounts of polariza- 
tion are less subject to error than those made upon small amounts. 
For, since the intensities of the two images compared are pro- 
