276 
Proceedings of the Royal Society of Edinburgh. [Sess. 
never unscrewed, so as to ensure that the zeros would remain constant 
during the subsequent measurements. 
At each selected point in the spectrum nine settings of the Nicol were 
made, and the mean taken. As regards the accuracy of these measurements 
two considerations come in. 
A large rotation makes the observer’s error in setting the Nicol pro- 
portionally of less account. But then, on the other hand, a large rotation is 
X = 465 
X = 487 
X = 515 
x = 650-5 
X = 698 6 
X = 666 
(0) (?) (i) (i) Concentration (|) (1) 
Fig. 3. 
usually accompanied by a large dispersion, which has the effect of somewhat 
increasing the observer’s error. 
I have accordingly calculated the probable error of the settings of the 
Nicol in two different cases. In the first of these the rotation dispersion 
was zero, the polarimeter tube containing water only; in the second it 
was large. The results are respectively (f'458 and 2 /, 46. The latter 
probable error is that obtained from the nine settings made at A = 515 on 
the strongest of the solutions. As the measured rotation in this case was 
