of the double refraction in strained glass. 311) 



Now the width of the slit used in the ordinary experiments 

 was 2"5 mm., so that fc=l , 25; 



•'• P = ~m = "go" = ' 02 approximately, 



w i 1 ~ \ z ^) < * ( ' 0004) I 1 + t) < ' 0008 ( numericall y)> 



and is certainly negligible. 



It is to be noted that this was a case where the stress was far 

 from uniform. It was certainly less uniform than in the average 

 observation. It seems reasonable therefore to assume that no 

 sensible part of the error is contributed by this cause. 



6. There remain two causes of error, (a) the possibility that 

 the stress distribution near the middle of the block is not exactly 

 of the form T + Sy; (b) the fact that the vertical centre-line of 

 the slit used (which had to be shifted for each experiment) was 

 not accurately opposite the vertical axis of the block. 



Theoretical considerations, developed in another paper (" On 

 an approximate solution for the bending of a beam of rectangular 

 cross-section under any system of load," Phil. Travis. A. Vol. 201, 

 pp. 63 — 155), make it unlikely that, at a distance from the points 

 of application of stress equal to about twice the breadth, the 

 influence of local perturbations should have been sensible. In 

 the experiments under consideration, this condition was nearly 

 satisfied for rays passing through the centre of the block, the 

 height being four times the breadth. There is still the possibility 

 that another type of stress may have made itself felt, for instance, 

 torsion about a vertical axis. Such torsion might have been 

 caused by friction against the compressing planes, owing to some 

 want of symmetry in the application of the load. But such want 

 of symmetry could hardly have caused a torsion couple sufficient 

 to affect the observations. 



(b) then remains as the probable cause of the discrepancies. 

 It was impossible to adjust the diaphragm so that its axis was 

 exactly central. If there was a small error e in the position of 

 the diaphragm, the stress corresponding to the observed black 

 band was no longer T but T + Se. Now S was not simply pro- 

 portional to the total load ; this followed clearly from the experi- 

 ments mentioned in § 3 above. Accordingly the stress varied 

 according to some unknown law, and although this variation was 

 probably small as a whole, yet it may account for the discrepancies 

 observed in the curves deduced from observations of different 

 sets. 



