558 Mr. H. T. Jessop on Cornus Method of 



values of s as large as possible were used. The number of 

 fringes observed on each side of the centre being 2s, the 



quantity - (A n+S 2 — A ?i 2 ) was calculated for values of n ranging 



from 1 to s. A series of values so obtained will of course 

 overlap, but any progressive change shown by such a series 

 must indicate a progressive change in the mean curvature. 



In cases where the number of fringes has rendered it 

 possible, a set of values of £(A )i+2 2 — A,, 2 ) for totally distinct 

 sets of fringes has been calculated, the result given in such a 

 case (e.g. col. 7 in Table 1(a)) being the mean value of 

 -|(A„ +2 2 — A ?l 2 ) for two consecutive values of n, and indicating 

 the mean Radius of Curvature over 4 fringes. 



Errors of Measurement. 



Several readings were taken of the position of' each fringe 

 and their mean value used. The probable error in measure- 

 ment varied with the breadth of the fringe, ranging from 

 01cm. for the most diffuse centre fringes to -001cm. for 

 well-defined fringes near the edge of the beam. The influence 

 of a given error upon the value of A n 2 , however, also varied 

 with the value of A n , being least where the error was greatest, 

 giving a fairly constant error over the range of fringes 

 measured . -. 



The probable error in the values of -(A n+S 2 — A n 2 ) given 



in the tables is estimated at *03 x - . 



s 



In the case of the longitudinal fringes the measurements 



were much more accurate, and the error would not amount 



to more than half that for the transverse fringes. 



Tables I. (a) and I. (b). Glass I. a. 

 Width of beam = 5*15crn. Thickness = '571 cm. 

 Bending nioment = 43'2 kg. cm. 



Table I. (a) shows the radii of transverse curvature 8 minutes 

 and 100 minutes after loading. 



In column 6 the mean values over 6 fringes show an increase 

 with time, but that this does not result from an increase at all 

 points on the transverse axis is shown by the values in column 7. 

 These latter are subject to greater error (approx. -015) than the 

 former, but they clearly show a decrease in radius of curvature 

 for the inner and outer fringes, and a marked increase for the 

 intermediate ones. This flattening out in the region of the 

 intermediate fringes may very well be due to the effects Ol 

 the knife-edges, which in this case were little more than twice 

 the width of the beam apart. 



