246 Mr. W. H. Dines. [June 19, 



At H the value was about 15 per cent, less than at G, increasing 

 uniformly from H to G. 



At H and F the values were very nearly equal, F being slightly 

 lower, but the difference was quite within the limits of an accidental 

 error. 



These results point clearly to an eddy from the frame of the 

 apparatus, and it was owing to the accidental discovery that the 

 pressure at C was 7 or 8 per cent, less than at B that I was led to 

 try the effect of placing the plate exposed normally in various parts 

 of the circle, as just explained. 



Reference to fig. 1 will show how small a surface the frame pre- 

 sents to the wind ; actually it is about 14 square inches, or only one- 

 tenth of the surface of the pressure plate, and being as much as 

 15 inches laterally from the centre of the plate, it certainly seems 

 surprising that it should exert so great an influence. 



Observations to find the value of the moment about for positions 

 I. II, III, and IV were made indiscriminately in the top and bottom 

 segments, before the existence of the eddy was suspected ; but sub- 

 sequent trial showed that the values in the case of oblique exposure 

 were symmetrical about the line AO, excepting in the neighbourhood 

 of the points B and C. 



It is clear that the disturbance due to the frame is an important 

 matter, and some attempt must be made to eliminate it. 



Since the normal pressure upon the plate was found to increase 

 uniformly from F to E, and also from H to G, it seems to be the 

 least objectionable plan to assume that the moment when the plate is 

 exposed obliquely varies in the same way. This assumption will, at 

 least, bring the position of the central line of pressure, as deduced 

 from positions I and II, more into accordance with the position which 

 it is known to take up. 



Also, since the normal pressures at H and F are practically 

 identical, and these are presumably the positions where the eddy 

 from the frame should have the least effect, I think it will be best 

 to take the values of the normal pressure in these positions as the 

 basis with which to compare all other pressures. It will be seen that 

 this is equivalent to taking about 15 per cent, less than the value 

 found at B, or about 1\ per cent, less than the value at C, as the 

 numerical value of the normal component. 



Assuming that the tangential component is zero, we have now four 

 ways of determining the normal component for any angle of incidence. 

 They are (1) by combination of the corrected values found for posi- 

 tions I and II; (2) by combination of positions II and IV. These 

 give also the position of the central line of pressure. If we take the 

 position of the central line of pressure obtained from the experi- 



