252 Mr. C. Spurge. On the Effect of Polish on the 



Table VI. 



w. 



tan w. 



cos p. 



P- 



I. 



o 



17 



39 



5 



0-3184 



0-14595 



81 



36 



47-296° 



17 



43 









U 14Uoo 



81 



55 



46 ' 983 



17 



52 







3223 



0-14797 



81 



30 



47 -007 



17 



57 







0-3239 



0-15006 



81 



22 



47 132 



17 



46 



5 



0-3206 



-14814 



81 



29 



47 -232 



17 



50 



5 



0-32L8 



0-14768 



81 



30 



47 -J 12 



17 



52 







3224 



15355 



81 



10 



47 • 424 



17 



38 



5 



0-3181 



• L4999 



81 



22 



47-317 



From Table V we may find with, what accuracy we can work with 

 the elliptic analyser. 



The greatest difference between the mean and a single observation 

 is 0"003, so that the error of determination of tan w is less than a per- 

 centage. The mean Talne of I is 47*188°, and the greatest error 0-236° 

 or 14'. The mean value of />is 81° 29', and the greatest error is some- 

 what less than 26'. Let d be the space retardation measured in air 

 for wave-length X, then — 



p = 2ttc?/X. 



Thus the error in the determination of p expressed in wave-lengths 

 is — 



As in the experiments tan-ar = 0"3209, this is very nearly the case 

 in which the accuracy of the analyser has been determined by Pro- 

 fessor Stokes. The present results agree well with his limits of 

 accuracy. 



" The mean error of single observations amounted to about ^° in 

 the determination of the azimuth of the principal axis, about three or 

 four thousandths in the ratio of the minor to the major axis, and a 

 little more than a thousandth part of an undulation in the determina- 

 tion of />."* 



We notice that in the mode of expressing p in degrees an error will 

 be made more apparent, for the formula p = 27rdj\ shows that the 

 space retardation is divided by the small quantity A. Thus, express- 

 ing the difference 5° between the two sets of observations in wave- 

 lengths, we find that the error of determination of p is only about 

 1^ per cent, of a wave-length. 



* ' Phil. Mag.,' vol. 2, 1851. 



