JN^o. 479] 



variationZin lotus 



763 



Of course to get absolute accuracy, areas instead of ordinates 

 should be compared with the observed frequencies, but inasmuch 

 as the number of groups is here large, the error made bv comparing 

 ordinates will not be serious. 



The frequency polygon aiul fitted vnvxvs arc shown in Fig. 4. 



The fit is seen to be .'xccllcnt in tlie case of l)oth the t-urv,-s, but 

 the slight superiority of \\w 'iy[)(> 1 rwvw is apparent. The dif- 

 ference, as has l)e('n jx.intcd ont al»ove ip. 7i;L'i. Ix'tween this and 

 the normal cnrv*- is not signitii-mt. The greatest (hserepanty 

 between the observations and the curves is in the region abont the 

 mode. I am luiable to account for the curious irregularity in the 



