762 



THE AMERICAN NATURALIST [Vol. XL 



by Pearson,^ the hurtosis, the case is somewhat different. The- 

 oretically the normal curve is mesokurtic, or ^2 - 3 = 0. Now in the 

 present case ^2 — 3 differs from zero by more than its probable error. 

 The deviation is less than twice the probable error of so cannot 

 be considered as significant on this basis. As we shall see, how- 

 ever, we get a somewhat better fit to the data given by the actual 

 samj)le if we use a curve which takes into account this deviation 

 from the mesokurtic condition of the normal curve. In so far, 

 however, as we may infer from the sample regarding the conditions 

 in the general j)0])ulation from which the sample is taken, we can 

 conclude with a liiuli <lc<:n'e of j)r()hal)ility that in the variaiiori in 

 number of sev(h pn- vap^»h X,lawhiu,n jolhws the normal law of 



From tlie values of k, and k, given in ^rable 3 we see that what- 

 ever deviation from normality exists, is in the direction of a curve 

 of '^rype ]. In order to compare the gratliiation given by a normal 

 and a skew cur^e, I ha\e fitted both t\pes of curve to the data, 

 'i^he e(iuati<)n to the normal curve is 



while the e(inati.)n to' the Tvpc 1 curve is 



y- 127.(1121 (1 -i^s7^s-,)'' ''"' M -.J. 

 Calculating out the cnlinatr^ of tlic.c two ( iirvo (■()rresj)onding 

 to the different mnnl)ers of seeds, we have the results shown in 

 Table 4. 



TABLE 4 



Comparison of Observations mid Fitt< <l Curves 



