132 HEFFERAN. [VOL. II. 



tendency than on the left towards the production of few indefi- 

 nite teeth. On the right side, then, it is clear that the prob- 

 ability of a large number of definite teeth is associated with 

 that of a small number of indefinite teeth. The same thing is 

 shown for the left side, for although definite and indefinite 

 teeth both show negative skewness, the negativeness is much 

 greater in the indefinite than in the definite teeth. Therefore, 

 relatively, the skewness of the indefinite and the definite teeth 

 may be said to be, here also, of opposite sense. This agrees 

 with results shown in the correlation table, to be noted later. 



A peculiar result is obtained in regard to the distribution 

 curve of the total number of teeth. The left total falls into a 

 curve of Type I, while the right total is of Type IV. The 

 negative skewness of the latter is 0.050, while that of the 

 former is about two and one-half times as much. The table 

 of frequencies shows that the right total includes two classes 

 more, one at each end of the series, than the left total. There 

 is one individual in each of these two classes. It seemed 

 probable, by inspection of the calculation, that the critical 

 function, F, which was negative 0.5314, might be made posi- 

 tive by dropping these two extreme individuals, thus giving a 

 curve of Type I. I found this to be the case, and obtained 

 for F the value +0.210; but I found further that Type I 

 might be obtained by dropping only the individual of Class 5, 

 making F +0.0389. The skewness in this case was very 

 slight, only -0.00706. 



In order to determine which was the closer fit of the ob- 

 served curve to the theoretical curve in the two types, I cal- 

 culated the theoretical curves from the observed data with the 

 following result. 



TYPE IV. 1 



n = 400 d = 0.06822 M = 10.055 



s = 25.66 m= 13.830 y = 96.34 



a = 6.5798 ff -- 1.3386 zero ordinate = 9.1114 (M-md) 



v = 3.6796 = 8 9' 7" tan e = x / a 



y = y (cos. 0) 2m e-i-0 



1 For the methods of calculating the results given in the following tables, see 

 Davenport, '99, pp. 20, 23, 24. 



