178 THE BIOLOGY OF TWINS 



measurable or enumerable characters. In human twins 

 only dimensional differences have been determined, 

 and these are subject to too large an element of environ- 

 mental control to be used as a test of heredity. 



In the armadillo, however, we find an ideal material 

 for statistical study in the armor scutes in the nine 

 bands of armor and in the armor shields. A determina- 

 tion of the coefhcient of correlation for large numbers 

 of sets of quadruplets reveals the fact that on the 

 average this coefficient approaches within about 7 per 

 cent of complete identity. For example, the coefficient 

 of correlation derived from an array of 56 sets of male 

 quadruplets gives a coefficient of o. 9294=1=0.0057. A 

 similar high coefficient was found for 59 sets of female 

 quadruplets. These figures were derived from a study 

 of the total numbers of scutes in the nine bands. Now, 

 when we compare with this the coefficients of correlation 

 for individual bands, it is found that there is a very 

 decided drop from the near approach to identity seen 

 in the banded region as a whole, for the average coeffi- 

 cient of correlation determined for three sample bands 

 (i, 5, and 9) is about 0.4+. These results may be 

 interpreted as showing that the total number of scutes 

 in a large armor region is rather definitely predeter- 

 mined, but the alignment of the scutes into rows or 

 bands is a process involving developmental mechanics 

 of a cruder sort which appears to be largely beyond 

 the limits of hereditary control. Here then we would 

 appear to be in possession of facts that should enable us 

 to draw the line between "nature and nurture" or to 

 determine the limits of hereditary control. But, as 

 is usual with statistical results, the averaging up of 



