INHERITANCE OF VARIATES IN THE ARMADILLO 171 



widely. In the banded region the cofficient of polyembryonic 

 correlation for individual bands ranges from 0.5051 to 0.7458; 

 in the caudal region that of individual tail rings ranges from 

 0.5241 to 0.7454. Here again there is close agreement between 

 the two regions of armor. 



This striking similarity must mean that in both regions the 

 same factors are in operation and that what is true for one region 

 is true for the other. Already it has been stated more than once 

 that the comparatively low coefficient that obtains for individual 

 bands means that there is practically no predetermination with 

 respect to the alignment of scutes into individual rows. Although 

 this may be true for groups or averages I think I can show that 

 the statistical method once more conceals a considerable part of 

 the truth; for it is quite clear, as we shall see in the following 

 paragraphs, that with respect to individual rings of individual 

 members of sets of quadruplets there is a strong tendency for an 

 exact repetition of the maternal scute numbers, which means 

 not only germinal predetermination but also exclusive or alter- 

 native inheritance. 



In table 4 certain peculiar conditions come to light. In some of 

 the rings the uniparental (maternal) correlation is almost as high 

 as the polyembryonic correlation, in others — the probable error 

 being taken into account — there is practically no correlation 

 between mothers and offspring, while in the remaining rings the 

 coefficient of correlation falls about midway between no correla- 

 tion and that which obtains for the polyembryonic relation. This 

 state o^ affairs does not in the least comply with my notion of a 

 consistent scheme of blended inheritance, but does strongly indi- 

 cate at least a strong predominance of some sort of alternative 

 inheritance. In some bands the resemblance between mother and 

 offspring is almost as strong as that which exists among the quad- 

 ruplets; in others there is practically no resemblance to the mother 

 and we must assume that the paternal influence is dominant. 

 On this basis one would expect, if the influence of both parents 

 is equal, an average uniparental coefficient of correlation of just 

 half that of the average polyembryonic coefficient of correlation 

 for the ten rings, which is 0.3131. The actual average maternal 



