VARIATION AND HEREDITY IN TWINS 127 



for each band. These studies are a necessary pre- 

 liminary for determinations of the coefficients of correla- 

 tion among quadruplet sets, but need not be dealt with 

 here. To illustrate the ways in which heredity works 

 in polyembryonic species only the most essential facts 

 about the modes of inheritance of these scute groups 

 need be presented. 



In all of this w^ork a limitation upon any complete 

 analysis of the situation is imposed by the fact that it 

 has been possible to study the heredity from one parent 

 only. Breeding in captivity was not found feasible 

 for several reasons. First, so large a number of sets 

 was required for statistical study that it w^ould have 

 been an enormous undertaking to capture and keep the 

 necessary number of parents. Secondly, attempts to 

 keep armadillos in confinement showed that, as a rule, 

 they become badly diseased and die. Thirdly, in order 

 to obtain knowledge of the pairing and symmetry rela- 

 tions of the embryos, it was found necessary to remove 

 the unborn fetuses from their mothers; this involved 

 killing the mothers, a practice hardly feasible in the case 

 of painstakingly domesticated animals. Finall}', in the 

 few cases where offspring were born in captivity, it was 

 found that the mothers ate the offspring, thus totally 

 nullifying the results of breeding experiments. Con- 

 sequently our method of capturing and killing preg- 

 nant females, removing and preserving the fetuses, and 

 also preserving the armature of the mothers for compari- 

 son with those of the fetuses, gave the maximum results 

 possible wath the material. 



This limitation of the study of heredity to the 

 maternal side only is less of a disadvantage than might 



