18 GENETIC STUDIES OF RABBITS AND RATS. 



In the case of MacDowell's observations (Castle, 1914), correlation 

 coefficients were obtained somewhat lower than those here recorded, 

 but still notably high. The leg-bone correlation coefficients (the 

 highest of any in both sets of observations) were in MacDowell's 

 rabbits 0.858, 0.857, and 0.791, where the present observations give 

 0.927, 0.906, and 0.904. The skull-length correlation with tibia 

 was, in the former case, 0.701, in the latter 0.806. In general, the 

 correlations run about 10 per cent higher in the present set of obser- 

 vations, which is probably due to the greater range of variation in 

 size in the lot of rabbits on which these observations have been made. 

 The number of rabbits studied in the two cases is comparable, a 

 maximum of 376 in the case of MacDowell's rabbits, of 348 in the 

 present case. But the range of size-differences is greater in the 

 present case. For example, the tibia classes in MacDowell's rabbits 

 range from 88 to 110 millimeters; the range in the present lot is from 

 80 to 112 millimeters. The correlation is strongest where the range 

 is most extended, for at the ends of the range, where only individuals 

 of pure race occur, genetic differences are greatest and non-genetic 

 agencies sirk into insignificance, whereas at the middle of the range 

 non-genetic agencies exert a relatively greater influence. 



That this is so can be shown by a fuller analysis of one of the 

 correlation tables. Let us take, for example, the correlation between 

 femur and humerus (table 24). The correlation coefficient (r) for 

 all the rabbits (343) taken collectively is 0.906. For the pure races 

 only (table 25) it is 0.980, almost perfect correlation, indicating nothing 

 but genetic agencies at work. For the Fi cross-breds (table 26), 

 which should be no more variable than the more variable of the 

 parent races, r is 0.888, much lower than for the pure races, because 

 now only intermediate forms are present, the extremes represented 

 by the pure parent races not being present. For Fa by itself (table 

 27), r is 0.878, very nearly the same as for Fi, it will be observed, 

 although the genetic diversity is greater. For Fs is more variable 

 than Fi as regards both femur and humerus, yet the correlation indi- 

 cated is practically the same, the difference being no greater than the 

 probable error. If there were independent inheritance of factors 

 affecting the size of femur and of humerus respectively, the correla- 

 tion should be less close in F2 than in Fi because of recombination of 

 independent genetic factors, but such is not the case. Hence we 

 are forced to conclude that exactly the same genetic agencies affect 

 the size of femur and humerus. Similar reasoning would lead to 

 the conclusion that the same is true of all size features studied, in- 

 cluding not only bone-dimensions, but also weight and ear-length. 



