70 GENETIC STUDIES ON A CAVY SPECIES CROSS. 



investigators; but the comparison is interesting and shows how uni- 

 form the skeletal dimensions of adult guinea-pigs really are. It is 

 probable that the wild cavy species is likewise very uniform. 



The coefficients of variability of the one-half wild hybrids were calcu- 

 lated from such small numbers (5 males and 8 females) that their value 

 is doubtful. Such coefficients are most valuable and accurate when 

 the number of variates is large. When the total number of variates 

 is small, a few wide deviates greatly increase the standard deviations, 

 and therefore increase the coefficients of variability also. According 

 to the theoretical scheme involving multiple, independent size factors, 

 incompletely dominant, the Fi generation should be a blend and no 

 more variable than the parents, if the parents were practically pure. 

 As a matter of fact, the | wild were larger than either parent. We 

 say that such phenomena accompany the heterozygous condition, but 

 we can not adequately explain it. Taking the coefiicients as they 

 stand, the variability of the | wild females was slightly greater than 

 that of the guinea-pig parent, but the male hybrids were on the whole 

 no more variable than their parent. In view of the fact that the 

 chances of error are great, no conclusions can be drawn. 



The one-quarter wild hybrids, or Fz generation, showed no great 

 increase in variability, as one would expect on the hypothesis of many 

 interchangeable factors without dominance. The males were no more 

 variable than the guinear-pig, and the females were only slightly so. 

 Here, again, the numbers were small (16 males and 20 females) and the 

 results are subject to a serious objection. 



The one-eighth wild hybrids, or F3 generation, were on the whole only 

 sUghtly more variable in both sexes than the guinea-pig. 



It can be readily seen that all the coefficients of variabiUty are small 

 and form no series consistent with the hypothesis advanced, according 

 to which the Fi generation should be no more variable than the parents, 

 but the F2 generation should show an increased variability, while the 

 F3 should be less variable than the F2 generation. The whole 128 

 coefficients in tables 67 and 68 are very small and close together. 

 Moreover, if one considers the probable errors, the chances are small 

 that the differences in variability are not due to random sampling. 

 Practically every coefficient in any particular dimension would over- 

 lap every other one in that dimension if the probable error is multiplied 

 by four. Therefore, from the standpoint of pure random sampling, 

 the chances are large that a repetition of these experiments, under 

 similar conditions and involving the same numbers, might easily give 

 results with no significant differences between the coefl&cients of varia- 

 bihty. It must be stated that probable errors for the I and | wild 

 are very unreliable, since the numbers are so small. 



Examining the data as they stand, to ascertain which dimensions 

 . show a series of coefficients most variable in the j wild and grow less 



