EFFECTS OF INBREEDING AND CROSSBREEDING. 15 
The probable errors derived from these formule are given in 
Table 31 for comparison with the values shown in Table 29, and the 
differences between experiments shown in Table 30. The probable 
errors of the differences are from 10 to 20 per cent larger than the 
figures shown. 
Before closing the discussion of probable errors there is one other 
point which should be brought up. We have based the probable 
errors on the number of litters or on a compromise between number 
of litters and number of individuals designed to allow for the cor- 
relation between litter mates. But the characteristics with which 
we are dealing here are wholly or in part characteristics of the parents. 
Should we not therefore base the probable errors on the number 
of matings? We have, however, previously (Part I, Bulletin 1090) 
called attention to the fact that there is no significant correlation 
between the successive records of a given mating as to size of litter, 
interval between litters, birth weight, and gain (Experiment B). This 
can also safely be affirmed of the mortality among the young. Thus 
persistent good or poor health of the dam or sire can be of very little 
importance in determining these characters. 
In most experiments the parents are either inbreds or first crosses 
between inbred families and thus should be homogeneous genetically. 
In these cases there would seem no question that the litter, if not 
the individual, is the proper unit in calculating probable errors of 
experiment averages. In the cases in which segregation should be 
taking place among the parents (C2, CL, CG, and B) the propriety 
of this course depends on the number of factors involved. If only 
one allelomorphic series of factors were involved in the differences 
among the families in a given respect, the testing of only a few 
second generation females in C2 might give very misleading results, 
whatever, the number of litters produced by each one. If, however, 
many factors are involved, as is probably the case, the genetic 
heterogeneity should be less. It is believed that a sufficient number 
of matings was made in all cases to make the parents a fair sample 
of their respective experiments. The number of matings which 
produced young is given for each experiment in Table 10. 
RESULTS. 
We have concluded that the differences between each experiment 
and the inbreds as shown in Table 2 give a satisfactory basis for 
comparison. 
The same results are shown graphically in Figures 8 to 19. A 
general survey of the figures establishes beyond question that marked 
improvement in every respect is brought about by crossbreeding. 
The characters, however, fall into two rather sharply opposed cate- 
gories. In some of them there is little or no improvement in the 
