200 



PHYSIOLOGIC GENETICS 



resemblance tends most often to be augmented by nongenetic causes. Full sibs 

 always have the same mother and they tend, especially if they are also littermates, to 

 be subjected to similar environmental circumstances. For these reasons the correlation 

 between full sibs is to be avoided as a means of estimating the heritability. 



Fig. 35. Heritability of the number of urethan-induced pulmonary tumors in mice of 

 a random-bred strain (JC), by regression of offspring on the mean of both parents. 



MID-PARENT 



Based on preliminary data from a current experiment by Miss J. L. Bloom, a total of 

 299 offspring from 46 pairs of parents are represented. Each point shows the mean of all 

 offspring from parents with the given mean tumor-number, the approximate number of off- 

 spring in each point being indicated by the size of the dots: small = fewer than 10, 

 medium = 10-20, large = more than 20. The straight line is the computed regression 

 line, each point being weighted by the number of offspring contributing to it. The regres- 

 sion coefficient, which estimates the heritability, is 0.173 ± 0.043. 



The absence of nongenetic causes of resemblance in a relationship is the most 

 important criterion in choosing a method of determining the heritability. After that, 

 the choice of method is a matter of deciding which gives the most precise estimate with 

 a given expenditure of effort. In general, low heritabilities are more efficiently esti- 

 mated from the half-sib correlation, and heritabilities higher than about 20 per 

 cent are more efficiently estimated by offspring-parent regressions. The offspring- 

 parent regression is usually the more convenient method for work with laboratory 

 mammals. It is usually also the more efficient because the total number of measure- 

 ments made is not usually the limiting factor, but rather the number that are made at 

 one time. By the measurement of parents and offspring, the work is spread over two 

 generations. The rest of this section is concerned with the technical problems that 

 arise in the estimation of heritability, first from the offspring-parent relationship and 

 then from sibs. 



