334 GORDON E. DICKERSON 



(Hazel et al., 1948) likewise have shown little advantage accruing from 

 selection during development of the inbred lines. 



The apparent inability of selection to offset the decline in performance 

 from mild inbreeding casts doubt on the assumption that epistasis or ordi- 

 nary dominance (between none and complete) can account for the major 

 influence of inbreeding on performance in swine. Unless one assumes a pre- 

 ponderance of tight repulsion phase linkages, selection should have increased 

 the frequency of favorable dominant genes. Similarly, under epistasis in 

 which the genetic intermediate is optimum, selection should have prevented 

 fixation of the more extreme homozygous combinations, particularly if a 

 rather large number of loci determine the genetic range for each primary 

 function. 



The type of genetic mechanism that would most surely produce an in- 

 breeding decline relatively unresponsive to selection is heterozygote superior- 

 ity (^ > 1). Here selection would maintain gene frequency near some inter- 

 mediate equilibrium value, rather than move it toward fixation of any one 

 allele (qa smaller). Linear regression of genotype on phenotype (heritability) 

 would be lower than for lesser degrees of dominance, making selection rela- 

 tively ineffective. Inbreeding depression for dominance, which is propor- 

 tional to 2 5'a(1 — Qa) kf, would increase with k, particularly since qa would 

 be smaller and Qa (1 — q) larger than under partial or complete dominance. 



"Controlled" Selection Experiments 



Results have been published from two ''controlled" experiments on selec- 

 tion with minimum inbreeding in swine. In both the Illinois study of growth 

 rate (Krider et al., 1946) and the Alabama study of feed efficiency (Dickerson 

 and Grimes, 1947), the high and low selection lines separated appreciably 

 and significantly. However, it is difficult to judge from the time trends 

 whether the difference came partly from improvement in the high line or 

 almost entirely from decline in the low line. Taken at face value, the time 

 trends indicate that the separation was due to decline in growth rate of the 

 low line in the Illinois experiment, but that efficiency increased in both lines 

 in the Alabama study. 



In these experiments, the low line involved a reversal in the usual direction 

 of selection. This amounted to assigning new selective values to genes affect- 

 ing growth and feed utilization, and hence selection might be expected to be 

 unusually effective for the first few generations in moving toward some new 

 equilibrium. In both experiments, selection was most effective in the first 

 generation. 



In Goodale's (1938) and in MacArthur's (1949) selection for size in mice, 

 there is no question that a steady increase in size was produced. However, 

 these experiments with adult size in mice are not directly analogous to those 



