342 GORDON E. DICKERSON 



within linecrosses, and 



<THb 



n 



among linecrosses, where/ is the inbreeding of the population of lines and n 

 is the efifective number of segregating chromosomal units. 



Range in degree of heterozygosity among all Fi crosses of a population of 

 lines for likely values of n {n = 100 and/= 1 in Fig. 21.4) is small rela- 



.3 .4 .5 .6 .7 .8 .9 1.0 



PROPORTION OF LOCI HETEROZYCX)US 



Fig. 21.4 — Frequency distribution for proportion of 100 loci heterozygous when k = 2 and 

 initial q = .75 within lines 50 per cent inbred (.4), between Fi crosses of homozygous lines 

 or in a non-inbred population (B), between Fi crosses of hnes inbred only 50 per cent (C), 

 and in a cross between complementary strains (^i = .95, go = .15) attainable only through 

 recurrent selection for cross performance {D). 



tive to the potential range. Hence recovery of Fi crosses much above the 

 average for all Fi's or for non-inbred stock cannot be expected. Inbreeding 

 provides a means for steadily reducing the proportion of heterozygous loci. 

 What is needed is recurrent selection in complementary strains to make them 

 steadily approach opposite extremes in gene frequency at each locus exhibit- 

 ing heterozygote advantage. The best Fi of a population of 100 Fi crosses 

 would average about 2.6 am above the mean, whereas the best 1 of 10 

 would average about 1.54 am above the mean and cumulative selection of 

 the best 1 of 10 in each of 10 recurrent cycles of selection would amount to 

 choosing Fi's that were 10 (1.54) am = 15.4 am above the original mean. 



Homozygous Tester versus Reciprocal Selection 

 Hull (1945) has proposed recurrent cycles of selection in crossbred material 

 based on progeny test in crosses with a single homozygous line (alternatively, 

 with two related lines or their Fi) as a method of producing highly comple- 



