HETEROSIS IN POPULATION GENETICS 159 



experiment two stocks differing for visible mutants. One stock was while- 

 and Bar-eyed, the other stock was normal for both characters. The two 

 stocks differed, too, in a number of quantitative characters such as fecundity, 

 fertility, rate of development, longevity, and size. 



After about thirty generations the two mutant genes had been wiped out. 

 This could have been expected on the basis of previous data of L'Heritier and 

 Teissier on the elimination of such genes in artificial populations. At that 

 time, however, I did not discard the populations, but kept them going for 

 some seventy more generations. All the individuals present in the popula- 

 tions were phenotypically normal. But testing from time to time the values 

 of the above mentioned characters, I could establish that natural selection 

 was continuously operating and favoring higher fecundity, higher fertility, 

 higher longevity, and quicker developmental rate throughout the four years 

 that the experiment lasted. At the end, the flies present in the population 

 were superior by a factor of more than six to the mean of the considered 

 characters in the two original parental stocks. When I measured such values 

 in the Fi hybrids between the two stocks I could observe values higher than 

 those obtained after more than one hundred generations of selection. 



The selection experiment could then be interpreted in two different ways. 

 Either (a) selection had picked up a new genotype made out of a new com- 

 bination of polygenes derived from the two parental stocks, or (b) selection 

 had just preserved by means of a heterosis mechanism a certain amount of 

 heterozygosity, which was at its highest value at the beginning of the experi- 

 ment. The fact that in the course of the experiment the factors had been 

 steadily improving seemed to be against hypothesis b, but I could not be sure 

 that was the case. 



I then set up a new selection experiment, whereby I put in competition the 

 original stock white Bar with the normal type derived from the population 

 which had been subjected to natural selection for more than one hundred 

 generations. The result was clear. The genes white and Bar were elimi- 

 nated in this second experiment at a much higher rate than in the first ex- 

 periment. In the first experiment the gene frequency of the gene Bar after 

 ten generations had dropped from .50 to .15. In the second experiment, after 

 as many generations, the Bar gene frequency had dropped from .50 to .03. 

 It seems that the genotype produced by a hundred generations of natural 

 selection under constant conditions was so much better adapted to its en- 

 vironment that it could get rid of the competing genes with much greater 

 ease than the original wild type flies. But could it not be that all or at least 

 part of this result could be accounted for by the action of some heterosis 

 effect? 



Another example of a similarly puzzling condition is an experiment on 

 artificial populations under way now in my laboratory. I would like to find 

 out whether it is possible to produce so-called small mutations or polygenic 



