132 CARNEGIE INSTITUTION OF WASHINGTON. 



Selection of Unstable Determiners vs. Plural Determiners, 

 E. C. MacDowell. 



Dr. MacDowell is studying the question of the stability of Mendelian 

 genes or detemniners which is a central topic for many investigations, 

 chiefly because of the extensive work of Dr. W. E. Castle. He reports 

 as follows: 



"If genes are stable and are passed from generation to generation unmodified 

 by their associations with other genes, the work of selection consists merely of 

 sorting out certain combinations of homozygous factors. After this is done, 

 selection can produce no further changes. If genes are modified by the other 

 genes in contact ^ith which they come, new genes as well as new somatic 

 appearances can be produced by selection. The work of Professor W. E. 

 Castle and Dr. J. C. Phillips on modifying the color patterns of rats by selec- 

 tion, reported in the Year Book for 1913, has been interpreted by these 

 investigators as showing that selection is effective in modifying Mendelian 

 genes and so in producing new kinds of genes. Their interpretation is not 

 the only one applicable. It is possible to suppose that the apparent success 

 of selection was due to the sorting out of multiple genes, which have been 

 assumed by the authors to explain certain phenomena of crossing. Work 

 bearing closely on this subject has been undertaken on the banana fly (Dro- 

 sophila ampelophila) . A race of flies has been produced lacking a restrictive 

 gene that limits the number of bristles on the back of the thorax in the normal 

 flies to four. In the absence of this gene the number of bristles is variable, 

 but greater than four. By the closest inbreeding and selection the number of 

 extra bristles was steadily increased for six generations ; after this no further 

 increase has been detected. At present the race is in the twenty-first genera- 

 tion. This result, supported by certain other very definite phenomena, has 

 led to the tentative adoption of the hypothesis that there are accessory genes 

 which, in the absence of the main restricting gene, hold down the numbers of 

 extra bristles, and that the success of selection at first is to be accounted for 

 by the dropping out of these accessory genes by using parents with high bristle 

 numbers, which in successive generations lacked more and more of these 

 accessory genes. After the sixth generation a state of equilibrium was pro- 

 duced in which these genes either were all removed or homozygous. This 

 interpretation, involving more than one determiner for a single somatic char- 

 acter, is much hke that reported by Shull in the Year Book for 1913, for his 

 work on shepherd's purse. The attempt is being made to test this hypothesis 

 by isolating lines of extra-bristled flies that show constant differences in bristle 

 number, and by crosses between these lines to determine whether or not the 

 differences l)etween the lines are due to accessory restrictors. Before this 

 can be satisfactorily attempted, special studies on the causes of the fluctuations 

 in the bristle-numbers must be undertaken in order to reduce these non-genetic 

 variations. It has already been found that the size of the fly, which varies with 

 the amount of food eaten in the larval state, influences strongly the number of 

 extra bristles, as a small fly is apt to have very few extra bristles (1 or 2) and 

 a large fly many (5 to 8). This is a good example of the influence of environ- 

 ment on the development of a Mendelian character, and emphasizes the neces- 

 sity of careful investigation of the role of environment in all genetic work. 



' 'Besides the plan outlined above , future work will include ( 1 ) further selection 

 for increase in the bristle numbers, so that the failure of selection can be abso- 

 lutely unquestioned, and (2) return selections, i. e., starting with the inbred 

 and selected race, low-grade parents wall be selected instead of high-grade. If 

 factors or genes may be modified, a complete return selection should be possi- 

 ble. If the first success of selection was due to the dropping out of ac- 



