114 



ing beforehand, or else to have run at 

 least half a dozen different pairs of 

 parallel lines of the control and treated 

 series, and to have obtained a definite 

 difference in the same direction be- 

 tween the two lines of each pair; 

 otherwise it can be proved by the 

 theory of "probable error" that the 

 differences observed may have been a 

 mere matter of random sampling 

 among genie differences originally 

 present. Accumulating large numbers 

 of abnormal or inferior individuals by 

 selective propagation of one or two 

 of the treated lines— as has been done 

 in some cases— adds nothing to the 

 significance of the results. 



At best, however, these genetically 

 unrefined methods would be quite in- 

 sensitive to mutations occurring at 

 anything like ordinary frequency, or 

 to such differences in mutation rate as 

 have already been found in the analyt- 

 ical experiments on mutation fre- 

 quency. And it seems quite possible 

 that larger differences than these will 

 not easily be hit upon, at least not in 

 the early stages of our investigations, 

 in view of the evidence that mutation 

 is ordinarily due to an accident on an 

 ultramicroscopic scale, rather than di- 

 rectly caused by influences pervading 

 the organism. For the present, then, it 

 appears most promising to employ or- 

 ganisms in which the genetic composi- 

 tion can be controlled and analyzed, 

 and to use genetic methods that are 

 sensitive enough to disclose mutations 

 occurring in the control as well as in 

 the treated individuals. In this way 

 relatively slight variations in mutation 

 frequency, caused by the special treat- 

 ments, can be determined, and from 

 the conditions found to alter the mu- 

 tation rate slightly we might finally 

 work up to those which affect it most 

 markedly. The only methods now 

 meeting this requirement are those in 

 which a particular mutable gene is fol- 



MULLER 



lowed, and those in which many 

 homozygous or else genetically con- 

 trolled lines can be run in parallel, 

 either by parthenogenesis, self-fertil- 

 ization, balanced lethals or other spe- 

 cial genetic means, and later analyzed, 

 through sexual reproduction, segrega- 

 tion and crossing over. 



V. OTHER POSSIBILITIES 



We can not, however, set fixed 

 limits to the possibiHties of research. 

 We should not wish to deny that some 

 new and unusual method may at any 

 time be found of directly producing 

 mutations. For example, the phenom- 

 ena now being worked out by Guyer 

 may be a case in point. There is a 

 curious analog>^ betu^een the reactions 

 of immunity and the phenomena of 

 heredity, in apparently fundamental 

 respects,® and any results that seem to 

 connect the two are worth following 

 to the limit. 



Finally, there is a phenomenon re- 

 lated to immunity, of still more strik- 

 ing nature, which must not be ne- 

 glected by geneticists. This is the 

 d'Herelle phenomenon. D'Herelle 

 found in 1917 that the presence of 

 dysentery bacilli in the body caused 

 the production there of a filterable 

 substance, emitted in the stools, which 

 had a lethal and in fact dissolving ac- 

 tion on the corresponding type of bac- 

 teria, if a drop of it were applied to 



^ I refer here to the remarkable specificity 

 with which a particular complex antigen 

 calls forth processes that construct for it an 

 antibody that is attracted to it and fits it 

 "Hke lock and key," followed by further 

 processes that cause more and more of the 

 antibody to be reproduced. If the antigen 

 were a gene, which could be slightly altered 

 by the cell to form the antibody that neu- 

 tralized it — as some enzymes can be slightly 

 changed by heating so that they counteract 

 the previous active enzyme — and if this anti- 

 body-gene then became implanted in the 

 cell so as to keep on growing, all the phe- 

 nomena of immunity would be produced. 



