MANNER OF PRODUCTION OF MUTATIONS 505 



have been exactly controlled (a requirement which probably cannot be 

 met unless very young males are used, since the ejaculates may contain 

 sperm of different ages), the recovered offspring will represent selected 

 material. For the sperm which were in a more "susceptible" stage when 

 treated, in the sense of a stage in which more not-to-be restituted breaks 

 (and probably more breaks in general) were produced, will, because of this 

 greater abundance of dominant lethals, leave fewer offspring, relatively 

 to the less susceptible sperm. Thus the frequency of recessive lethals 

 recovered will not be as high as the frequency with which they were pro- 

 duced in the total population of sperm which engaged in fertilization. 

 This selective influence will work disproportionately strongly at high 

 doses since at such doses the frequency of dominant lethals is in consider- 

 able measure caused by multibreak aneucentric rearrangements which 

 increase more than linearly with dose. Hence at high doses the fre- 

 quency of recessive lethals recovered from mixed sperm will, as the dose 

 rises still more, fall more and more below that of the recessive lethals 

 which had actually been produced. This would tend to flatten a con- 

 cavely rising curve and might well result in its appearing straight or even 

 convex. 



If the mutation-by-breakage interpretation of the linear frequency- 

 dosage relation is examined in the light of dosage data from organisms 

 other than Drosophila in which there is reason to infer the existence of 

 abundant position effects, the case against it is further strengthened. 

 Some of the fungi appear to fall into this category. Thus, in Neuros'pora* 

 mutations produced by X rays are, as in Drosophila, much more often 

 found to be connected with semisterility, evidently caused by transloca- 

 tions, than would be allowed on a random distribution of the two phe- 

 nomena with regard to each other (Sansome et al., 1945). It is a sig- 

 nificant finding of these investigators that the frequency of these muta- 

 tions is dependent on the time-intensity distribution of the dose, falling 

 off at lesser intensities just as is expected for structural changes but not 

 for effects connected with the primary breaks. Although the variation in 

 frequency with dose, even at the higher intensity, appeared to them to 

 follow a straight-line graph, nevertheless, low-dose determinations for the 

 high-intensity exposures, which would have provided the critical evidence 

 on this point, were lacking, and a careful comparison of the data for the 

 two intensities indicates that for high intensities the frequencies actually 

 fit better to a concavely rising curve, such as would be expected of events 

 showing time dependence.^ 



^ The details of this revision are as follows: The low-dose data obtained at the low 

 intensity in these experiments are obviously plus deviants since the values found are 

 too high for the rest of the low-intensity curve. However, they are shown in the 

 graph as if they were high-intensity data (although there are actually no low-dose 

 data for the high intensity), and this method of representation makes the high-inten- 



