DELAYED EXPRESSION OF MUTATIONS 471 



It is realized that this early low division mortality may not persist through- 

 out the whole of the growth cycle but may rise as the population in a culture 

 approaches the saturation density. In order to minimize the possibility of this 

 effect, the end population of cultures grown on agar (method 4) was not al- 

 lowed to reach that period at which growth rate begins to decline (approxi- 

 mately 5X10 9 bacteria on a 100 millimeter plate of broth agar). If, despite 

 these precautions, there was an increase in division mortality, then as pointed 

 out earlier, the estimates of mutation rate obtained would be somewhat greater 

 than the true rate. It will be noted that this would tend to weight the data ob- 

 tained by method 4 against the interpretation arrived at. We may therefore 

 conclude that, errors from this source do not affect the main conclusions. 



DISCUSSION 



From the experiments described, it is apparent that those methods which 

 take into account only the number of resistant clones developing in a given 

 period, and not the number of resistant individuals in these clones, give a low 

 estimate of the mutation rate (see table 11 for summary). 



Method 1 falls in this category, since it is based upon the proportion of cul- 

 tures in which no phenotypic mutants have appeared — this being a function 

 of the total number of phenotypically resistant clones in the series of cultures — 

 and is independent of the size of the clone. 



Method 4 also falls in this category, since the members of a mutant clone 

 are confined to one spot on the agar and after one member has become re- 

 sistant others may change in the same manner without being detected. 

 Method 4, which uses the number of resistant clones appearing during rapid 

 growth, provides a necessary check on method 1, which uses the number of 

 resistant clones developing during growth to saturation, and which would be 

 strongly biased if resistant clones appeared at a different rate during the few 

 divisions prior to the cessation of growth. The estimates of rate obtained by 

 these two methods are, within the limits of experimental error, similar, and in 

 the present experiments — using resistance of strain B/r to the phage Tl — 

 average approximately 0.5 X 10~ 8 . 



Methods that are based on the increase in numbers of resistant individuals 

 developing during growth, however, as distinct from the numbers of resistant 

 clones, produce much higher estimates of mutation rate. The following fall in 

 this category: method 2, which uses the average number of resistant bacteria 

 in a series of similar cultures started with small inocula; method 3, which uses 

 the number of resistant bacteria in a single culture started with a large in- 

 oculum; and method 5, which uses the highest number of resistant bacteria in 

 any one culture of a series. 



Method 3, which is based on a relatively direct calculation, provides a neces- 

 sary check on method 2, which is subject to a statistical bias. Similarly, 

 method 5, which is based on the early mutations, provides a check on method 

 2, which would be biased if the delay in developing phenotypic resistance were 

 excessive in an appreciable proportion of the lines of descent within a mutant 

 clone. Method 3 is also of use in this connection in the one experiment in which 



