S. E. LURIA 



such clones were expected to be present an appropriate fraction of the largest 

 clones (see Table 3). This correction is indeed a small one. 



Figure 1 compares the experimental distribution of clones with one or more 

 mutants with the expected distribution from the hypothesis of exponential 

 reduplication. The data corresponding to the "corrected" columns in Table 3 

 are plotted for r mutants, for w mutants and for the two together. They fit well 

 the expected relationship (Imear relation between log Y^ and log x with slope — 1) 

 for low clone size, up to mutant clone sizes of the order of 10-15. Above that 

 point, the frequency of mutant clones falls below the theoretical values. This 

 behavior is precisely what we should expect. The linear relation between log y 

 and log X — see Equation (1) — should only obtain for clones so small that they 

 have equal chances to be formed in all bacteria. For clone sizes of the order of 

 the burst size, a limitation is placed on the frequency with which these clones can 

 be observed. In the curve log Yx versus log x this limitation will manifest itself 

 as a downward concavity, which becomes appreciable aromid the value corre- 

 sponding to the lowest class of frequent burst sizes (about 20 phages per bacteri- 

 um in our experiments) and progressively more pronounced as the median burst 

 size is approached. Of course, there cannot be any clone larger than the maximum 

 burst size. An additional factor (suggested by Dr. S. Dancoff) that works in the 

 same direction is nonsynchronization itself; in fact, this results in the existence, 

 within each burst, of subclones that have origmated at the same generation but 

 have different sizes, thus producing effects similar to those of the burst size 

 differences. 



No closer analysis of the concave portion of the distribution frequency curve is 

 feasible beyond these qualitative considerations, since the clones in this region 

 are few and fluctuations affect the results strongly. Altogether, our results fit 

 quite well the hypothesis that the genes responsible for the investigated pheno- 

 types reproduce exponentially by successive reduplications. Let us now analyze 

 some of the factors that might affect the experimental results. 



1. Failure to recognize mutants. This cause of error is difficult to assess; we 

 believe the error to be very small. All plates were scored by the same observer 

 after the optimum incubation period, and every plate that might have presented 

 difficulties in scoring, because of crowding or of faulty layering, was discarded 

 before exammation. All the doubtful plaques were picked and replated for the 

 phenotype test. Any residual error from this source would probably result in 

 underestimation of the frequency of clones with one mutant, since the finding of 

 the first mutant on a plate might sharpen the alertness of the observer, thereby 

 increasing the chances of detecting other mutants on the same plate; the classes 

 of clones with more than one mutant might thus have been favored in our 

 observations. 



2. A more definite and more easily evaluated source of error is the coincidence 

 of more than one clone of a given mutant type (r or w) on the same plate because 

 of coincidence of two mutations, either in the same bacterium or in the group of 

 bacterial bursts examined on one plate. The expected coincidences ("doubles") 

 were calculated to be 2.6 r and 4.3 w. It is not easy to correct for these "doubles," 



146 



