58 A. D. HERSHEY AND RAQUEL ROTMAN 



diagrams in which the disproportionate yields (k^0.20) are indicated by open 

 circles. Actually this effect is of minor importance, since the variations due to 

 unknown causes are so much greater than those due to variations in k. Con- 

 sequently, the uncorrected data lead to the same conclusions as the corrected 

 data; namely, that the proportionate yields of the recombinants are uncor- 

 rected in the crosses hXrl, hXrl3, and the corresponding reverse crosses, but 

 that there is a weak positive correlation for the crosses hXr7 and hr7X wild 

 type. 



As mentioned earlier, the correlation data in table 4 are not unselected. The 

 diagrams of fig. 7, however, show all the mixed bursts for the respective ex- 

 periments, those omitted from the table being indicated by crosses in the dia- 

 gram. It is plausible that some of the discrepant bursts came from bacteria 

 infected with a spontaneous mutant present in one of the parental stocks of 

 virus. In fact, two bursts from the cross hXr7 were found to contain a large 

 proportion of mutants, m in one case and weak inhibitor (Hershey 1946b) in 

 the other, which almost certainly arose in this way. The h stock used in these 

 crosses contained about 0.1 percent of h r virus, so that in crosses with r at 

 least one bacterium in 200 was infected with three types of virus, and would 

 be likely to yield an excess of one of the recombinant phenotypes. Unfortu- 

 nately the recombinant progeny in the exceptional bursts were not checked by 

 crossing with the parental stocks, which should be done in any further experi- 

 ments of this type. For the present these bursts throw some doubt on the sig- 

 nificance of the positive correlation between proportions of recombinants in 

 the crosses hXr7 and h r7X wild type. 



The results of the cross hXrU are of special interest because of the small 

 yield of recombinants. The distribution of the bursts with respect to absence 

 of recombinants is shown in table 5. Nine of the 125 mixed bursts fail to show 

 either recombinant, and 31 more lack one recombinant or the other. One can 

 test the hypothesis that the two sister recombinants arise independently as 

 follows. About 20 percent of the bursts fail to show a specified one of the two 

 possible recombinants. If the absence of the one were independent of the ab- 

 sence of the other, about (0.2) 2 or four percent of the bursts should show 

 neither recombinant. The number found, 9/125, is larger than this, but not 

 significantly so. Moreover, bursts lacking one recombinant do not show less 

 than the average proportion of the other (table 5). The data evidently fail to 

 exclude the hypothesis of independent origin of the two recombinants, but do 

 not, of course, rule out the hypothesis of reciprocal exchange. 



Another question that arises in connection with the data is concerned with 

 the number of genetic exchanges per bacterium. For explicitness, we consider 

 separately the hypotheses of reciprocal and non-reciprocal exchange. If ex- 

 changes are reciprocal, the bursts lacking a single recombinant are the result 

 of failure to recognize the few plaques of either type, of losses in the ten per- 

 cent of each culture not examined, and of unspecified biological accidents. As 

 previously computed, four percent or 5 of the 125 bursts fail to show either 

 recombinant for one or another of these reasons, leaving only 4 without re- 

 combinants possibly owing to failure of exchange. This number is too small 



252 



