SUMMARY:Some hypotheses are considered which describe the aetiology of a fatal infection in a partially resistant host ; i.e. a host which does not invariably die after inoculation with one bacterium. The hypothesis of independent action postulates that the mean probability per inoculated bacterium of multiplying to cause (or help to cause) a fatal infection is independent of the number of bacteria inoculated and, for a partially resistant host, is less than unity ( 1 > p > 0). It predicts : (1) that the slope, b, of the probit-mortality/log-dose curve will be 2.0 or less a t the LD 50 point ; (2) that, while hosts dying after inoculation with many LD 50 die as a result of the multiplication of many of the inoculated bacteria, most of those dying from 1 LD 50 or less do so following the multiplication of only one of the inoculated bacteria, regardless of the total number of bacteria inoculated. When a mixture of several equally virulent, distinguishable variants of a given pathogen are inoculated, fatal infections caused by the growth of one bacterium should result in the predominance of only one variant a t post mortem. The hypotheses of maximum and of partial synergism postulate that inoculated bacteria co-operate so that the value of p increases as the size of the dose increases. They predict : (1) that the slope of the dose-response curve may be more than 2.0 a t the LD50 point ; (2) that all observed fatal infections will be initiated by more than one bacterium and that consequently the inoculation of a mixed inoculum will lead to the predominance of several variants at post mortem.Variants of Salmonella paratyphi-B which carried one of the three flagellar antigens, b, i, or e,h, were prepared by transduction. Variants of S. typhimurium which either fermented or did not ferment xylose were also prepared. In each case, the variants did not differ detectably in growth rate in vivo, or in virulence. The value of b was 1.8 for mice inoculated with S. paratyphi-B by intraperitoneal injection (LD50 = 74' x lo6 organisms). As predicted by the hypothesis of independent action, the relative frequencies of the variants in samples of heart blood obtained post mortem varied greatly from mouse to mouse when the dose was 1 LD50 or less, and became progressively more uniform (and similar to the inoculum) as the size of the dose increased. The value of b was 0.66 for mice challenged by S. typhimurium by subcutaneous injection (LD 50 = 2 x lo8 organisms) ; and challenge with a mixed inoculum gave similar results.Despite this general conformity with prediction, far fewer mice than expected yielded only one variant at post mortem. This discrepancy possibly resulted from * Present address.