26 Perspecfives in Microbiology 



by Mendelian and populational analysis elsewhere. To be 

 sure, such pioneers as Massini, Beijerinck, and Barber did 

 distinguish fluctuating or impressed variations, which are 

 reversible physiological responses to the environment, from 

 fixed mutations representing innate genetic alterations. 

 Bacterial mutations are reversible, however, in the same 

 sporadic fashion as the primary events; reversion is no less 

 characteristic of other organisms, but this was overlooked 

 in support of cyclogenic or more obscure special theories 

 of bacterial dissociation (18). The confusion was com- 

 pounded by the tremendous size of bacterial populations 

 and the aggressiveness with which bacterial mutations and 

 reverse mutations often present themselves (6, 33). 



These features were put to good use in a renewal of 

 exact study by a biometric approach in the 1940's, cul- 

 minating in the analysis of clonal variance of mutations to 

 phage resistance by Luria and Delbriick (34). The out- 

 standing qualitative result of these experiments was statis- 

 tical proof of the uncontrollability, if not the absence, of 

 purposive regulation of adaptive mutations, such as re- 

 sistance to phage or to streptomycin. The statistical pro- 

 cedures have since been expanded for the calculation of 

 mutation rates and for more detailed analysis of spon- 

 taneous mutation (2, 38). But population dynamics is so 

 complex (7) that foreseeable progress here may consist of 

 balancing the complications omitted from the approxi- 

 mate calculations against the imprecisions of measurement. 

 The clonal variance that is the keystone of the qualitative 

 argument for preadaptive mutation also requires tedious 

 repetition to measure mutation rates with a scarcely ac- 

 ceptable precision of ±: 50 per cent (34, 38, 42). Ordinary 

 cultural methods also entail an ever-changing chemical 

 environment (7) in addition to an inexhaustible reservoir 

 of systematic as well as sampling errors (26). 



Novick and Szilard (39) have resolved these obstacles 

 by means of a simply engineered continuous-flow device, 



