204 BAKNETT COHEN 



bacteria. For instance, the optimum for growth of Bad. ty- 

 phosum lies between the pH limits 6.2 and 7.2, though it will 

 tolerate greater extremes. Yet here we encounter the significant 

 fact that its optimum for maintenance under conditions of starva- 

 tion Ues between pH 5.0 and 6.4. It would, however, be unwise 

 at present to dwell upon the possibilities in this direction. 



Of interest in the same connection is the work by Shohl and 

 Janney (1917) who established the pH limits for the growth of 

 colon and tj^phoid bacteria in urine. Marsh (1918) found that 

 Bad. typhosmn is sensitive to the degree of acidity occurring 

 in sour milk; and Beckwith (1920) reports that this organism 

 is capable of surviving in rabbit bile in vivo even when the hy- 

 drogen ion concentration is depressed to pH 9.4. 



It may be asked how far we are justified in concluding that 

 the control of the pH with buffers conditions uniformity in 

 results between duplicates. Striking graphic proof is given in 

 Experiment 5 in whifch only one (that at pH 7.1) out of 8 tests 

 failed in this respect. In experiments 6 to 13, the data for which 

 are tabulated at the end of this paper, we find the large majority' 

 of duplicates to run closely parallel. That there are a few ex- 

 ceptions is evident but even in these, the average divergence is 

 very much less than in duplicates from de\iating unbuffered 

 solutions. 



A further test of this fact was made six months later when 

 Bad. iyphosum was studied in triplicate samples buffered around 

 pH 6.4. In this experiment, no. 13, the technique was especially 

 careful, and the individual bacterial counts are the most accurate 

 we have made, the probable error being below 5 per cent. If the 

 resulting data be plotted against time, it will be found that each 

 of the tripUcate samples yields an almost identical curve. We 

 have reason to believe that non-uniformity of results under these 

 conditions is due mainly to laxity in technique. 



' Of 33 such tests, about 4 may be considered to have shown any considerable 

 divergence between duplicates. 



