VIABLE BACTERIA IN YOUNG CULTURES 



423 



800, and 800 and over developed. The results are shown in 

 table 4. 



These figures, of course, do not represent the sampling error 

 itself; this can only be obtained by putting up comparable series 

 of tubes from the same emulsion; but they do show that the 

 fewer colonies there are per tube, the less chance is there of ob- 

 taining an accurate sample of the emulsion under investigation. 

 When more than 800 develop, the increasing accuracy of repre- 

 sentation is overshadowed by the error introduced by the diffi- 

 culty experienced in the actual counting of the tubes. The con- 

 clusion may therefore be drawn that in order to keep the sam- 

 pling error as low as possible, somewhere between 400 and 800 

 bacilli should be inoculated per tube. Naturallj^, if more than 

 three tubes are put up, the sampUng error will be smaller, but as 



TABLE 4 



Number of experiments. . . . 

 Percentage mean deviation 



13 

 19.6 



50 TO 



100 



24 

 9.42 



100 TO 



200 



43 

 7.13 



200 TO 

 400 



42 

 6.08 



400 TO 



800 



42 

 4.44 



800 AND 

 OVER 



11 



6.89 



three has been selected as a suitable number for this work, the 

 results have been calculated in accordance with this figure. 



b. The error of overcrowding. We now come to consider 

 the second factor deternadning the optimum number of bacilli 

 to be inoculated in putting up viable counts by the tube method, 

 namely the error of overcrowding. More or less in proportion as 

 the error of sampUng decreases as the number of developing colo- 

 nies increases, so the error of overcrowding increases as the number 

 of developing colonies increases. The two vary in opposite direc- 

 tions; the greater the number of colonies the less the sampUng 

 error; the fewer the number of colonies, the less is the overcrowd- 

 ing error. A point must be chosen between the two which will 

 permit of the minimum combined error being experienced. Be- 

 fore this could be done however, it was necessary to ascertain the 

 actual effect of overcrowding on the development of colonies in 

 tube preparations. As mentioned above, tliis overcrowding error 

 is one which seems to have been neglected by the majority of 



