442 CITY MILK SUPPLY 



the same plate is counted by two different laboratories the results are not infre- 

 quently 100 per cent, apart, and occasionally even more. 



" A third factor modifying the counting is the method adopted by each labora- 

 tory of estimating numbers rather than actual counting. In plates that contain 

 large numbers of colonies it is, of course, necessary to make an estimate rather than 

 to count them all. The laboratories adopt different methods in such conditions, 

 the results being slightly different. 



"IV. Irregularities in samples from the same bottle of milk. Two samples 

 taken from the same bottle of milk, even after thorough shaking, are by no means 

 identical. This is easily explained, and is due (1) to the clumping of bacteria 

 and (2) to the fact that inasmuch as bacteria are not in solution but are solid 

 objects they cannot be expected to be uniformly distributed through the liquid. 

 The tests show that one sample of a bottle of milk when tested by the plate 

 method may sometimes not contain more than one-fourth as many bacteria as 

 other samples from the same bottle. 



"V. In low counts variation between duplicate samples is sometimes con- 

 siderable, due to the irregularity of the distribution of bacteria. 



"In high counts variation in reports is also sometimes very great, due to ex- 

 cessive crowding and to methods adopted in estimating the number of colonies. 



"6. The extent of the variation in the results obtained from the analysis of 

 duplicate samples varies widely with the care that is taken in the laboratory 

 technique. It has been found in these tests that at first the above causes of 

 irregularity were sufficient to give results disagreeing as much as tenfold in the 

 number of bacteria that would be reported from any sample of milk. The vast 

 majority of results, however, were much closer than this, even at the beginning 

 of this series of tests. 



" 7. After attention had been called to the points of irregularity and the lab- 

 oratories had adopted methods of bringing about uniformity in technique so far 

 as possible, the variations were very greatly reduced, the last tests showing that 

 when sufficient care is given the variations need not be more than twofold. It 

 is not possible to rely upon a greater accuracy than 100 per cent, even when an 

 average of more than one analysis is obtained, although most of the results fall 

 considerably below this limit. 



"8. There is no essential difference in the results whether milk dilution is 

 directly inoculated into the petri dish and the agar poured upon it, or whether 

 the milk dilution is inoculated into the melted tube of agar and subsequently 

 poured into the petri dish. The difference between the two methods is so slight 

 as to make it impossible to determine which is the superior of the two. But 

 when examinations of cream are made the plate inoculation is unreliable, and the 

 inoculation must be made in the agar tube followed by thorough agitation. 



"9. Five days' incubation (48 hr. at 37 and 72 hr. at 20) gives a very 

 slightly greater count than a 2-day incubation. The increase in count is so slight 

 that for general regulation purposes it is hardly superior to a 2-day count; and 

 considering the superior value of obtaining the count promptly there seems to be 

 no reason for changing from a 2-day count to a 5-day count. 



"10. A 24-hr, count gives on the average about one-half as high numbers as 

 a 48-hr, count. 



"11. In spite of all these irregularities the results with duplicate samples in 



