328 Method of estimating Bacterial Densiti/ 



for development (Condition I). In a long experiment, however, where a 

 number of different batches of medium are used, this will be the case 

 only if the medium can be accurately reproduced, if, that is, different 

 batches of medium, prepared independently, give significantly the same 

 results. This reproducibility has been confirmed for Thornton's agar 

 medium (Thornton, 1922 (ii)). 



Again condition (IV) would fail if from any cause the dilution was 

 carried out in an irregular manner. This may be tested directly by carrying 

 through the whole dilution process independently with different portions 

 of the same sample. The following experiment is an example of such 

 a test. 



Four portions of a sample of Barnfield soil, simultaneously analysed 

 by four different workers (Aug. 14, 1921), gave the following counts: 



Table I 



Portion 



The four sets of plates are indistinguishable from random samples 

 from a single population. The variance estimated as from a single sample 

 of 20 is 8-52, actually less than the mean value for the variance within 

 each set, 9-15. An equivalent test is provided by the correlation between 

 different plates of the same set; this is —-089 ±-108, negative and quite 

 insignificant. In spite of the fact that the different plates of the same 

 set agree very closely, the variation between the four means is quite 

 insignificant. 



Table II 



Portion 



