R. A. Fisher, H. G. Thornton, and W. A. Mackenzie 329 



Equally close is the agreement between the sets of seven plates pre- 

 pared from four parallel series of dilutions (June 22, 1922), shown in 

 Table II. No trace of differentiation is observable, and the four sets 

 must be regarded as random samples from a single population. 



On certain occasions the same point is established by the analysis of 

 simultaneous samples from the same field. An agreement in such cases 

 shows the uniformity in bacterial density of the portion of the field 

 sampled; it also serves to show that no significant differences are 

 introduced by variations in the process of dilution. Thus four simul- 

 taneous samples from Broadbalk (Aug. 14, 1921) gave the following 

 counts. 



Table III 



Sample 



Plate 

 I 

 2 

 3 

 4 

 5 

 Mean 



From the whole set of 20 the variance is 56-27, from the four sets 

 of 5, 56-97, not a significantly greater value. The correlations between 

 plates of the same group is + -014 ± -108, an insignificant positive value. 

 By the most sensitive tests possible, no differentiation is observable. 



There is thus reason to claim that the manipulative technique can 

 be so efficiently standardised that no significant variations in it are 

 detectable, having regard to the variance that occurs between the colony 

 numbers developing on parallel plates from a single final dilution. 



Our attention is thus drawn to this variance between parallel plates, 

 which may be due solely to the chance distribution of organisms within 

 the final dilution, or may in addition be influenced by the mutual 

 interference between organisms on the plates, or by the failure of certain 

 organisms to develope into single discrete colonies. 



It is therefore necessary, in interpreting the results of the counting 

 technique, to discover the relative importance of these influences, on the 

 colony numbers, and on the variance between them. It is on the experi- 

 mental evidence as to the actual nature of this variance between parallel 

 plates that our further conclusions will be based. 



Nevertheless, the two questions of the reproducibility of the medium 

 and of the equivalence of results obtained by independent series of 



