R. A. Fisher, H. G. Thornton, and W. A. Mackenzie 347 



of plates of a suspension of Rothanisted soil, one set of which was 

 inoculated with this organism. The reduction of, and increased variation 

 in colony numbers are again well seen. 



Table XV 



Effect of toxic organism from Rothamsted soil on colony development 

 from a soil suspension 



It is of course impossible to decide, with certainty, from a simple 

 record of colony numbers, whether the presence in the soil of some such 

 organism was the cause of the epidemics of variable plate-sets in Cutler's 

 series. However, the above two cases of high variance between parallel 

 plates, which have been traced to the presence of definite organisms, 

 show that this factor, though apparently of infrequent occurrence, is 

 capable of causing a disturbance in the colony numbers of precisely the 

 kind actually observed. It is important to notice that this, probably like 

 all other causes, that produce a sensible departure from the Poisson 

 Series, seriously disturbs the mean value. 



9. The Occurrence of Subnormal Variation 



It has been shown that in a small proportion (about 34 cases) of 

 Cutler's data, the variation between parallel plates has been apparently 

 lowered by some disturbing agency. The same phenomenon in a much 

 aggravated form appears in Owen's data (section 10), and has from time 

 to time occurred in Thornton's work. For example the 20 plates shown 

 in Table I display an unduly low variation, and though this fact does 

 not detract from the value of the data in proving the equivalence of 

 parallel dilutions, it does throw suspicion on the value of the mean as 

 an estimate of bacterial density. A similar depression appears in Table 

 XIV, Series B. 



