R. A. FiSHEK, H. G. Thornton, and W. A. Mackenzie 355 



duplicate counts of the same plates have been taken as independent 

 observations, x^ h^'S been increased by this cause to some extent short 

 of doubling, so that we may say that in reality x^ lies between 3-77 and 

 7-54, while P lies between -584 and -185; neither value could be taken as 

 indicating a significant departure from expectation. 



We believe, therefore, that in this material, at all events in Series C, 

 the somewhat severe conditions under which the Poisson Series is 

 produced, were in reality fulfilled, and that the departure of the observa 

 tions from expectation could have been eliminated had precautions been 

 taken to secure a sufficiently accurate counting of the colonies. It must 

 however be borne in mind that the material employed consisted in 

 nearly all cases of almost pure cultures of B. coll in milk. The case cannot 

 therefore be compared closely to the different problem of counting such 

 a mixed bacterial flora as occurs in soil, where many different types of 

 organisms, whose growth may be mutually harmful, occur on the plates. 



The interference on the plates between dissimilar organisms cannot 

 here be seen, neither can the capability of the medium to check this 

 interference be studied. In this material, for example, there would be 

 little danger of frequent interference by "spreading" organisms, whose 

 growth, had they occurred, would probably have been stimulated by 

 such a medium as was used, containing peptone and meat extract. 



The lessened accuracy in counting a mixed flora on this medium is 

 illustrated in Table XXII, where the second group of platings, which 

 contains counts of uninoculated milk, shows a noticeably higher variance 

 in counting than the adjoining groups made from milk cultures of B. coli. 



The data show, however, that when such a simphfied flora is studied, 

 an agreement between parallel platings comparable with the expectations 

 of random sampling can be obtained. 



(D) Tlie data of W. Owen(i^) 



One of the most remarkable bodies of data which we have examined 

 is that provided by W. Owen in his investigation of various culture 

 media for the counting of micro-organisms in cane sugar products. In 

 this work, a variety of different media w^ere employed, varying in 

 composition, reaction and osmotic pressure. These were tested in counting 

 bacteria from a variety of sugar refinery products. From the variety of 

 media employed, and from the fact that most of them were new and of 

 untested value, it was to be expected that a rather high variance between 

 parallel platings would be found over the whole series taken together. 

 Had this been the case, separate tests would have been needed of the 



