242 Development of a Standardised Agar Medium 



bacterial numbers in milk, in which the numbers obtained by the plate 

 method were compared with those derived from direct counts made 

 under the microscope. The plate method, therefore, cannot tell us the 

 total numbers of bacteria in a soil sample, but it affords a means of com- 

 paring the numbers in two or more samples, by enabling us to count a 

 percentage of the total numbers. To make comparison possible, this 

 percentage must not appreciably vary. 



The sources of error in the method, which are not preventable, must 

 therefore be standardised, so that they will afPect the calculated numbers 

 to a constant degree. Thus, when comparing the bacterial numbers in 

 different soil types by this method, errors may be introduced by the 

 adherence of groups of bacteria to soil particles. The extent to which this 

 occurs is not at present known, but it is likely to vary in different types 

 of soil. In working with a single soil type, however, this source of error 

 is likely to remain constant and therefore loses its primary importance. 

 The close agreement that we have found (2) between bacterial numbers 

 calculated by this method from parallel soil samples taken at Rothamsted 

 seems to indicate that this is the case. The percentages of organisms which 

 are lost in the diluting and plating process, can be rendered sufficiently 

 constant by careful standardisation of the technique. A variation in 

 numbers obtained, resulting from random sampling, is necessarily in- 

 volved in making the dilutions. This variation can, however, be cal- 

 culated and due allowance made for it. 



The remaining sources of error are connected with the medium used 

 in the plating and with the development of the colonies therein. The 

 medium appeared to be so great a cause of variation as to render its 

 investigation a matter of first importance, and a necessary prelude to 

 the study of the other factors before mentioned. Its investigation has 

 therefore formed the subject of the present work. 



The qualities to be looked for in an ideal count medium have been 

 described by Conn (3). For the purpose of the general bacterial count, 

 constancy in the results obtained with a medium is by far its most 

 important property. The medium should be exactly reproducible by 

 different workers or by the same worker at different times. Also, a 

 suspension of soil, if plated on different samples of the medium, should 

 give rise to the numbers of colonies diftering only within the limits of 

 random sampling variance. 



Constancy in the results obtained with a medium depends mainly 

 on the following features. 



A. The composition of the medium must be constant. 



