In comparing the bacterial numbers in two soil samples, however, the case is different. Within each bacterial group investigated the plate method should give counts proportional to the bacterial numbers in the soil. Thus, by the method one should be able to tell whether the bacterial numbers are increasing or decreasing over a period of time, or whether a certain soil treatment produces an increase or a decrease. With this end in view the technique of the method has been improved by recent workers. It was found that, when carefully standardised, the process of dilution of the soil could be carried out without significant variation in result (Table IV.), and that the accuracy of the method is limited mainly by the variation in colony numbers on parallel platings, due in part to random distribution of bacteria throughout the final suspension, and partly to the uneven development of colonies on the medium. The question of the medium was therefore taken up with a view to improving the uniformity of results obtained with it. Lipman, Conn, and others effected an improvement by using chemical compounds as nutrient ingredients, thus making their media more closely reproducible. On most agar media, an important disturbing factor is the growth of spreading colonies, which prevent the development of some of the other colonies. A medium has been devised at Rothamsted on which these spreading organisms are largely restricted.[61] A statistical examination[19] has shown that on this medium errors due to the uneven development of colonies, except in special cases, are prevented, so that in fact the variation in colony numbers between parallel plates is found to be that produced merely by random distribution of bacteria in the diluted suspension (see Table IV.). In this case the accuracy of the counts of the bacteria in the diluted suspension depend directly on the number of colonies counted, and can be known with precision. TABLE IV.--BACTERIAL COUNTS OF A SOIL SAMPLE. PARALLEL PLATE COUNTS FROM FOUR SETS OF DILUTIONS MADE BY DIFFERENT WORKERS. +------------------------------------------+ | Counts of Colonies on each Plate. | +------+--------+--------+--------+--------+ |Plate.| Set I. | Set II.|Set III.| Set IV.| +------+--------+--------+--------+--------+ | 1 | 72 | 74 | 78 | 69 | | 2 | 69 | 72 | 74 | 67 | | 3 | 63 | 70 | 70 | 66 | | 4 | 59 | 69 | 58 | 64 | | 5 | 59 | 66 | 58 | 62 | | 6 | 53 | 58 | 56 | 58 | | 7 | 51 | 52 | 56 | 54 | +------+--------+--------+--------+--------+ | Mean | 60·86 | 65·86 | 64·28 | 62·86 | +------+--------+--------+--------+--------+ Standard deviation between the four sets = 5·62. Standard deviation between plates within the sets = 7·76.