Summary 1.Three series of bacterial counts from samples of fresh, unpasteurized milk have been completed. Six or seven analysts participated in each, working in two groups in laboratories located within fifty miles of each other. 2.In all cases, counts were made both by the agar plate method and by direct microscopic examination, thus permitting a check upon the accuracy of the counts not possible where only one method of counting is used. 3.In two series (B and C), the samples analyzed were carefully prepared so as to present the most favorable conditions possible for accurate counting and to allow checks to be made upon the accuracy of the results. This was accomplished by inoculating three lots of freshly drawn milk known to contain very few bacteria, with a skim milk culture of the colon organism. The amount of inoculum used was such that the final counts were expected to show the ratio 1:2:4. The colon organism was chosen because it grows well under normal conditions, and exists in milk largely as isolated individuals. 4.Under the above conditions, the results met all of the checks upon their accuracy so perfectly that there can be little doubt but that they actually were fairly accurate counts of the number of individual bacteria present. 5.The results obtained in the final series (C) were so uniform that the coefficient of variability was reduced to less than 15 in all cases. Under the conditions present, the variability of the microscopic counts was slightly greater than that of the agar plate counts. 6.In the counts made from samples containing a miscellaneous flora (series A), wide variations were found between the plate and microscopic counts. The primary cause of these variations appeared to be the existence of clumps of bacteria which were not separated into their component individuals in preparing the agar plates. The uniformity of the agar plate counts was generally good, indicating that the technique used was satisfactory. The greater lack of uniformity in the microscopic counts was in part due to the inexperience of some of the analysts, several of whom had never before attempted to make accurate counts by microscopical methods. 7.The average number of individuals in the clumps of bacteria present commonly varied between two and six; but at times (when streptococci were present) greatly exceeded these numbers. As the data indicate that the clumps are only very poorly broken apart in the processes ordinarily used in preparing dilution waters, the plate counts did not represent the full number of bacteria present. 8.The chief limitations upon the accuracy of the microscopic counts appear to be those involving the skill of the analyst making the microscopic observations, and the patience necessary in order to examine a sufficiently large quantity of milk to give an accurate average. Given unlimited time, and numerous duplicate preparations from a sample of milk, a skilled microscopist can secure reasonably accurate counts of the number of individual bacteria present in any ordinary sample of milk. Yet the laboriousness of this proceeding limits its usefulness, and makes it impossible to actually count the bacteria in examining large numbers of samples. 9.Fortunately neither the inaccuracy of the plate counts caused by the clumping nor the limitations of the microscopic technique just noted, appear to be. so great as to prevent the use of either technique where the purpose is to grade miscellaneous samples of unpasteurized milk into two or three grades. However, the information at present available indicates that attempts to use simplified methods for analysis for the purpose of making finer distinctions in quality introduces gross errors. When a finer classification is desired (as is now the case in many grade A plants) the use of the so-called simplified routine control methods should not be regarded as satisfactory. The present ituation suggests the desirability of the State exercising controls over bacteriological methods for analyses, whenever the results are to be used as a basis for payment in order to insure the use of more accurate methods of analysis just as it now does in the case of the Babcock test for determining the percentage of butter fat.