milk containing large numbers of bacteria is due to the increasing predominance of the lactic acid bacteria.

6. The clumps of bacteria vary greatly in compactness and therefore vary greatly in the amount to which they break apart into smaller clumps in preparing dilutions for the plate method. This causes an error in plate counts which cannot be measured directly. Indirect data indicate that the average clump of bacteria in market milk breaks apart 2.6 times in preparing dilution waters. This varies greatly in individual cases, however, as in 24 instances out of 345 the clumps broke into 7 or more groups" each. In one instance the resulting number of groups" was 16.5 times the original

number.

7. Because of the serious errors introduced into the plate counts by the presence of clumps of bacteria, as well as by the presence of living bacteria which do not grow on agar, popular ideas as to the number of bacteria in milk, based as they are on plate counts, are incorrect. The actual number of bacteria in milk is usually greatly in excess of the figure obtained by the plating method.

INTRODUCTION.

Quantitative determinations of the germ life in milk, as a means of ascertaining and controlling its wholesomeness as a food, have been quite generally used in the United States. The judging of the quality of milk according to bacterial content has necessitated the establishment of numerical standards which were to be used as guides in classifying it into grades. The increasing use of such standards makes it very important to the commercial interests involved as well as to the officials charged with the enforcement of regulations governing the control of milk supplies, that definite information be secured to determine the accuracy of the analytical methods used.

Bacteriologists have, from the beginning, been aware of the difficulties involved in making accurate bacterial counts and of the wide fluctuations possible in duplicate counts. Many attempts. have been made to measure the amount of these errors but they have not proved successful because of the lack of an absolute standard of comparison. The recent improvements in the technique of the direct microscopic method of counting bacteria in milk permit

comparisons to be made between counts thus obtained and counts made by the officially recognized agar-plate method. Results secured in this way permit more definite conclusions concerning the accuracy of bacterial counts than have hitherto been possible.

PREVIOUS WORK.

Much valuable work has been done by way of investigating the relationship between such factors as temperature, food requirements, chemical reaction, dilution, etc., and the number of colonies that grow on agar plates. This work, together with the reports drawn up by the Committees on Standard Methods of Analysis appointed by the American Public Health Association,1 have tended to make the methods used in plating milk samples more uniform. In spite of this, it is well known that there still exists great difficulty in getting uniform plate counts from the same sample of milk when the analyses are made in two different laboratories, or even when the analyses are made by two different workers in the same laboratory.

Probably the most extensive attempt which has been made to so standardize the plate method for making counts of bacteria in milk as to secure uniform results from different laboratories was carried out under the direction of Dr. H. W. Conn 2 working in conjunction with four New York City laboratories which were accustomed to doing routine work. The aim was, first, to determine how great the variation in count would be when duplicate milk samples were plated in each of these laboratories; then, to determine in what way the technique used by each differed. After making as careful a standardization of the methods as found practicable, claim is made that, in subsequent trials, more accurate counts were secured because there was a better agreement in the counts obtained by the different laboratories.

An objection which might be made to this claim is that it is impossible to tell when a count of bacteria is accurate. This is because it is exceedingly difficult to count accurately large numbers of any object, and the smaller the objects are, the nearer the task of counting

1

1 Report. Amer. Jour. Pub. Hyg., 20 (N. S. 6):315–345. 1910. Also Report Amer. Jour. Pub. Health, 6:1315-1325. 1916.

2 Conn, H. W. Standards for determining the purity of milk. U. S. Pub. Health Serv., Pub. Health Reports, 30:2349-2395. 1915.

approaches impossibility. A large office force is required to count the several million dollars of gold, silver, and bank notes stored in the United States Treasury vaults. But money is large enough to be handled separately and the different denominations lend themselves readily to systematization. Since this is not true with reference to objects so numerous and tiny as bacteria, actual counting is out of the question. Therefore the best that can be done is to make estimates of the numbers present; and the exact number of bacteria not being known, the accuracy of any "count or method of making estimates cannot be determined absolutely, because there is no standard with which to compare.

Even if a perfect standardization of the plate method results in securing ideally uniform counts in duplicate samples of milk, this does not prove the counts to be accurate. There may be a uniform error common to all the samples. Such an error, for example, might be caused by the existence of clumps of bacteria in the milk. Also, uniform counts from duplicate samples do not prove that the same relative error exists thruout a series of counts made from different samples of milk. The count on one sample may be affected by one type of error and the count on a second by another type. For instance, one sample may contain a large number of bacteria not adapted to growing on agar while another sample may contain approximately the same number of bacteria all of which are capable of growing on agar. Tho the percentage error may be uniform in duplicate samples of milk it can be readily understood how the percentage error can be widely divergent in different samples.

Altho the number of bacteria in milk can be stated as estimates only, it is possible to learn something of the probable accuracy of these estimates when made by two or more different methods. The direct microscopic method as devised by Breed, and the plate method used simultaneously on the same samples of milk furnish a comparatively simple means for making such comparisons.

Such a study has been carried on at this Station in two ways. First, coöperating with the Department of Dairy Industry at Cornell University, comparative analyses have been made by several men on a small number of duplicate milk samples. A preliminary

3 Breed, R. S. The determination of the number of bacteria in milk by direct microscopical examination. Centbl. Bakt., Abth. II, 30:337-340. 1911.

account of this work is reported in the Journal of Dairy Science.1 Second, comparative studies have been made at this Station on miscellaneous market milk samples. The first report of the latter work was a preliminary one submitted in 1914.5 A second bulletin described the technique after it had been further developed in routine work. The present bulletin contains a report on 643 comparative counts made by both the plate and microscopic methods.

PLAN OF WORK.

The milk selected for this investigation was fresh, unpasteurized market milk from the Geneva city milk supply. Each sample was taken by means of a sterile aluminum pipette which was long enough to reach to the bottom of a 40-quart can. The night milk was always older than 12 hours and usually less than 16 hours old, while the morning milk was rarely older than four hours. The samples were packed in ice as soon as taken. It required approximately four hours to take the samples, bring them to the laboratory and to plate them.

Plating method. The medium used for plating the first 491 samples was made from meat extract, peptone and agar according to the following formula:

The acidity of each lot of medium was determined by direct titration and was found to be very constant. The percentage acidity to phenolphthalein ranged from 0.8 to 1.1 and no attempt was made to adjust the reaction. The H-ion concentration of the medium made according to this formula, as determined by the colorometric method, has been found to be between PH 7.0 and PH = 6.5.

=

Ten grams of lactose per liter were added to the medium used for plating the remaining 152 samples.

4 Breed, R. S., and Stocking, W. A., Jr. A preliminary report on a series of cooperative bacterial analyses of milk. Jour. Dairy Sci., 1:19–35. 1917.

.5

Brew, J. D. A comparison of the microscopical method and the plate method of counting bacteria in milk. N. Y. Agr. Exp. Sta. Bul. 373, pp. 1-38, 1 pl., 2 figs. 1914.

6 Breed, R. S., and Brew, J. D. Counting bacteria by means of the microscope. N. Y. Agr. Exp. Sta. Tech. Bul. 49, pp. 1–31, 5 figs. 1916.

Method of plating.-Three dilutions (1/100, 1/1000, 1/10,000) were used in all cases and each dilution was plated in triplicate, Each sample of milk was shaken vigorously 50 times and each diluted sample 25 times. Care was taken to keep the interval of time elapsing between the making of the dilutions and the pouring of the plates less than 20 minutes.

Incubation.—All plates were incubated at 21° C. for five days and then counted with the aid of a magnifying lens. They were then incubated two additional days at 37° C. and recounted. The plate counts herein reported are the higher of the two, but since there was generally a slight increase in the number of colonies after the additional incubation at 37° nearly all of the counts reported were those obtained after this incubation.

Counting plates. In selecting the plates for counting, only those were chosen which had more than 20 and less than 400 colonies." There were necessarily a few exceptions which have been noted in the tables. In only 10 cases were all plates so crowded as to make it necessary to estimate the number of colonies on each. Triplicate platings in three dilutions made it possible to secure a final count which was an average usually from three or more plates. All of those cases in which it was impossible to count more than one plate have been noted in the tables.

Microscopic method. As soon as the dilutions for plating were completed, preparations were made for microscopic examination according to the procedure described in Technical Bulletin No. 49.8

The microscope was adjusted so as to use a 300,000 factor for computing the number of bacteria found in the microscopic fields in terms of bacteria per cubic centimeter. In order to obtain this factor with the microscope at hand a 5x ocular, and a 1.9 mm. oil immersion objective were used, and the draw tube was placed at 170. This gave a microscopic field .205 millimeters in diameter.

Counting. In order to secure as great uniformity as possible in microscopic counts, certain definite rules were observed which may be stated as follows:

1. The bacteria in 100 microscopic fields were counted in all cases where they were few in number. With a 300,000 factor this

7 Breed, R. S., and Dotterrer, W. D. The number of colonies allowable on satisfactory agar plates. N. Y. Agr. Exp. Sta. Tech. Bul. 53, pp. 1-11. 1916.

8 See footnote 6.