Figures showing the means and the standard deviation of butterfat percentage of milk samples, at the afternoon and morning milkings res-pectively, during the four quarters of the year are presented.

The probable sources of error due to the nature of the sample are discussed and the seasonal variability of mean and of standard deviation is considered.The likelihood of the mean of afternoon or of morning milk failing to reach certain specified minima during the respective seasons is calculated.

This paper contains the results obtained in the fourth and fifth, years of a study of the milk of a typical herd of healthy Shorthorn cows. The period of drought in the summer of 1933 is included and furnishes analyses which differ so much from the four previous years that the daily figures are given in Chart 1. In the earlier paper (i) weekly averages only could be published. The greatest seasonal variations which have been demonstrated by the daily analyses are:

(1) A fall of the percentage of fat in the morning's milk in the spring when the cows were on young grass. This was clearly shown in each of the five years.

(2) A fall in the percentage of solids-not-fat in the evening's and morning's milk during the period of drought in 1933, to an extent which had not previously been recorded.

The causes of these changes in the composition of milk will be further studied using the milk of individual cows.

The weekly analyses of morning's milk have also shown seasonal variations.

Relatively high values of the titratable acidity, formol titration, ash, calcium and phosphorus were observed from October 1931 to February 1932, followed by a general decrease in the summer months, while in the late autumn somewhat higher values were again found.

During the summer and early autumn of 1933 values which were lower than any previously obtained during this investigation were recorded for formol titration and for phosphorus, while the time of coagulation with rennet was longer.

Preliminary consideration of the data has revealed the possibility that the fall in non-fatty solids may be particularly associated with the phosphorus content.

The writer seeks to establish the interrelationship and variability of the three tests used for gauging the hygienic quality of milk, plate count, coliform content and keeping quality. Keeping quality is denned, and after discussing previous work and source of material, and noting the technique followed, the method of classification is described. Plate count and coliform content are treated logarithmically. Treatment of keeping-quality results presents especial difficulties, and the test is described in detail to disclose these, and reasons given for tabulating by half-days. The relationship between plate count and keeping quality is linear, but non-linearity occurs in those relationships which concern coliform content. This is because the first group “absent from 1 ml.” is not truly a definite class, but contains samples of superior quality. It is shown that an increase of one “stage” of coliform contamination reduces the average keeping quality as much as an increase of 0–54 in the logarithm of the plate count (approximately seven times). Variability of each term is discussed. The standard deviation of keeping quality is of the order of 1 half-day, that of logarithm of plate count is 0·77 and that of logarithm of coliform content in excess of unity. These variabilities are too great to permit of reasonable forecast of one term from the other two.

The isolation from cow's milk of an organism allied to Actinomyces farcinicus merits recording on account of the rarity in this country of such organisms and on account of failure to find in the literature a description of the organism.

The various methods tested in this work have revealed the criteria of utility, U, shown in the following table for the detection of added water as measured by for the “difference” modes, and for the ratio modes by , where M3/M4 is the ratio of the maximum and minimum seasonal values of M. All the figures are for 10 per cent, added water. Utility values equal to and less than unity in the table are useless.

The titratable acidity of fresh milk has been studied by several investigators, but the effects of the various buffers present have not been precisely defined. Early statements as to the cause of the titratable acidity were expressions of opinion unsupported by experimental evidence; the theories of indicators and buffer action were not considered. The acids and bases present were calculated by arbitrary rules to various combinations of salts, and the acidity was stated to be due to the presence of acid salts in these hypothetical mixtures.

The previous section of this study in which the titration curves of various milk constituents were discussed approximately denned their influence on the titration. Some of the conclusions arrived at are amplified and more clearly defined when the buffer index of the curves is considered.

The reactions which occur when sour cream is “neutralised” with calcium hydroxide and sodium bicarbonate are discussed from the point of view of the acid-base equilibria.

Experimental results are recorded showing that:

1. When lime is the neutraliser, pasteurisation has little effect on the pH. or titratable acidity of the neutralised cream.

2. The reduction in acidity when lime is used is shown to vary from that calculated from the sodium hydroxide titration to an extent which depends on the pH of the neutralised cream. The reasons for and extent of this variation are defined.

3. When sodium bicarbonate is used as a neutraliser the loss of carbon dioxide (which is influenced by the method of pasteurisation adopted) has most influence on the degree of neutralisation obtained and the experiments show the extent of this at various acidities.

The variations in the pH-titratable acidity relationship of creams neutralised with calcium hydroxide and sodium bicarbonate are defined.