The article presents results from two series of milk samples where samples from individual cows were collected at each milking over a period of two weeks. Fat content, protein content and cell count were determined for each sample. The cell counts (in thousands per ml) were transformed to logaritmic values and statistically analysed. As much as 75 to 80% of the variation (expressed as proportion of the total sums of squares) was due to differences "among cows" and about 3% due to differences "among days" during the sampling period. Samples from morning milkings had about 20% lower cell counts than samples from afternoon milkings. The standard deviation among samples from two consecutive milkings of the same cow (after removal of systematic differences between days and between morning and afternoon milkings) was computed as about 30% of the mean (i.e. coefficient of variation approximately 0.3). By combining samples from two milkings the random "error" in cell count can be reduced to about 20% of the mean. The samples exhibited a highly significant, but nummerically small, positive within-cow correlation between fat content and cell count, and between protein content and cell count, while the corresponding correlation between milk yield and cell count was slightly negative (non-significant).