The effect of 6-, 12-, 18-, 24-, 30-and 36-h milking intervals on milk secretion has been investigated. The effect of the previous milking intervals was eliminated by interposing a recovery period consisting of 12-h milking intervals between the experimental intervals. Bias due to the carry-over of residual milk was reduced by removing the residual milk at the end of each milking, after an injection of oxytocin.The rate of secretion of milk and of the individual constituents decreased curvilinearly, with duration of the interval, but the degree of curvilinearity differed between constituents. The rates of decrease were in the following increasing order: sodium, chloride, fat, whey proteins, casein N, water, non-protein N, lactose and potassium. The effects of the treatments persisted for some time after the end of the experimental intervals, but the original rates of secretion were regained by the end of the recovery period. The differing effects on the secretion of individual constituents resulted in increases in the concentrations of sodium, chloride and whey proteins in the milk and decreases in those of lactose and potassium.The earliest work on the rate of secretion of milk was carried out by Ragsdale, Turner & Brody (1924), who concluded that there was a progressive decline in the rate of milk secretion with increase in the length of milking interval, and that there was an even more marked effect on the secretion of fat. In the interpretation of the results the authors assumed that the milk obtained at a milking was that which had been secreted in the previous interval. Johansson (1940) demonstrated, however, that there is a considerable carry-over of milk and of fat from one milking interval to the next. He found that the carry-over of residual fat was in direct proportion to the amount of milk in the udder before milking. This was confirmed by Bailey, Clough & Dodd (1955), and Turner (1953, 1955 showed that it was true also for residual milk. In the type of experiment devised by Ragsdale et al. (1924) the carryover would therefore bias the results so that the quantity of milk removed from the udder after short intervals would be greater than that secreted during the interval and, conversely, after a long milking interval the quantity removed would be less than the amount secreted. The effects would be even more marked for fat than for milk yield.• Present address: Department of Agriculture, The University, Leeds.