The effect of pregnancy on milk yield was studied on a 601 lactation sample. One hundred and sixty-nine lactations of non-pregnant cows, managed tinder identical conditions, served as controls. Lactations were divided into five groups according to parity (primiparous or multiparous) and production potential (two ranks for potential primiparous cows, three ranks for multiparous cows). The difference between the individual milk yield of each pregnant animal, and the mean production of the corresponding controls was computed weekly from the week of conception. The reducing effect of pregnancy began to be measurable from the 20th week of pregnancy, regardless of the group studied or the week of conception. Analysing the amplitude of this effect restricted the investigation to the three following groups: (i) primiparous cows, (ii) low- and medium-producing multiparous cows, (iii) high-producing multiparous cows. The effect of pregnancy was lower in primiparous than in multiparous cows. In the fatter, it was higher in high-producing cows, partly because of the greater calf weight. At the 29th pregnancy week, this effect reached -1 . 5, -2 . 4 and -3 . 6 kg/day in the three groups, respectively. In each of these three lactation groups, the following non-linear model Y = -e(-a)[(Pw - 18)e(-bPw)] was fitted, where Y is the difference between the milk yields of pregnant and non-pregnant cotes at a given stage (week), Pw is the pregnancy week, and a and b are parameters. The overall milk yield losses over a complete lactation, as estimated from the integral of that equation, reached 89, 137 and 203 kg, respectively in the three groups