First, second and third lactation curves of 21,885 first lactation dairy cows were allowed to vary depending on the months in milk at which the cow became pregnant. The curves were modelled using a 4th degree polynomial of days in milk. In the models for second and third lactation daily yield, days open in the previous lactation was included as a quadratic regression. Allowing the curve coefficients to depend on the month in milk when pregnancy occurred significantly improved the goodness of fit of the model to the data. The linear and quadratic coefficients of second and third lactation daily milk yield on days open in first and second lactation, respectively, were also significant. We suggest that, in the evaluation of test day milk yield, the shape of lactation curves are allowed to vary depending on the months in milk when pregnancy occurs and that days open in the previous lactation, if appropriate, is included in the model.