Three sets of data, standardized 120-day milk yield, corresponding test-day yields and the first two test-day yields of Churra ewes were considered to estimate genetic and environmental parameters of milk yield. A total of 22599 lactation milk yields and their corresponding 84488 test-day yields for 12785 ewes, collected between 1992 and 1994, were used to estimate environmental parameters employing fixed models. A subset including data from the daughters of the rams participating in the breeding program was used for the estimation of genetic parameters (3379 lactations and their corresponding 12662 test-day yields for 2102 ewes, the daughters of 121 rams). Heritabilities and repeatablities were estimated by derivative free restricted maximum likelihood and breeding values of the rams were estimated by a repeatability animal model. For test-day milk yield, the environmental effects of the fixed model, including flock test-date (FTD), age at lambing, type of birth, and linear, quadratic, and cubic coefficients of days in milk and their inverses, were all highly significant (P < 0.001). For 120-day milk yield, flock-year-season, age at lambing, and type of birth all had highly significant effects (P < 0.001). However, interval from lambing to first test had a less significant effect (P < 0.05). The heritabilities estimated for all test-day yields, the first two test-day yields and 120-day milk yields were 0.14, 0.15 and 0.18, respectively. These heritabilities were all low, mainly because of the higher phenotypic and residual variances obtained. The corresponding repeatability estimates for the same traits were 0.44, 0.46 and 0.42, respectively. Product moment and rank correlations between breeding values of rams based on 120-day milk yield and those based on all test-day milk yields and the first two test-day yields ranged from 0.77 to 0.88. Product moment and rank correlations between evaluations based on all test-day yields and on the first two test-day yields were 0.94 and 0.92, respectively. FTD models presented a possible alternative to models fitting standardized yield records. (C) 1998 Elsevier Science B.V.