The estimation of covariance components for daily milk yield, fat and protein content was performed in three Slovenian dairy sheep breeds (Bovec, Improved Bovec, and Istrian Pramenka). In the period 1994–2002, 38 983 test-day records of 3 068 ewes were collected according to ICAR regulations (method A4). All the available relationships between animals were considered. For that reason, information on 3 534 animals was included. Test-day records were analysed by a multiple-trait repeatability animal model. In its fixed part, the model contained breed and season of lambing as classes. Days after lambing, parity, and litter size were treated as covariates. Days after lambing were modelled with modified Ali-Schaeffer’s lactation curve, parity with quadratic, and litter size with linear regression. The random part of the model consisted of flock-test month effect, additive genetic effect, permanent environment effect over lactations, and permanent environment effect within lactation. Covariance components were estimated using the restricted maximum likelihood method (REML). The estimated heritabilities were 0.11 for daily milk yield, 0.08 for fat content, and 0.10 for protein content. A relatively high variance ratio for all milk traits was explained by the flock-test month effect (from 0.27 for daily milk yield to 0.57 for protein content), while ratios explained by both permanent environment effects were lower (up to 0.13). Additive genetic correlations between daily milk yield and fat content, and daily milk yield and protein content were negative and similar (–0.36 and –0.37). A high and positive (0.67) additive genetic correlation between fat and protein content was found. Correlations for environmental effects showed a pattern similar to additive genetic correlations. Genetic parameters estimated in Slovenian dairy sheep showed that genetic progress in milk traits could be achieved using test-day milk records.