Twelve models were fitted to early growth data of two Swiss sheep breeds to investigate their suitability for evaluation of breeding values. Models were identical for fixed parity, litter size, sex and lambing season effects, random flock-year, and direct genetic effects but differed for combinations of random litter, permanent environmental, ram-flock, and maternal genetic effects. Records of average daily gain to 30 d of 25,564 lambs of the Black-Brown Mountain Sheep (SBS) and of 26,391 lambs of the White Alpine Sheep (WAS) born 1989 to 1995 and their pedigrees were available. A single-trait animal model was fitted by the restricted maximum likelihood method. The information criterion of a particular model (i.e., the maximum of the likelihood function adjusted for the number of independently estimated parameters) was used to evaluate the models for their fitting power. The litter effect accounted for between 26 and 31% of the phenotypic variance, with little variation within breed. Models containing the ram-flock effect provided a better fit of the data than otherwise identical models. This effect contributed 6 and 4% to the phenotypic variance in the two breeds and strongly influenced estimates of other components. The proportion of phenotypic variance due to the flock-year effect was 23 and 25% without and 19 and 23% including the ram-flock effect in the model in the two breeds. Including permanent environmental effect of the ewe in addition to litter effect led to a better fit of the data. Depending on the model, it then contributed between 3 and 6% to the phenotypic variance. Fitting the ram-flock effect reduced heritability considerably and increased the breed difference of the estimates of this parameter. Estimates ranging from .16 to .10 and from .08 to .14 were obtained for the SBS and WAS breeds, respectively. For models without the ram-flock effect, negative estimates of the direct-maternal correlation of between -.38 and -.45 were observed. Including the ram-flock effect reduced this correlation substantially to between -.08 and -.17. Including the direct-maternal covariance in addition to the ram-flock effect did not improve the fit any further in either breed. Ranking of the models investigated differed between breeds, but the same model provided the best fit. It contained the random litter, permanent environmental, ram-flock, direct, and maternal genetic effects, but not the covariance between the last two.