Weaning weight records of 24 758 Bonsmara calves produced by 503 sires in 30 herds, from 1980 to 1993, were used to examine the importance of the inclusion of herd-year-season by sire interaction in the estimation of genetic parameters. Three separate models were used in the DFREML analysis of the data. In the first, permanent maternal environment was included as an additional random factor, while in the second, herd-year-season by sire interaction was included, also as an additional random factor. Both these factors were included as additional random factors in the third model. The estimates of the (co)variance components and heritabilities for direct additive (a), maternal additive (m) and additional random factor (c) for the three models analysed, were as follows:(σ2a: 138.04, 65.23, 67.46; σ2m: 84.77, 143.80, 66.83; (σam: -57.47, -28.24, -21.97; (σ2perm: 73.72, 71.01;( σ2HYSxS: 41.79, 41.39; (σ2e: 251.31, 274.78, 265.97; (σ2P: 490.37, 497.35, 490.74; h2a: 0.28,0.13,0.14; h2m: 0. 17, 0.29, 0.14). The results indicate the importance of both permanent maternal environment and herd-year-season by sire interaction in the estimation of the genetic parameters. The exclusion of herd-year-season by sire interaction could lead to a serious overestimation of the direct additive components. The same applies for permanent maternal environment. The exclusion of this factor could lead to a serious overestimation of the maternal components. Both these factors should therefore be included in the across-herd weaning weight analysis of Bonsmara cattle.