In an effort to quantify herd-year-season x sire interaction in Bonsmara ca
ttle, four different models were fitted to the data used in Neser at al. (1
996). The herds were also divided into two (bushveld and sour grassveld) re
gions in order to estimate the effect of region x sire interaction. Heritab
ility estimates from herd-year-season x sire and year-season x sire interac
tion models were more or less equal (0.138 vs 0.136), while estimates of he
ritability of herd x sire and region x sire models were considerably lower
(0.080 and 0.060). The separate analysis of the regions yielded heritabilit
y estimates similar to those obtained with the model which included herd-ye
ar-season x sire interaction (0.142 and 0.166). The estimate of the interac
tion as a proportion of the total variance remained more or less constant i
n the models which included year-season x sire, herd x sire and region x si
re interaction (0.064, 0.061 and 0.059). It was, however, slightly lower th
an the model which included herd-year-season x sire interaction (0.084). Th
e product moment correlation between the breeding values of sires in the tw
o different regions was only 4.59% in the separate analysis of the two regi
ons compared to 32.74% when the two regions were analysed together. This lo
w correlation would indicate that separate analyses of regions should be co
nsidered. Furthermore, testing of sires for stability of their progeny's pe
rformance across regions, should receive attention.