An experimental evaluation of HI-BET (heterosis inheritance backcross evaluation technique) involving the comparison of the performance of F3 and backcross (BC) families with common sires is described. These crossbred populations were derived from an F1 cross between Australorp and White Leghorn fowls with substantial heterosis for egg production variables. There was considerable evidence for the presence of a statistically significant sire family by population type interaction for survivor's total egg mass. The ranking of sires varied substantially depending upon whether they were being assessed on the performance of their F3 or their BC daughters. The BC population yielded substantially higher sire component heritability estimates for egg weight, egg production and total egg mass than did the F3 population. For the latter two traits, this improvement in the correlation between genotype and phenotype is in accordance with the expectations of HI-BET. It is concluded the HI-BET should be effective strategy for selectively focussing upon the non-additive genes that are only present in one ancestral line at a high frequency, whilst ignoring non-additive genes unique to the other ancestral line.