1. An experimental evaluation of Robertson's (1960) theory of limits in artificial selection was attempted. A number of lines from the Canberra base population were selected for abdominal bristle number over 50 generations with population sizes of 10, 20, and 40 pairs of parents and selection intensities of 10, 20 and 40% as well as unselected controls.
2. In general, the total response obtained increased with an increase in Nī (product of population size and standardized selection differential).
3. Thus, total response increased with increase in the number of individuals scored, or, for a fixed number of parents, increase in selection intensity increased both rates of response per generation and total response.
4. But for the same total number scored, the response increased as selection intensity decreased. However, the proportion selected had only a small effect as compared with that of the total number scored.
5. Sublines in which the population size was reduced after 16 generations of selection but with the selection intensity kept constant, immediately fell behind their parent lines and gave much less response.
6. Agreement between replicate lines was generally poor, particularly for the 10- and 20-pair lines.
7. Patterns of response in individual lines were frequently irregular and ‘waves of response’ were not uncommon.
8. The results are discussed in terms of several theoretical models of selection limits. In general, agreement with these models was poor, as much of the response appeared to be due to a few genes (or gene combinations) with large effects on bristle number.