Summary In north-eastern New South Wales (NSW) the Regional Forest Agreement process has transferred more than 400 000 ha of state forests to national park, and restricted silviculture to ‘single tree selection’ and a light form of ‘Australian group selection’. While these silvicultural systems are theoretically well suited to ecologically sustainable forest management, there is concern that in their current form they are not achieving adequate regeneration or optimising the growth of that regeneration. This is of particular concern for mixed-species blackbutt forest, for which there is no quantitative research concerning the growth and composition of regeneration within group-selection gaps. We address this issue by: (1) quantifying the effect of gap size, and other gap characteristics including distance from gap edge, on the growth of regeneration; and (2) assessing the effect of gap size on the composition of regeneration. We use the answers to these questions to recommend a gap size for group selection silviculture in mixed-species blackbutt forests in north-eastern NSW. We measured attributes describing the growth and composition of regeneration in nine circular group-selection gaps in mixed—species blackbutt forest near Coffs Harbour and Wauchope. These gaps contained 14.5–15.5-y-old regeneration and provided three replicates of small (0.27–0.3 ha), medium (0.45–0.67 ha) and large (0.93–0.97 ha) gaps. ANOVA testing indicated significantly (P < 0.05) lower height, diameter and volume growth of dominant blackbutt stems up to five metres from gap edge. Outside this zone growth remained fairly constant, indicating dominant blackbutt trees were susceptible to suppression only in close proximity to gap edges. Multiple regression analysis confirmed the relatively short distance from gap edges over which suppression occurred, with distance to closest gap edge explaining a small proportion of the variation in the models fitted for tree—and plot-level growth. The origin of blackbutt regeneration within gaps was a significant effect in tree-level growth models, with planted stems having increased diameter and volume growth compared with stems regenerated from natural seedfall. Gap size had no significant effect on the composition of regeneration. We conclude that for the range of gaps tested, 1-ha gaps are optimal for growth because they minimise the proportion of gap within 5 m of the retained forest edge, without altering composition. Larger gaps have also been shown to have operational and economic benefits compared with smaller gaps.