Although stocking density is perceived as a topic of major importance, no consensus has been reached on what density would allow for good welfare. In the present study, the welfare of 4 replicates of birds stocked at 8, 19, 29, 40, 45, 51, 61, and 72 broilers per pen (or 6, 15, 23, 33, 35, 41, 47, and 56 kg actually achieved BW/m(2)) was studied using 6 welfare indicators. Density did not affect bursa weight, mortality, or concentrations of corticosterone metabolites in droppings but did influence leg health (P = 0.015) and footpad and hock dermatitis (P < 0.001) and tended to influence fearfulness (P = 0.078). However, not every increase in density or group size, or both, led to poorer welfare for the affected indicators: leg health and fearfulness showed unexpected peaks at intermediate densities. Furthermore, the indicators were influenced at different densities: leg strength showed a steep decrease from 6 to 23 kg/m(2), hock dermatitis rose from 35 to 56 kg/m(2), and footpad dermatitis and fearfulness were only significantly higher at the highest density of 56 kg/m(2). No threshold stocking density above which all aspects of welfare were suddenly altered was found in this study. Instead, different aspects of welfare were influenced at different densities or group sizes, or both. Thus, evaluating the effects of stocking density on welfare as a whole would require either identification of acceptable levels for each separate indicator or a weighting of the indicators in an integrated welfare score. A tentative attempt to such an integration, made using equal weights for all parameters, showed a decrease in welfare as density increased (P < 0.001). The lowest 2 densities (6 and 15 kg/m(2)) scored better than most middle densities (23, 33, 35, and 47 kg/m(2)), whereas all densities scored better than the highest density (56 kg/m(2)).