Several errors occurred in this paper by Williams and others (13 March 1999, pp 693-8). Because earlier versions of the figures were inadvertently submitted with the final manuscript, some of the data are incorrect. Figure 1 should have shown that 42 [not 41] infants who received iron supplemented formula milk remained in the study. Figures 2 and 3 should have shown that, due to missing data, comparative studies were carried out on 38 pairs of observations [not individual] in infants receiving unmodified cows' milk; correct versions of figures 2 and 3 appear below. In the results section of the abstract the 95% confidence intervals for personal and social scores should have read 1.2 to 16.8 [not −5.4 to 17.2]. The conclusions remain unchanged.