The results of a number of experiments carried out at Hillsborough on the growth of chickens, covering two seasons and involving 45 groups of 30 to 100 chickens, are discussed with particular reference to the variability of the live weights. It is concluded that the weights of cockerels or pullets taken separately should not show a standard deviation very much greater than 15 per cent. of the mean weight for the group, unless the experimental treatment or other conditions have seriously interfered with the rate of growth. It is pointed out that comparisons of mean weights and of standard deviations may be misleading unless cockerels and pullets are considered separately. Smallest significant differences between the mean weights of groups are tabulated for standard deviations of 15 per cent., 17·5 per cent, and 20 per cent., and for groups of 10 to 100 birds.