Tables of confidence intervals for a proportion based on the sample proportion are presented, calculated by a slight modification of the method proposed by Sterne (1954), for fixed sample sizes up to 30 and confidence coefficients of 0 90, 0 95 and 0 99. This system is compared, especially in shortness, with Sterne's system, Clopper & Pearson's (1934), and another, intermediate system. It is assumed that a random sample of fixed size n is drawn from an infinite population containing a proportion oT of individuals with a given characteristic, that r individuals are observed to have the characteristic, and that it is desired to estimate 7 by means of a two-sided confidence interval. While relatively tedious to calculate, Sterne's system and its modification have in common the advantage that no system is possible which has shorter total length of the n + 1 confidence intervals for r = 0, 1, .. ., n. However, they are at a disadvantage if one