The experimenter who desires an interval estimate of a parameter should take into consideration two very fundamental questions: (1) Does the interval contain the parameter? (2) Is the interval too wide? In practice these questions may be answered in the form of probability statements. Let the confidence coefficient, 1 a, denote the probability the interval will contain the parameter and let the width coefficient, O3, denote the probability that w, the width of the interval, will be less then a specified number, d. Usually 1 a is specified, a confidence interval is computed, and the width is then governed greatly by the sample size, n. The larger n becomes, the more precise the estimator becomes; hence a smaller interval. Consequently, the larger the sample size, the larger 32 becomes. The crux of the problem may now be stated: Determine the sample size, n, necessary for (A) a 1 a confidence interval and (B) a 32 width coefficient. In order to provide an answer some a priori information is needed. This suggests a two step procedure. The purpose of this paper is to utilize a two step method to determine the size of the samples from two independent normal populations necessary to satisfy (A) and (B) when obtaining an interval estimate on the ratio of variances. The technique requires taking preliminary samples of sizes ml and mi2, making a simple calculation after specifying a, A, and d, and looking for the sizes of n, and n? in Table I to insure obtaining the desired results. Table I gives choices of n, and n2 for I a and 3 = .90, .95, and .99. It is of importance to note that there is not a unique solution for n1 and n, but in general n, + n2 can be minimized without much difficulty.2 Also it is important to bear in mind that the width co-