An analysis of variance approach to estimating reliability is examined. This approach uses an internal-consistency index for estimating the reliability of survey instruments containing ranked items, and is recommended when the relative ranking of item means is of interest to survey researchers. The computation of this index is demonstrated using data from three American College Testing Program surveys: Student Opinion Survey, Entering Student Survey, and Adult Learner Needs Assessment Survey. Eighteen of the 250 postsecondary institutions that utilized these surveys between 1980 and 1983 were studied. Survey records were analyzed using a coefficient reliability procedure available through the Statistical Package for the Social Sciences. The reliability estimate requires little computation, since the mean square values necessary to calculate the estimate may be obtained directly from analysis of variance tables. Since the approach is based on the internal-consistency of raters' responses, only one administration of the instrument is necessary. The major problem in using the proposed approach centers on the treatment of missing data. A bibliography is appended. (SW)