The formula and derivation are given for the sampling variance of the "random-halves" reliability coefficient when items, rather than examinees, are sampled. When the number of test items is large, this sampling variance is found to be larger than the Kuder-Richardson formula-20 coefficient by a factor equal to the number of test items. The usual matched-halves reliability coefficient will likewise usually have a large sampling variance, since it likewise is based on an arbitrary split of the test items. An adaptation of the Jackson-Ferguson "battery reliability coefficient," used as an estimate of the matched-forms reliability of the test, avoids this relatively large sampling variance.