John W. French has recently suggested an index for keying "multiple-choice tests where no 'a priori' key exists" and in the discussion has made brief reference to its relationship to the point biserial correlation, mentioned by others and often used for the same purpose. This memorandum attempts to point up the relationship more clearly and comments on some of the relative merits of the two indices. (It is assumed that the reader has seen French's article and a complete review of its contents is given not given here.) It is concluded that no rigorous justification for use of either index (French's or the biserial correlation) can be given. Practically speaking, where the item p's are mostly moderate or are nearly all the same, the items keyed by the two methods will probably not vary greatly. However, if the investigator is at the stage where he has little notion of how to key the items the added meaningfulness of a correlation coefficient should prove to be of particular value.