Reliability of Learning and Recall Scores as Affected by Length and Difficulty of Lists

Abstract

The effects of length of list, level of difficulty, and method of determination upon the reliability estimates of learning and recall scores (CVC trigrams) were investigated in the paired-associate situation. For all conditions estimates of reliability were obtained by the equivalent-form and split-half methods. The two methods gave fairly comparable results for the reliability of the learning scores on lists of moderate length. The obtained reliability estimates tended to increase with increased list length, although these did not increase in the manner predicted by the Spearman-Brown prophecy formula. The reliabilities of the recall scores were low, but they did increase with increased length of list more than did the learning scores. Means of learning scores increased with practice while those of recall scores decreased. Means of recall scores were relatively greater for the longer lists than for the shorter ones.