Conditional Recall and Bias in Memory Experiments

Abstract

The use of recall measures conditionalized on other performance introduces the possibility of bias due to item and/or subject selection. Several possible cases are considered. While conditional recall measures are not impossible to interpret, there are restrictions that must be observed. Progress in the understanding of psychological phenomena usually results from the innovation of new analytic procedures that more clearly throw the phenomenon in question into focus. Most theories of verbal learning are concerned with processes presumed to act on the storage and retrieval of the individual verbal unit, while data analysis still most often depends upon gross measures of list acquisition, retention, and transfer. These measures may be quite insensitive to the processes of major theoretical interest, since performance is typically averaged across list items and subjects. A common means of conducting a more fine grained analysis is to conditionalize recall on some other performance measure. That is, the individual items are classified according to some performance criterion, then the items in the various classes are compared on some other performance criterion. For example, the subjects may be given a recognition test for the stimulus terms on a list of paired associates and then asked to recall a response to the stimuli (Martin, 1967a). The probability of a correct recall is then measured only for items for which correct recognition occurs.