How many dimensions underlie judgments of learning and recall redux: Consideration of recall latency reveals a previously hidden nonmonotonicity

Abstract

Jang and Nelson (2005) used state-trace analysis to examine factors that affect judgments of learning (JOLs) given as a prediction of future cued recall success. Koriat’s (1997) cue-utilization framework predicted that intrinsic cues (e.g., item difficulty) would have approximately the same effects on recall as they would have on JOLs whereas extrinsic cues (e.g., number of presentations) would have greater effects on recall than on JOLs. In contradiction to the prediction from the cue-utilization framework, Jang and Nelson repeatedly found a monotonic state-trace solution, suggesting that a single latent variable (e.g., memory strength) explained both JOLs and recall. However, performance can be measured in many ways, and dissociations between JOLs and recall may arise from factors other than intrinsic and extrinsic cues. Thus, an apparent monotonic solution could be an artifact of the particular choice of behavioral measures or experimental manipulations. In light of this possibility, we reanalyzed Jang and Nelson’s data and conducted a new experiment, considering recall latency as well as recall accuracy and including the manipulation of immediate versus delayed JOLs. Even when additionally including both immediate- and delayed-JOL conditions, state-trace analysis with JOL magnitude and recall accuracy generally suggested a single latent variable, except for a single case in which immediate JOLs produced high overconfidence. However, the state-trace results with JOL magnitude and recall latency primarily revealed a nonmonotonic function, indicating that more than one latent variable is needed to explain the relationship between JOLs and recall.