The deceptive nature of associative word pairs: the effects of associative direction on judgments of learning

The accuracy of judgments of learning (JOLs) in forecasting later recall of cue–target pairs is sensitive to associative direction. JOLs are generally well calibrated for forward associative pairs (e.g., credit-card), but recall accuracy is often overestimated for backward pairs (e.g., card-credit). The present study further examines the effect of associative direction on JOL accuracy by comparing forward and backward pairs to unrelated pairs and symmetrical associates (e.g., salt–pepper)—a novel comparison. The correspondence between initial JOLs and recall accuracy was examined when study was either self-paced with concurrent JOLs (Experiment 1), when study/JOL duration was equated across pair types (Experiment 2), when JOLs were made immediately following study (Experiment 3), and when JOLs were made after a delay (Experiment 4). Across experiments, JOLs accurately estimated correct recall for forward pairs, but overestimated recall for symmetrical, backward, and unrelated pairs—an overestimation that was particularly robust for backward pairs. Calibration plots depicting JOL ratings against their corresponding recall accuracy indicated overestimations occurred for all pair types, though overestimations only occurred at high JOL ratings for symmetrical and forward pairs, a qualitative difference that was not captured in standard analyses of mean JOL and recall rates.