Guess quality moderates how semantic relatedness influences the pretesting effect

Students are often taught that testing themselves will help them retain new material. Less well known, however, is that testing oneself before learning can also produce benefits (the “pretesting effect”). Prior studies have indicated that the semantic relatedness between the provided information (cue) and the answer (target) may be critical for this effect. Here, we investigated this by measuring the trifold relationship between cue, target, and guess using continuous relatedness measures. In Experiment 1, participants learned cue-target word pairs by studying or generating guesses based on the cue word before taking a test on both cue-target pairs and the original cue-guess combinations. We found a robust difference in accuracy favoring pretesting over studying, regardless of cue-target relatedness. In Experiment 2, we further examined if errorless generation contributed to the pretesting effect by adding a condition in which cue and target were present, but subjects still generated a related word. Cue-target memory was better in both the pretesting and generation conditions than the study condition. In both studies, while cue-target relatedness benefitted accuracy overall, these benefits did not interact with pretesting or generation pair conditions. Importantly, we observed an interaction between the cue-target semantic relatedness and guess-target semantic relatedness only in the pretesting condition, demonstrating a benefit for (incidentally) high relatedness between guesses and targets when cue-target relatedness was low and an impairment when cue-target relatedness was high, likely reflecting interference between target and guess. Our results illuminate the importance of considering guess quality in uncovering the mechanisms in pretesting.