The role of mediators for the pretesting effect

ABSTRACT

Taking a pretest (e.g., smoke – ?) before material is studied (smoke – fog) can improve later recall of that material, compared to material which was initially only studied. The goal of the present study was to evaluate for this pretesting effect the potential role of semantic mediators, i.e., of unstudied information that is semantically related to the study material. In all three experiments, subjects studied weakly associated word pairs (e.g., smoke – fog), half of which received a pretest. Subjects then either completed a recognition test (Experiment 1) or a cued-recall test (Experiments 2 and 3), during which they were presented with both the original study material and never-before-seen semantic mediators that were strongly related to the cue item of a pair (e.g., cigarette). Strikingly, presenting semantic mediators as lures led to higher false alarm rates for mediators following initial pretesting than study only (Experiment 1), and presenting semantic mediators as retrieval cues led to better recall of target items following pretesting than study only (Experiments 2 and 3). We argue that these findings support the elaboration account of the pretesting effect but are difficult to reconcile with other prominent accounts of the effect.

KEYWORDS:

• Testing effect
• pretesting effect
• elaboration
• mediator

Open Scholarship

This article has earned the Center for Open Science badges for Open Data and Open Materials through Open Practices Disclosure. The data and materials are openly accessible at https://osf.io/5vdu4/ and https://osf.io/5vdu4/ .

Disclosure statement

No potential conflict of interest was reported by the author(s).

Open practices statement

The data and materials for all experiments are available at https://osf.io/5vdu4/, and none of the experiments was preregistered.

Notes

1 Since the study material consisted of German translations of word pairs that were standardized with English materials, the association strengths between cue and target items of these translations may not perfectly match the association strengths of the original English version of the materials.

2 This approach follows the specifications of Carpenter (2011). Indeed, previous research has shown that a (reduced) pretesting effect can occur even when there is no clear relationship between the participant's guess during the pretest (which is often also likely to be a mediator item) and the target item (Cyr & Anderson, 2018).

3 To make sure that, across all subjects, mean forward association strength between the cue and mediator items was comparable for the study and pretest conditions, the following proceeding was applied: If participant X guessed the mediator of the standard list (e.g., gymnast) during pretesting, the mediator of the spare list (e.g., circus) was used on the final recognition test, as already pointed out. Critically, for the subsequent participant (X + 1), the same mediator item then was used as the mediator item for the study condition. Across subjects, all mediator items thus were presented equally often as pretest and study mediator items in the final test.

4 The cue items served as filler items on the final recognition test to obtain as many “old” items as “new” items (i.e., 48 items each). Since we were primarily interested in the hit rates of the target items, we decided to report the hit rates of the cue items separately, and not as part of an ANOVA (see also Carpenter, 2011).

5 However, one difference between the present findings and the Carpenter findings is that in the earlier study, the magnitude of the posttesting effect was most pronounced at the highest confidence level, while in the present study, level of confidence did not appear to have an effect on the size of the effect. The reason for these minor discrepancies in results may not necessarily be due to the fact that the earlier study employed a posttesting procedure and the present study used a pretesting procedure but could also be the result of other procedural differences between studies (i.e., Carpenter had subjects study fewer word pairs overall during the initial acquisition phase and informed subjects about the presence of a final test).

6 While pretest-induced elaboration processes may increase the memory strengths of both target and mediator items, ROC analysis is not able to unambiguously determine whether such processes occur with the present experimental setup.

Additional information

Funding

This work was supported by a grant from the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) awarded to Oliver Kliegl [grant number: KL 2967/2-1].