Does testing enhance mediation in paired-associate learning?

ABSTRACT

The testing effect (TE) is the finding that testing on previously studied material leads to better long-term retention as compared to restudying that material. Pyc and Rawson (2010) proposed the Mediator Effectiveness Hypothesis (MEH) as an explanation for the TE in paired-associate learning. The MEH states that review testing on cue-target word pairs strengthens semantic/associative mediators, which help participants recall targets to their cues on a later test. Pyc and Rawson found support for the MEH with Swahili-English word pairs and explicit mediation instructions, using the most rigorous means of assessing the MEH (i.e. asking participants to recall both the mediator and target to each cue). Using these procedures with spontaneous mediation conditions and unrelated English word pairs, the results of Experiments 1 and 2 failed to support the MEH. When semantic/associative or phonological mediation was explicitly encouraged in Experiment 3, the MEH was supported only for phonological mediation.

KEYWORDS:

• Testing effect
• paired-associate learning
• mediator effectiveness hypothesis
• mediator shift hypothesis

Data Availability

The data that support the findings of this study are openly available in OSF at https://osf.io/pxb4e/?view_only=126e38b4b2e34063a9d79fcdfbb64f63, DOI number 10.17605/OSF.IO/PXB4E.

Author note

This manuscript is based on my dissertation, which was submitted to the University at Albany in partial fulfillment of the requirements for a doctoral degree in cognitive psychology. I thank Jim Neely (chair), Jeanette Altarriba and Heather Sheridan for their helpful comments as members of my dissertation committee.

Disclosure statement

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

Notes

1 Eleven participants (five in the test group and six in the restudy group) were excluded from the analysis because they recalled all of the mediators during Session 2.

2 One participant in the restudy group was excluded from the analysis for recalling all of the mediators during Session 2.

3 The 2 (Type of Review: Test vs. Restudy) × 2 (Type of Final Test Recall: Mediator vs. Target) mixed-factor ANOVA was repeated to assess if participants were recalling mediators that were not their most recently generated mediators. Mediator recall remained similar when the participants’ Review Cycle 1 mediators were considered as correct mediator recall (73 ± 5%) and when Review Cycle 2 mediators were considered as correct mediator recall (73 ± 5%).

4 Two participants in the test group were excluded from the analysis because they recalled all of the mediators during Session 2.

5 The same pattern of results emerged when considering mediators recalled from Review Cycles 1 and 2, instead of the most recently generated mediator. There was a nonsignificant 3 ± 17% difference in TEs for target recall when the mediator from Review Cycle 1 could be recalled in Session 2 as compared to when it could not, and a nonsignificant 0 ± 19% difference in TEs for target recall when the mediator from Review Cycle 2 could be recalled in Session 2 as compared to when it could not.

6 The 2 (Type of Review: Test vs. Restudy) × 2 (Type of Final Test Recall: Mediator vs. Target) mixed-factor ANOVA was repeated to assess if participants were recalling mediators that were not their most recently generated mediators. Mediator recall remained similar when the participants’ Review Cycle 1 mediators were considered as correct mediator recall (59 ± 9%) and when Review Cycle 2 mediators were considered as correct mediator recall (60 ± 8%).

7 Three participants (one from the test group and two from the restudy group) were excluded from the analysis because they recalled all of the mediators during Session 2.

8 The same pattern of results emerged when considering mediators recalled from Review Cycles 1 and 2, instead of the most recently reported mediator. There was a significant 22 ± 21% difference in TEs for target recall when the mediator from Review Cycle 1 could be recalled in Session 2 as compared to when it could not, and a significant 23 ± 21% difference in TEs for target recall when the mediator from Review Cycle 2 could be recalled in Session 2 as compared to when it could not.