Retrieval practice (self-testing) reliably improves long-term retention of studied material, but whether it improves the ability to draw deductive inferences — conclusions that follow necessarily from combinations of studied premises — is contested: published experiments span positive, null, and negative effects. Eglington and Kang (2018) showed that practice testing improves later deductive inference from studied sentence sets when the related sentences are presented simultaneously during study and practice, but not when they are presented one at a time, and they left open four linked questions: (1) whether retrieval success during practice causally moderates the inference benefit; (2) the boundary conditions across materials, inference types, and degrees of element interactivity (including interactivity at study versus at test); (3) which mechanism — episodic-context updating, reconsolidation, enhanced organizational processing, or transfer-appropriate processing — produces the benefit; and (4) the roles of learners' test expectations and feedback timing in enabling relational processing during practice. The problem is to determine, jointly, under what conditions practice testing improves, leaves unchanged, or impairs deductive inference from studied material, and through what mechanisms. Later literature sharpens rather than resolves the question: meta-analysis shows transfer overall is positive but moderated by initial test performance; transitive-inference experiments show retrieval practice can even impair inference; and no study has causally manipulated retrieval success for deductive inference or discriminated the candidate mechanisms.

Background

Retrieval practice (also called the testing effect) refers to the finding that actively retrieving information from memory during learning — for example, taking fill-in-the-blank or free-recall practice tests — produces better long-term retention of that information than restudying it for an equivalent time. It is among the most robust effects in learning science, but most demonstrations concern memory for the practiced material itself, and its benefits for using the material in new ways are less well established.

A central case of such use is deductive inference: answering questions whose correct answers are never stated verbatim in the studied material but follow necessarily when several studied premises are combined (e.g., after learning a set of related facts about who lives where and who owns what, answering which pet belongs to a person never directly paired with one in the text). Tran, Rohrer, and Pashler (2015) reported a striking failure of transfer: after participants studied sets of related premises, fill-in-the-blank practice improved later recall of the premises but produced no advantage over rereading on multiple-choice questions requiring deductive inferences from them.

A candidate explanation comes from cognitive load theory. Element interactivity denotes the degree to which the elements of to-be-learned material must be processed jointly rather than independently, because they interact with one another; material with high element interactivity is intrinsically harder to learn. Leahy, Hanham, and Sweller (2015) proposed that the testing effect diminishes or disappears as element interactivity increases. Deductive inference is a high-interactivity demand, because it requires integrating information across multiple premises.

Eglington and Kang (2018) revisited this question in three experiments. Participants studied four sets of seven to nine related sentences, practicing retrieval for two sets and rereading the other two, with a final test two days later that included deductive-inference questions. When sentences within a set were presented one at a time during study and practice (as in Tran et al., 2015), retrieval practice gave no benefit on the inference test. When all sentences of a set were presented simultaneously, so that relational processing — integrating the individual sentences into a coherent structure — was possible during practice, retrieval practice (fill-in-the-blank testing in Experiments 1–2, free recall in Experiment 3) improved later deductive inference relative to rereading. The authors concluded that retrieval practice can improve deductive inference, provided the practice format does not hinder relational processing of the material.

They explicitly left open four issues. First, the causal role of retrieval success: Experiment 3 (free recall) showed lower intermediate practice performance, which may have constrained the effect, so experiments that manipulate or systematically vary retrieval success during practice are needed to establish whether the level of learning reached during practice moderates the later inference benefit. Second, boundary conditions: it remains unknown how far the simultaneous-presentation rescue generalizes to other high-interactivity educational materials, and whether element interactivity at study versus at test has different roles. Third, the mechanism producing the transfer is undetermined among at least four candidates: reconsolidation (the hypothesis that retrieval renders memory traces labile and rewritable), episodic-context updating (the account of Karpicke, Lehman, and Aue, 2014, that retrieval updates the contextual cues associated with remembered items), enhanced organizational processing (retrieval encourages organizing material into coherent structures), and transfer-appropriate processing (practice-test processing overlaps with final-test processing). Fourth, the roles of learners' intentions — for example, expecting an inference test — and of the timing of feedback after practice attempts in enabling relational processing during retrieval practice are unknown.

Subsequent literature has made the picture more, not less, contingent. A meta-analysis of transfer of test-enhanced learning (Pan and Rickard, 2018) found overall positive transfer of testing (d = 0.40 across 192 effect sizes), strongest for application and inference questions, with initial test performance a strong moderator — but publication-bias adjustments reduce the predicted effect to near zero when supportive moderators are absent. Wissman, Zamary, and Rawson (2018) replicated the null result of Tran et al. with fill-in-the-blank testing and proposed that deductive inference depends on complete memory for the premises, so testing helps inference only when it yields a memory advantage. Most strikingly, Picklesimer, Buchin, and Mulligan (2019) and Mulligan, Buchin, and Powers (2023) found that retrieval practice impairs transitive inference (inferring, from premises such as B > C and C > D, relations like B > D that were never directly studied), with a small-scale meta-analytic effect of d = -0.37, and this impairment persisted under simultaneous presentation and with pictorial materials. Related boundary-condition work shows no benefit for mathematical word-problem solving (Huang, Zheng, Yu, and Chen, 2023) but positive effects on inference questions in second-language discourse comprehension when free recall is combined with self-explanation (Liu, Zheng, and Song, 2024), and retrieval-practice benefits for far transfer via rule-based (inductive) learning of an artificial language (Opitz and Kubik, 2024). Retrieval success during practice has been causally manipulated for retention outcomes (Racsmány, Szőllősi, and Marián, 2020, showed feedback can reverse the testing effect at low practice success but not at moderate or high success), but not yet for deductive inference. The upshot is an active empirical controversy with directly contradictory published effects, whose resolution requires exactly the systematic work the source paper called for.

Problem Statement

Determine when and why retrieval practice, relative to restudy, improves the ability to draw deductive inferences from studied material, through four linked parts of one causal account of when and why testing changes deductive-inference ability: (1) the causal role of retrieval success during practice — whether and how the level of learning or retrieval success reached during practice quantitatively moderates the later deductive-inference benefit; (2) the boundary conditions — over presentation format (simultaneous versus sequential presentation of related material during study and practice), material type and degree of element interactivity (including whether interactivity at study and interactivity at test have different roles), and type of inference (sentence-premise deduction versus transitive inference), specifying the conditions under which the effect is positive, absent, or reversed; (3) the mechanism — which of the candidate accounts (episodic-context updating, reconsolidation, enhanced organizational processing, transfer-appropriate processing), or what alternative, produces the observed transfer; and (4) the roles of learners' expectations (e.g., expecting an inference test) and of the timing of feedback after practice attempts in enabling relational processing during retrieval practice.

References

  1. Eglington, L. G., & Kang, S. H. K. (2018). Retrieval Practice Benefits Deductive Inference. Educational Psychology Review, 30(1), 215-228. DOI: 10.1007/s10648-016-9386-y. https://doi.org/10.1007/s10648-016-9386-y
  2. Tran, R., Rohrer, D., & Pashler, H. (2015). Retrieval practice: the lack of transfer to deductive inferences. Psychonomic Bulletin & Review, 22(1), 135-140. DOI: 10.3758/s13423-014-0646-x. https://doi.org/10.3758/s13423-014-0646-x
  3. Wissman, K. T., Zamary, A., & Rawson, K. A. (2018). When does practice testing promote transfer on deductive reasoning tasks? Journal of Applied Research in Memory and Cognition, 7(3), 398-411. DOI: 10.1016/j.jarmac.2018.03.002. https://doi.org/10.1016/j.jarmac.2018.03.002
  4. Pan, S. C., & Rickard, T. C. (2018). Transfer of test-enhanced learning: Meta-analytic review and synthesis. Psychological Bulletin, 144(7), 710-756. DOI: 10.1037/bul0000151. https://doi.org/10.1037/bul0000151
  5. Picklesimer, M. E., Buchin, Z. L., & Mulligan, N. W. (2019). The Effect of Retrieval Practice on Transitive Inference. Experimental Psychology, 66(6), 377-392. DOI: 10.1027/1618-3169/a000467. https://doi.org/10.1027/1618-3169/a000467
  6. Mulligan, N. W., Buchin, Z. L., & Powers, A. (2023). Transitive inference and the testing effect: Retrieval practice impairs transitive inference. Quarterly Journal of Experimental Psychology, 76(10), 2356-2370. DOI: 10.1177/17470218231156732. https://doi.org/10.1177/17470218231156732
  7. Racsmány, M., Szőllősi, Á., & Marián, M. (2020). Reversing the testing effect by feedback is a matter of performance criterion at practice. Memory & Cognition, 48(7), 1161-1170. DOI: 10.3758/s13421-020-01041-5. https://doi.org/10.3758/s13421-020-01041-5
  8. Huang, X., Zheng, S., Yu, Z., & Chen, S. (2023). Retrieval practice may not benefit mathematical word-problem solving. Frontiers in Psychology, 14, 1093653. DOI: 10.3389/fpsyg.2023.1093653. https://doi.org/10.3389/fpsyg.2023.1093653
  9. Liu, D., Zheng, T., & Song, Y. (2024). Testing effect in L2 discourse comprehension: importance of retrieval-based learning. Current Psychology, 43(37), 29346-29363. DOI: 10.1007/s12144-024-06481-9. https://doi.org/10.1007/s12144-024-06481-9
  10. Opitz, B., & Kubik, V. (2024). Far transfer of retrieval-practice benefits: rule-based learning as the underlying mechanism. Cognitive Research: Principles and Implications, 9, 65. DOI: 10.1186/s41235-024-00598-y. https://doi.org/10.1186/s41235-024-00598-y
  11. Wang, L., & Yang, J. (2024). The influence of repeated study and repeated testing on the testing effect and the transfer effect over time. Memory & Cognition, 52(3), 476-490. DOI: 10.3758/s13421-023-01477-5. https://doi.org/10.3758/s13421-023-01477-5
  12. Wang, L., & Yang, J. (2025). Interaction of repetition and retention interval influences the transfer effect after answer feedback for episodic memory. Frontiers in Psychology, 16, 1638780. DOI: 10.3389/fpsyg.2025.1638780. https://doi.org/10.3389/fpsyg.2025.1638780
  13. Leahy, W., Hanham, J., & Sweller, J. (2015). High Element Interactivity Information During Problem Solving may Lead to Failure to Obtain the Testing Effect. Educational Psychology Review, 27(2), 291-304. DOI: 10.1007/s10648-015-9296-4. https://doi.org/10.1007/s10648-015-9296-4
  14. Karpicke, J. D., Lehman, M., & Aue, W. R. (2014). Retrieval-Based Learning: An Episodic Context Account. Psychology of Learning and Motivation, 61, 237-284. DOI: 10.1016/B978-0-12-800283-4.00007-1. https://doi.org/10.1016/B978-0-12-800283-4.00007-1