The testing effect is moderated by experimental design

Comparing the testing effect under blocked and mixed practice: The mnemonic benefits of retrieval practice are not affected by practice format

“…In one of these studies, Rowland, Littrell-Baez, Sensenig, and DeLosh (2014) reported no influence of list composition on the testing effect, whereas, in another study, Mulligan and Peterson (2015) found that the testing effect behaved similarly to the generation effect and was larger with mixed than with pure lists of retrieval-practice and restudy trials. In a follow-up study, Mulligan, Susser, and Smith (2016) replicated the original finding reported by Mulligan and Peterson under varying conditions and confirmed across four experiments that the testing effect can be modulated by list composition in a way similar as the generation effect-as long as specific experimental procedures are applied, ones under which the generation effect is also sensitive to list composition. For example, in accordance with theoretical accounts of listcomposition effects (e.g., the item-order account; see Nairne et al, 1991), such effects may hinge on the exact nature of the final test.…”

“…Although in one of these studies Rowland et al (2014) reported that the magnitude of the testing effect was not affected by whether pure or mixed lists of retrievalpractice and restudy trials were used, a second study by Mulligan and Peterson (2015) that was also designed in parallel to studies on the generation effect reported a significant interaction, reflecting a larger testing effect with mixed than with blocked practice. The inconsistency in results across the two reports was addressed in another recent study conducted by Mulligan et al (2016). There, four experiments confirmed that the testing effect can be affected by list composition in the same way as the generation effect, at least when experimental conditions are applied under which the generation effect is also expected to be larger for mixed than for pure lists.…”

“…There, four experiments confirmed that the testing effect can be affected by list composition in the same way as the generation effect, at least when experimental conditions are applied under which the generation effect is also expected to be larger for mixed than for pure lists. In particular, as was discussed by Mulligan et al, procedural details that are critical for whether or not effects of list composition will emerge (e.g., free recall as the final-test format and the application of separate free-recall tests after each list; see also the paragraph below) are often linked to the predictions of theoretical accounts of the examined effects (e.g., in this case, the item-specific-relational account or the itemorder account of the generation effect; for details, see Mulligan et al, 2016; for a review, see McDaniel & Bugg, 2008). In sum, the results of the recent studies on the role of list composition indicate that there may indeed be certain parallels between the testing and generation effects, suggesting that theoretical accounts that have been discussed for the generation effect may also hold relevance for the testing effect and should be scrutinized in further studies (but see Karpicke & Zaromb, 2010).…”

“…For example, in accordance with theoretical accounts of listcomposition effects (e.g., the item-order account; see Nairne et al, 1991), such effects may hinge on the exact nature of the final test. For example, they may only emerge when final free-recall tests are conducted soon after study of each single (pure or mixed) list, thus creating unique retrieval sets for each pure list that aid recall, relative to mixed lists (for details, see Mulligan et al, 2016; see also our General Discussion).…”

“…Yet, because the three previous studies on the role of list composition (Mulligan & Peterson, 2015;Mulligan et al, 2016;Rowland et al, 2014) had purposefully applied conditions that are conducive for list-composition effects for the generation effect, their results may not apply to the issue of whether the testing effect can be modulated by practice format when experimental conditions typical for testing-effect studies are used (e.g., when the final test comprises both restudied and retrieval practiced items and is conducted not only after short but also after prolonged delays, or more generally under conditions that make recall more difficult).…”