Retrieval practice reduces relative forgetting over time

ABSTRACT

There is overwhelming evidence in the literature that retrieval practice of studied material can lead to better final recall than restudy of the same material. Far less clear is whether this recall benefit is accompanied by reduced subsequent forgetting over time. This study revisited the issue in two experiments by comparing the effects of retrieval practice – with and without feedback –, restudy, and a no-practice condition on recall across different delay intervals ranging between three minutes and several days. We fitted power functions of time to the recall rates of each practice condition and compared relative forgetting rates between conditions. The comparisons showed that relative forgetting was reduced after retrieval practice relative to restudy, the relative forgetting rate after retrieval practice was unaffected by the presence of feedback, and forgetting after restudy did not differ from the no-practice condition. Together with other findings in the literature, the results provide evidence that retrieval practice reduces relative forgetting over time.

KEYWORDS:

• episodic memory
• retrieval practice
• time-dependent forgetting
• forgetting

Disclosure statement

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

Credit authorship contribution statement

A. T. Nickl: Data curation, Methodology, Visualization, Formal analysis, Writing – original draft, Investigation K.-H. T. Bäuml: Conceptualization, Methodology, Supervision, Resources, Formal analysis, Writing – review & editing, Investigation

This research is part of A. T. Nickl’s dissertation.

Declaration of interests

The authors have no conflicts of interest with respect to the authorship or the publication of this article.

Open practices statement

Materials and data are available on the Open Science Framework (https://osf.io/e8cg5/?view_only=6865e077babf42fea39ba1cf713107f8). Further requests for the data or materials can be sent via email to the corresponding author at [[email protected]].

Notes

1 A few other studies measured retention after retrieval practice without feedback across more than two delays, but did not include a restudy condition. Instead, they compared recall in the retrieval practice condition to recall in a no-practice condition, which introduces a confound with total exposure to material (Chan, 2010; Runquist, 1983; Slamecka & Katsaiti, 1988; Spitzer, 1939).

2 A two-parametric version of the power function of the form, r(t) = at^−b, has also seen wide-spread use in the forgetting literature (e.g., Rubin & Wenzel, 1996; Wixted & Ebbesen, 1991, 1997). Under most conditions, the two-parametric and three-parametric versions behave very similarly and lead to largely the same conclusions (see Wixted, 2004).

3 The parameter estimates for b were considerably higher in values in Experiment 1 than Experiment 2. This is related to the fact that the two experiments also differed considerably in best fitting scaling parameter c. Indeed, Experiment 1 showed a relatively low parameter c with relatively high parameters b, whereas Experiment 2 showed a relatively high parameter c with relatively low parameters b. When fitting the two-parametric power function model, r(t) = at^−b, in which no scaling parameter is included, to the recall rates of the two experiments, the values of parameter b were found to be much more similar between experiments. In Experiment 1, estimates of b were 0.113 for restudy, 0.084 for retrieval practice with feedback, and 0.089 for retrieval practice without feedback; in Experiment 2, estimates of b were 0.085 for restudy, 0.054 for retrieval practice without feedback, and 0.110 for the no-practice condition.

4 Like for the arithmetically averaged data, we estimated a common c parameter also for each experiment’s geometrically averaged data that we then used for all further analyses. Again, the two identified parameters differed clearly between experiments (0.48 for Experiment 1 and 28.13 for Experiment 2), but were numerically quite similar to the ones estimated for the arithmetically averaged data (0.45 for Experiment 1 and 27.78 for Experiment 2).

5 Wixted (2022) did not report any statistical analysis on whether forgetting parameters differed between stronger and weaker items. The conclusions therefore rest on the found numerical differences in forgetting rates between stronger and weaker items.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.