The benefits of retrieval practice on subsequent memory are firmly established (see Rowland, 2014 for a review). This benefit, often referred to as retrieval-based learning or the testing effect, has been found consistently in both laboratory and classroom contexts (e.g. McDaniel, Agarwal, Huelser, McDermott, & Roediger III, 2011; Roediger & Karpicke, 2006). However, the underlying mechanism is still unclear. With every act of retrieval, it is assumed that there is some change that improves one’s ability to retrieve that knowledge in the future. The exact nature of that change, however, is still unknown. Several theories have been proposed to explain the effect of retrieval practice on learning (e.g., Bjork, 1994; Carpenter & Delosh, 2006; Hopper & Huber, 2018; Pyc & Rawson, 2009; Rickard & Pan, 2017), but none are currently widely accepted by the field.
The episodic context account, proposed by Karpicke, Lehman, and Aue (2014), is one recent theory of why retrieval practice is beneficial. The authors suggest that the mnemonic benefit of retrieval stems from reinstating the episodic context associated with the target (or to-be-retrieved) memory. The account builds upon current context-based memory models of encoding and retrieval (e.g. Howard & Kahana, 2002; Lehman & Malmberg, 2013; Mensink & Raaijmakers, 1988). These models suggest that when people encode an event, they encode both the event itself and the context associated with the event. Later retrieval attempts are then facilitated by retrieving the context of the study episode and using it to guide memory search.
Karpicke et al. (2014) expand these context-based models to explain the benefits of retrieval on later memory. According to their account, during retrieval practice the reinstated context representation is strengthened and updated to include the context at the moment of recall. Since the updated context includes both the study context and recall context, it provides more cues to guide memory search on a future retrieval attempt (Siegel & Kahana, 2014). In contrast, when an item is restudied, the context representation is neither reinstated nor updated, thus there will be fewer and weaker retrieval cues available for a later retrieval attempt.
While the episodic context account offers a clear explanation for why and how retrieval benefits memory, the details of the account remain to be clarified. First, it is unclear what is reinstated during the memory retrieval. Context is a broad term and can refer to multiple features of an event (e.g., the modality of presentation, the speaker’s voice, the presentation font), as well as background features (e.g., the current location, the participants’ mood, background noises) (Bjork & Richardson-Klavehn, 1989; Geiselman & Bjork, 1980). Many memory models highlight temporal context as being especially important for memory retrieval. Temporal context is assumed to change slowly over time so that items that are encoded nearby in time share similar temporal contexts (Kahana, 1996; Polyn, Norman, & Kahana, 2009). Karpicke et al. (2014) emphasize the importance of temporal context, but no further details are given about what aspects of context are or are not encoded and reinstated during retrieval.
There is some evidence that temporal context is strengthened during retrieval practice. Participants who study multiple lists and are asked to retrieve the items after each list are better able to identify on which list an item occurred during a final test as compared to participants who complete some other distractor task after each list. That is, participants who practiced retrieving the items have better memory for the items’ temporal contexts (Brewer, Marsh, Meeks, Clark-Foos, & Hicks, 2010; Chan & McDermott, 2007).
However, it is still unclear if other types of context are strengthened by retrieval practice. To date, only two studies - Brewer et al. (2010) and Akan, Stanley, and Benjamin (2018) - have directly examined the effects of retrieval practice on non-temporal context memory. Both studies examined participants’ memory for incidental details from the original study phase after retrieval practice, but their results differ substantially.
First, Brewer et al. (2010) found that retrieval practice did not improve memory for context features of the items themselves. In their study, participants encoded two lists of words. Each word was presented visually and spoken by a male or female voice. After each list, participants either recalled the items or completed a distractor task. When asked to recall who presented each word on a final test, participants who had practiced retrieving the items were no better at identifying if the voice was male or female than participants in the control condition.
However, a recent study by Akan et al. (2018) did find evidence that retrieval practice improves memory for context features of the items themselves. Participants studied word pairs in one of eight locations circling the screen. During retrieval practice, participants were presented with one word of the pair in the center of the screen and were asked to recall the second word. On a final test, participants were more accurate at identifying the word pair’s original spatial location when they had previously retrieved the items as compared to when they restudied them. Thus, the current evidence surrounding the episodic context account is mixed. There is some evidence that retrieval practice does enhance memory for temporal context, but mixed results for contextual aspects of the item itself.
Across three experiments, we tested predictions from the episodic context account by examining participants’ memory for the initial study context after successful retrieval practice and comparing it to the memory of participants who restudied the same items. Specifically, we examined memory for source information (font color). During the initial study phase, cue-target pairs were presented in four different colors. Participants then either practiced retrieving the targets or restudied them and were later asked to remember the original font color they saw during the study.
If the episodic context account is true, and retrieval practice benefits memory by reinstating and updating the entirety of episodic context from the encoding phase, items that were successfully retrieved during practice should show enhanced memory for context details (i.e. font color) compared to those that were restudied. The successful retrieval practice should require participants to reinstate the prior episodic context and strengthen participants’ memory for that context. Therefore, we expect better color memory performance for items that were successfully practiced than those that were restudied.