Does reactivation trigger episodic memory change? A meta-analysis

Highlights

• Meta-analyses revealed reliable reconsolidation effects for episodic memories.

• Reactivation makes episodic memories susceptible to physiological and behavioral interference.

• Effects are more pronounced for remote memories and memories of narrative structure.

• New information presented after reactivation intrudes into the original memory.

• Findings support a dynamic view of long-term memory.

Abstract

According to the reconsolidation hypothesis, long-term memories return to a plastic state upon their reactivation, leaving them vulnerable to interference effects and requiring re-storage processes or else these memories might be permanently lost. The present study used a meta-analytic approach to critically evaluate the evidence for reactivation-induced changes in human episodic memory. Results indicated that reactivation makes episodic memories susceptible to physiological and behavioral interference. When applied shortly after reactivation, interference manipulations altered the amount of information that could be retrieved from the original learning event. This effect was more pronounced for remote memories and memories of narrative structure. Additionally, new learning following reactivation reliably increased the number of intrusions from new information into the original memory. These findings support a dynamic view of long-term memory by showing that memories can be changed long after they were acquired.

Introduction

Shortly after encoding memories are fragile and highly susceptible to both physiological and behavioral interference. Over time, memories stabilize, and once this consolidation process is complete, memories are resistant to change and stored permanently (for a review, see McGaugh, 2000). According to this view, memory failures and instances of misremembering reflect temporary problems of accessibility or source confusions, rather than loss or alterations of stored representations. Throughout the years, researchers have questioned whether consolidation is a “one-way street” converting fragile memories into permanent traces (see Dudai, 2004, Riccio et al., 2006 for reviews of the history of consolidation and reconsolidation). Alternatively, it was proposed and empirically demonstrated that memories can return to a fragile state when they are reactivated (e.g., Misanin, Miller, & Lewis, 1968), and if these memories were to survive, they needed to undergo another round of consolidation, a process which was later termed reconsolidation (Przybyslawski & Sara, 1997). This idea ultimately broke through in 2000 when Nader, Schafe and LeDoux showed that retrieval of a consolidated memory induces a transient state of plasticity which can lead to a radically altered memory. Specifically, blocking de novo protein synthesis in the basolateral amygdala after retrieval of a fear memory caused amnesic effects, that is, loss of the fear memory. This effect has been replicated numerous times, and modifications of consolidated memories have been obtained in other experimental paradigms using a variety of reactivation and post-reactivation treatments (for reviews, see Besnard et al., 2012, Nader and Einarsson, 2010, Nader and Hardt, 2009). Together, these findings support the view that reactivation transfers memories from an inactive to an active state re-opening a temporary window for memory modification. Reconsolidation processes are then needed to preserve memories and transfer them back to an inactive state (Lewis, 1979, Nader et al., 2000).

Reconsolidation has most frequently been studied in laboratory animals, particularly rodents. The invasive interventions that are used to block re-storage processes in animals, such as the injection of protein-synthesis inhibitors into targeted brain areas (e.g., Nader et al., 2000) are not suitable for human use. Therefore, the study of reconsolidation process in humans has lagged behind considerably. Human studies either employ less invasive physiological treatments, such as stress manipulations, the administration of propranolol or glucose, or utilize behavioral interference. In behavioral interference studies, new information that often bears some resemblance to the original information is presented shortly after the memory is reactivated. Reconsolidation is inferred when the delayed retrieval of the original information is impaired or “contaminated” with new information in comparison to a condition in which learning of new information was not preceded by reactivation, and in comparison to a condition in which reactivation was not followed by new learning. Using these types of post-reactivation treatments, human reconsolidation effects have been reported for amygdala-dependent, procedural and episodic memories (for reviews, see Agren, 2014, Hupbach et al., 2015, Schiller and Phelps, 2011).

The mechanisms underlying reactivation-induced memory changes remain controversial. Inherent in the reconsolidation account is the assumption that reactivation destabilizes memories, and that post-reactivation interventions either directly interfere with re-storage processes or lead to the incorporation of new elements into the original memory. Alternatively, reconsolidation effects could reflect temporary retrieval deficits that can be alleviated by contextual reinstatement (Gisquet-Verrier et al., 2015, Riccio et al., 2006, Sederberg et al., 2011). This debate resembles the old question as to whether forgetting reflects decay (i.e., loss of information) or interference (i.e., inaccessibility of information) which has been proven difficult to disentangle empirically because storage failure is almost impossible to demonstrate experimentally (Agren, 2014, Hardt et al., 2013, Miller and Matzel, 2006). Gisquet-Verrier and Riccio (2012) have advocated to consider the phenomenon of reactivation in its own right as an essential element contributing to the dynamic nature of memory, and several studies remain theoretically neutral as to whether reactivation-induced memory changes reflect storage or retrieval processes (e.g., Marin et al., 2010, St. Jacques et al., 2013).

In addition to the theoretical controversy, recent failures to replicate human reconsolidation effects have cast doubt on the ubiquity of reconsolidation effects and their reliability in humans (Hardwicke, Taqi, & Shanks, 2016). It is important to know whether and under which circumstances post-reactivation treatments are effective in altering what can be retrieved from an experience, as this knowledge has profound implications for the clinical (Schwabe, Nader, & Pruessner, 2014), educational (Bauer, 2009) and legal (Lacy & Stark, 2013) sectors.

The aim of the present study was to critically evaluate the evidence for reactivation-induced long-term changes in human episodic memory. Most studies on this topic follow the same general procedure. Some prior learning experience is either reactivated or not reactivated before some physiological or behavioral interference manipulations are applied. The interference manipulations aim to impair or enhance the reactivated memory and/or to update it with new content. The effects of these manipulations are then tested after a delay. At the core of the reconsolidation account is the idea that long-term memories will only be susceptible to memory-modifying manipulations if they are reactivated prior to applying the manipulations. Therefore, we based our effect size calculation on a comparison between reactivation and no-reactivation control groups. We used a meta-analytic approach in order to estimate the size of memory change and to evaluate several potential boundary conditions of this effect. Based on theoretical considerations, we included age of the memory, method of reactivation, type of study material, type of interference manipulation and the type of final memory test as potential moderators.

• Age of memory. Dudai and Eisenberg (2004) proposed that consolidation progresses over longer time periods than originally assumed, reinterpreting reconsolidation as lingering consolidation effects. During the prolonged consolidation process, endogenous activation and retrieval stabilize the memory network, integrate it with other memories and help establish retrieval links, but also render memories susceptible to interference. According to the lingering consolidation hypothesis, post-reactivation treatments should affect recent memories more than remote memories that have had a chance to stabilize and can be accessed through multiple retrieval routes. We analyzed whether memories that were reactivated between 24 and 48 h after acquisition were more susceptible to post-reactivation effects than memories that were reactivated between 7 and 28 days after encoding.

• Method of Reactivation. There is neural evidence that reactivation strength is related to memory change. Moderate reactivation of neural representations corresponding to specific items results in weakening of these representations and impaired recall of the corresponding items, whereas strong reactivation strengthens the representations and benefits subsequent retrieval (Detre et al., 2013, Norman et al., 2007). Similarly, retrieval practice that results in strong reactivation of learned material is one of the most effective ways to enhance memory and to promote long-term remembering (Roediger & Karpicke, 2006). Several studies further show that retrieval practice protects memories from proactive and retroactive interference (e.g., Halamish and Bjork, 2011, Potts and Shanks, 2012, Szpunar et al., 2008). Based on these findings, we assessed whether memory changes were affected by the method of reactivation. We compared indirect reactivation methods that should elicit moderate levels of reactivation, such as contextual reinstatements, with direct reactivation methods that should elicit strong reactivations, such as re-exposing participants to stimuli or asking them to recall previously presented items.

• Study material. Studies on human memory reconsolidation either use lists of items or material that has a narrative-like structure (e.g., videos, narrated slide shows, texts). Interference manipulations might have stronger effects for narrative-like materials, because reactivation of one component might trigger reactivation of the remaining elements. This might not be the case for unrelated items on a list.

• Interference manipulation. The reviewed studies used two different types of post-reactivation manipulations. Studies using physiological manipulations attempt to directly modulate re-storage processes by administering pharmacological agents that are known to affect memory consolidation or by subjecting people to stress, which causes hormonal changes that alter hippocampal and prefrontal functioning. In contrast, in behavioral interference paradigms, new information is presented after reactivation. This new information then becomes integrated into the reactivated memory or impairs it in a long-lasting manner. We included the interference manipulation as a moderator variable because these manipulations operate on fundamentally different levels: physiological manipulations target re-storage processes whereas behavioral interventions attempt to alter the content of the reactivated memory.

• Retrieval test. Studies use either recall or recognition tests as a final memory assessment. While recognition can be based on familiarity alone, recall requires conscious recollection (Jacoby, Toth, & Yonelinas, 1993). Test-type dependent differences in reconsolidation effects could shed light onto the storage versus retrieval debate.

The present meta-analysis focused exclusively on studies assessing reactivation-induced changes in episodic memory. Procedural or amygdala-dependent memories are supported by different brain systems and therefore were not considered in our analyses. We decided to compare reactivation to no-reactivation control conditions in our effect size calculations, because according to the reconsolidation account, only reactivated long-term memories should be susceptible to memory interfering treatments.