No evidence that arousal affects reactivated memories

,


Introduction
The capacity to remember previous experiences has been shaped by evolution to prioritize information that is relevant to survival (Klein et al., 2002;Nairne & Pandeirada, 2008).This is why elements of emotionally charged events, such as a car's head lights in near-fatal accidents or a gun in violent encounters, leave particularly vivid and long-lasting memories (Hirst et al., 2010;Talmi, 2013).As the capacity to store memories is presumed to be limited, it needs to be decided for each element of every experience whether it may prove useful in the future and should therefore be retained, or whether that would be a waste of precious energy (J.R. Anderson & Milson, 1989).A wealth of studies has demonstrated that noradrenergic arousal drives the enhanced consolidation of emotional events (Krenz et al., 2021;McGaugh, 2018), underscoring a key role for this neurotransmitter in biasing what information to retain.However, the future value of a memory is rarely immediately obvious, meaning its fitness-relevance can only be revealed over time.This suggests that an adaptive memory system would do well to retroactively strengthen elements of past experiences whenever these become more relevant.From this perspective, retrieval of a memory can be considered a beginning point for a variety of changes depending on the current internal and external state, rather than an endpoint of encoding and consolidation (Kensinger & Ford (2020).Previous studies have shown that the mere act of retrieving a memory may strengthen its future expressions (Karpicke & Roediger, 2008).Yet it remains unknown whether retrieval of a memory interacts with noradrenergic arousal to alter its course towards long-term persistence beyond the effect of retrieval alone.The present study was designed to address this outstanding question.
Some first clues on how noradrenergic arousal may affect a reactivated memory can be derived from studies that demonstrate its effects on inherently neutral stimuli that co-occur with an emotional event.That is, the emotional enhancement of memory is not limited to the central emotional feature of an event, but also extends to unrelated neutral elements that just precede it.In an early demonstration of this effect, participants viewed neutral images which were sometimes followed by emotionally arousing scenes (Anderson et al., 2006).One week later, recognition of neutral images that were presented near the onset of an emotionally arousing scene was enhanced.Similarly, it was shown by Schwarze and colleagues (2012) that memory for neutral images can be enhanced when they are immediately followed by a mild shock.As these studies investigated memory for neutral stimuli without a predictive or semantic relationship to the source or arousal, they rule out that the emotional enhancement occurs through cognitive factors which often confound studies comparing memory for neutral and emotional stimuli, such as increased attention or saliency (Sommer et al., 2008).This leaves the involvement of noradrenergic arousal as a likely explanation for this effect.Indeed, it has been proposed that noradrenaline released in response to an arousing stimulus increases amygdala activity, which in turn stimulates the hippocampus, thereby initiating enhanced memory consolidation for attended items (Inman et al., 2018;McGaugh, 2004).
Interestingly, it may be that noradrenergic arousal not only amplifies newly encoded items, but also enhances reactivated memories in much the same way.This would be consistent with the hypothesis that reconstruction of a memory along with an emotional state can strengthen its future expression (Rubin et al., 2008).Mechanistically, several models of episodic memory have proposed that the hippocampus functions as a central hub in a recurrent processing loop (Hintzman, 1984;Kumaran & McClelland, 2012;Nadel & Moscovitch, 1997), meaning that the memories it retrieves are also fed back in as inputs to be encoded just as if it were presently perceived (Koster et al., 2018).This idea is more generally rooted in multiple-trace theories of episodic memory (Hintzman, 1984;Nadel et al., 2000), which state that reexperiencing a stimulus either through repeated perception or reactivation in memory results in the encoding of an extra copy of that stimulus.Together, the mechanisms of noradrenergic arousal and recurrent processing may enhance an existing memory by boosting the consolidation of a second copy.Additionally, it may be possible to retroactively strengthen a reactivated memory by directly strengthening its original representation.Animal studies have shown that the retrieval of conditioned associations may render them susceptible to change through various manipulations, a process referred to as 'reconsolidation' (Elsey et al., 2018;Nader et al., 2000).Given that reconsolidation and consolidation are thought to depend largely on the same molecular signaling pathways (Alberini, 2005), it is possible that noradrenergic arousal during retrieval enhances reconsolidation just as it does for consolidation.Thus, there are plausible mechanisms by which heightened arousal could strengthen subsequent expressions of a reactivated memory, either by enhancing the original representation or a new one.Note however that at the behavioral level these hypotheses lead to the same predictions, and to disentangle them is beyond the scope of this paper.
No studies to date have specifically isolated arousal to investigate its effects on specifically reactivated memories, but some earlier work has addressed the question how more general negative emotional states affect reactivated memories.The pattern of results that has emerged from these studies is somewhat conflicting.Some have showed that a stress-inducing manipulation enhances just reactivated memories (Bos et al., 2014;Coccoz et al., 2011Coccoz et al., , 2013)), while others found an impairment (Hupbach & Dorskind, 2014;Schwabe & Wolf, 2010).However, as memories were retrieved in bulk to be affected by a single stressor, these effects are likely additionally moderated by heightened cortisol rather than solely due to fast-acting noradrenergic influences (Roozendaal et al., 2006).Also, since the reactivation procedures typically included attempts to induce some form of novel learning (i.e., to induce a prediction error), it is challenging to use these observations as basis to substantiate a clear hypothesis regarding a simple reactivation.The first study to investigate retroactive changes in memory strength with a design that targets and manipulates specific events reported an impairing effect of emotional faces on the subsequent free recall of preceding words (Strange et al., 2010).In contrast, later experiments found a retroactive emotion-induced enhancement of memory (de Vries et al., 2022;Zhu et al., 2022).Together, these studies broadly suggest that emotional factors can alter the strength of previously consolidated memories, but reveal seemingly contradictory effects.On closer inspection, these results may be difficult to directly compare given important differences in study designs.In short, they leave three important questions unanswered.First, since the majority of studies either induced some form of novel learning upon reactivation it is likely that these studies have targeted other memory processes rather than modelling the mere process of reactivation.Second, since none reported an analysis of trial-level modulation by some index of noradrenergic arousal such as amygdala activity or skin conductance, it remains unclear how arousal specifically contributes to this pattern of findings.It is possible that arousal had enhancing effects on reactivated memories across study designs, but that in some this was outweighed by other factors that have an impairing effect, such as heightened cortisol.Finally, no study to date has measured the degree to which specific memories were reactivated prior to inducing an emotional state.Studies of emotionally neutral episodic memory indicate that reactivation strength is an important determinant of whether it will be modified (Ritvo et al., 2019;Sinclair & Barense, 2019).Whether this also holds for emotional manipulations is currently unknown.
Both multiple trace and reconsolidation accounts of memory dynamics agree that newly encoded events and reactivated memories exist in a similar state.Yet, as it stands, even though there are studies showing event-specific noradrenergic enhancements of newly encoded items (A.K. Anderson et al., 2006;Inman et al., 2018;Schwarze et al., 2012), there are no studies adopting a comparable strategy to study the effects of arousal on reactivated memories.This hinders important theoretical discussions on whether memories are encoded and subsequently altered by the same or distinct mechanisms (Alberini, 2005;McKenzie & Eichenbaum, 2011), as well as our understanding of how emotional states during memory reactivation may have maladaptive consequences.To investigate whether stimulating noradrenergic arousal during memory reactivation strengthens their future retrieval, we designed an experiment consisting of three sessions.During the first session, participants encoded images of objects that were presented in unique context images.To promote the 'what', 'when', 'where' character of the encoded events that typifies episodic memories (Tulving, 1983), they were instructed to use each object-context pair to imagine a vivid story involving themselves (de Vries et al., 2022;Van Ast et al., 2013).The next day, two thirds of the context images were presented again, thereby presumably reactivating the memories of their corresponding objects, and followed by a shock in 50 % of trials.For each reactivated trial, the degree of reactivation was indexed using a subjective indication of reliving on a visual analogue scale (VAS), and noradrenergic arousal was indexed by measuring the skin conductance response (SCR) elicited by the shock.Finally, in the third session, participants completed a recognition test of objects encoded on day one.As the enhancing effect of noradrenergic arousal on newly encoded items has been shown to be time-dependent (Inman et al., 2018;Schwarze et al., 2012), we investigated whether this is also true for reactivated memories by having one group of participants complete the recognition test immediately after the manipulation, whereas another group was tested the next day.Specifically for the delayed testing group it was predicted that recognition would be improved for all reactivated objects (Chan & McDermott, 2007;Karpicke & Roediger, 2008), but further enhanced for objects whose reactivation was followed by the arousal inducing electrical stimulus.Additionally, we predicted parametric modulation of this effect by both trial-specific noradrenergic activity as indexed by skin conductance responses (SCR), and memory reactivation strength as indexed by subjective ratings of reliving.

Participants
Previous studies have demonstrated noradrenergic memory enhancements of initial encoding with sample sizes of around 20 participants (e.g.Schwarze et al., 2012), but did not consistently report effect sizes on which power analysis can be based.We sought to include a substantially larger sample for two reasons.First, it has been demonstrated that effect sizes of published studies are, on average, twice as O.T. de Vries et al. large as direct replications (Camerer et al., 2018).Second, instead of direct perceptions we are aiming to manipulate retrieved memories, which requires the memory to be both retained and accessible during the second session.This is very unlikely to be the case for all trials, thus somewhat reducing the sensitivity of our analyses relative to consolidation studies.For these reasons, we wanted to include at least three times the typical number of participants in consolidation studies in the delayed testing group where the effect of arousal on reactivated memories was hypothesized.The question of whether this effect is timeindependent was secondary, which is why most recruited participants were allocated to the delayed-testing group.Participants signed up via the university's recruitment website.Exclusion criteria as assessed via a screening based on self-report were recreational drug use at a frequency higher than once a month, average consumption of 14 or more units of alcohol per week, having experienced trauma, and having received treatment for a mental disorder listed in the DSM-5 by a psychologist or psychiatrist in the past year.A total of 97 participants signed informed consent and started the experiment.It was emphasized that participants were free to quit the experiment at any time without having to state a reason, but no participants made use of this option.One participant had to be excluded due to a technical failure, and another missed the appointment for the third day.Finally, one participant was excluded due to chance level performance on the recognition test.The final sample thus includes data of 94 participants, of which 69 were in the delayed testing group (mean age = 22.0, SD = 3.1; 70 women).For the key comparison of memory for objects reactivated with and without shock, a post-hoc analysis of achieved power (conducted in G*Power 3.1, Faul et al., 2007) showed that this sample size provided a statistical power of 0.8 to detect an effect size between small and medium (Cohen's d = 0.35), and of approximately 1 for effects of medium size and above (Cohen's d >= 0.5).The immediate testing group consisted of 25 participants (mean age = 22.7, SD = 4.0, 19 women).Skin conductance data of five participants were lost, meaning they were not included in analyses of SCRs.Participants were rewarded either with course credits or 30 euros for completing the experiment, or according to the total amount of time spent in the lab in case of dropping out.This study was performed in accordance with the Declaration of Helsinki and approved by the local ethics committee of the University of Amsterdam.

Stimuli
Stimulus presentation and recording of behavioral data was done using Presentation by Neurobehavioral Systems.

Images.
The 240 objects that functioned as targets and lures were selected from the Bank of Standardized Stimuli (BOSS; Brodeur et al., 2010).The 120 context images depicted equally many indoor as outdoor settings.These were taken from previous studies from our lab (e.g.Van Ast et al., 2013) and supplemented with images from the internet.

Electrical stimulation.
Shocks were administered to the wrist of the non-dominant hand through two 20 x 25 mm Ag/AgCl electrodes that were covered with a small layer of electrode gel, and controlled by a Digitimer DS71 (Welwyn Garden City, UK).

Skin conductance
Skin conductance data was measured using two 16 x 20 mm Ag/AgCl sensors that were attached to the medial phalanx surfaces of the index and ring finger, and registered with VSRRP98, University of Amsterdam's inhouse software for recording physiological data.The data was collected at a sampling rate of 1000 Hz, and down sampled to 100 Hz for analysis.

Design and procedure
The experiment consisted of three sessions: 1) encoding, 2) reactivation & manipulation, and 3) memory testing.Participants in the delayed testing group performed the sessions on three consecutive days, whereas in the immediate testing group session two and three were performed on the same day with a 10-minute break in between during which they watched a neutral underwater film fragment.For consistency between groups, participants in the delayed testing group also watched this video at the end of session two.Six different versions of the experiment were used such that each of the 240 object images would equally often be used as a target in one of the three reactivation conditions or as a lure.Before and after each computer task, participants filled in the state-section of the State-Trait Anxiety Inventory (STAI; Spielberger et al., 1983) and the Positive and Negative Affect Schedule (PANAS; Watson et al., 1988).

Encoding
At the start of the first session participants signed an informed consent form and filled in two trait questionnaires: the trait section of the STAI (Spielberger et al., 1983) and Beck's Depression Inventory (BDI; Beck, A. T., Steer, R. A., & Brown, 1996) Fig. 1.Then, a calibration procedure was done to set the shock intensity in advance of the next session.This was done gradually by presenting shocks beginning at 1 mA and gradually increasing the intensity until the participant indicated that it reached a level that was uncomfortable but not painful up to a maximum of 55 mA.During the encoding task of the experiment context images were presented for 7 s, in which after 3 s a smaller image of an object would appear and disappear again after 2 s.Participants were instructed to imagine a story involving the object in its presented context.The session consisted of 120 trials.Each trial was immediately followed by a visual analogue scale (VAS) that was used to rate the vividness of the imagined story from 0 (not vivid at all) to (extremely vivid).Participants were given 3 s to indicate their vividness ratings, after which there was an inter-trial interval (ITI) of 8, 9, 10, or 12 s (randomly chosen).

Reactivation and manipulation
The second session took place approximately 24 h after the first session.Out of 120 previously encoded episodes, 80 were reactivated by again presenting the contexts, but this time without the corresponding objects.Each reactivated context was shown for 7 s, during which participants were instructed to indicate the degree to which they relived their encoding experience of the previous day on a VAS from 0 (no reliving) to 100 (clear reliving).Critically, half of the context images were immediately followed by a brief shock to the left wrist.The experiment was programmed such that no more than 3 contexts could consecutively be presented with or without a shock.There again was a randomly chosen ITI of 8, 9, 10, 11, or 12 s.

Memory testing
The final session consisted of a recognition test of objects presented during session one, and either took place immediately following the second session (immediate testing group) or one day later (delayed testing group).Targets were mixed in with unseen lures and presented in a pseudo-random order such that no more than 3 of each type would be presented sequentially.For each image, participants had to indicate whether this was an 'old' or 'new' object, and how certain they were about each judgement (guess, moderately sure, very sure).The recognition trials were self-paced, and followed by randomly chosen ITIs of 3, 4, or 5 s.When the test was completed, participants filled in an exit questionnaire with questions regarding their experience of the experiment and were then debriefed on the real purpose of the experiment.
O.T. de Vries et al.

Skin conductance data
Skin conductance responses were computed by finding the peak value in the 6 s that followed the offset of each context presentation on day two, and subtracting a baseline which was by computed taking the mean value of the 1 s preceding this window.

Statistical tests
All analyses were conducted in R version 4.0.3(R Core Team, 2021).To test the main hypothesis that reactivation in the presence of heightened arousal retroactively enhances memory, we calculated corrected hit rates by subtracting the proportion of false alarms (lures classified as old) from the proportion of hits (targets classified as old).The corrected hit-rates were used as the dependent variable of a mixed ANOVA with Group (immediate testing, delayed testing) as betweensubjects factor and Reactivation Condition (no reactivation, reactivation, shock) as within-subjects factor.We predicted a Group x Reactivation interaction.Significant effects were further investigated using Tukey-adjusted planned comparisons of marginal means with the package 'emmeans' (Russell et al., 2023).In the delayed testing group, we predicted that reactivated memories would be stronger than nonreactivated memories, and those reactivated with shock on their turn to be stronger than those reactivated without shock.For the immediate group, we predicted that reactivation would have the same enhancing effect, but without an added effect of shock.The hypothesis that noradrenergic arousal modulates an enhancement of memory that depends on the magnitude of reactivation was further tested by running a multilevel probit regression using the package 'lme4′ (Bates et al., 2015) with Reactivation Condition, reliving rating and SCR as trial-level predictors of correct 'old' responses.This approach has two important benefits as compared to traditional subsequent memory analysis (Wright et al., 2009): First, it computes a beta coefficient and corresponding standard error that refers to the effect of a predictor at the level of specific memories, as opposed to an individually aggregated average across within-subject factors.Second, it uniquely enables the inclusion of interactions between multiple continuous predictors.The continuous variables SCR and reliving were Z-scored within participants to account for individual variation in physiological responsivity and in the range of the VAS that was used.If heightened arousal specifically leads to enhanced memory for successfully reactivated shock trials, this predicts a significant three-way interaction between Reactivation Condition, SCR, and reliving.

Vividness and reliving ratings did not differ across groups and conditions
A mixed ANOVA of mean vividness ratings given on day one of the experiment with Group (immediate test, delayed test) as betweensubjects factor and Reactivation Condition (no reactivation, reactivation, reactivation with shock) as within-subjects factor revealed no significant effects (all F < 0.774, all p > 0.463).Similarly, a mixed ANOVA of mean reliving ratings given on day two with the same factors, but without the level 'no reactivation' as those trials were not presented during session two, showed no significant effects (all F < 1.612, all p > 0.207).Together, these results indicate that both vividness (mean = 54.57,average within-participant sd = 26.00)and reliving (mean = 53.66,average within-participant sd = 29.58)ratings did not differ across testing groups and reactivation conditions.

Shocks successfully triggered heightened skin conductance
A mixed ANOVA of mean SCRs with Group (immediate test, delayed test) as between-subjects factor and reactivation Condition (no reactivation, reactivation, reactivation with shock) as within-subjects factor revealed a main effect of Condition (η p 2 = 0.36, F 1,87 = 49.537,p < 0.001), but no significant effect of Group (η p 2 = 0.00, F 1,87 = 0.012, p = 0.915) nor an interaction between Group and Condition (η p 2 = 0.00, F 1,87 = 0.038, p = 0.846).This suggests that the arousal manipulation was successful and equally effective across testing groups.There was gradual habituation to the shock, as evidenced by a multilevel regression that showed a significant interaction between shock condition and trial number (β = -0.005µS, CI 95 = [-0.009,− 0.002], p = 0.003), indicating that the difference between conditions decreased across trials.However, SCRs to later shock trials still remained higher on average (Fig. 2).

No evidence that arousal enhances reactivated memories
A mixed ANOVA with Group (immediate testing, delayed testing) as a between-subjects factor and Condition (no reactivation, reactivation, reactivation with shock) as within-subjects factor revealed a significant interaction (η p 2 = 0.07, F 2,184 = 6.64, p = 0.002), but no main effects of Group (η p 2 = 0.01, F 1,92 = 1.03, p = 0.313) or Condition (η p 2 = 0.01, F 2,184 = 1.17, p = 0.312).Tukey-adjusted comparisons of marginal means in the delayed testing group revealed that reactivated memories in both the shock and no shock conditions were better remembered than nonreactivated memories (reactivation vs.  3).
To explain the unexpected impairing effect of shock on reactivated memories in the immediate testing group, we considered the possibility that the arousal induced by shock trials impaired retrieval of the following trial.Due to the constraint of presenting a maximum of three contexts in the same reactivation condition sequentially, this would occur disproportionally often in trials reactivated without shock.This was tested by classifying each trial both by whether it was Preceded (shock, no shock) and Followed (shock, no shock) by a shock, and conducting a 2x2 rmANOVA on the corresponding hit-rates.There were no significant effects (all F < 3.38, all p > 0.070), thus providing no evidence for the hypothesis that impairment of reactivated trials was carried by shocks delivered after preceding trials.We additionally explored whether reactivated memories were affected by preceding shock trials in the delayed testing group, reasoning that shocks may have recruited additional attention to the task whenever they occured.However, this analysis also did not yield any significant effects (all F < 0.187, all p > 0.666).

Moderation by SCR and reliving
To test the complementary hypothesis that arousal specifically moderates the strength of reactivated memories in a time-dependent manner, we conducted a multilevel regression on the data of the delayed testing group with Reactivation Condition (reactivation, shock), subjective reliving, and SCR as interacting predictors.Even though the previous analyses did not reveal any overall effects of condition, it remains possible that within-participant variance in arousal responses during reactivation is predictive of subsequent memory strength.This revealed a main effect of reliving (β = 0.17, CI 95 = [0.10,0.23], p < 0.001), but not of SCR (β = 0.02, CI 95 = [-0.06,0.09], p = 0.685), or condition (β = 0.01, CI 95 = [-0.08,0.10], p = 0.813).There were no significant interactions (all p > 0.103).These results indicate that stronger reliving is associated with an increased likelihood of subsequent object recognition.This effect is however not moderated by the magnitude of elicited arousal as indexed by SCR, and independent of the Fig. 2. Skin conductance responses for trials reactivated with and without shock throughout session two.SCRs to shock trials were significantly higher and habituated slightly over time.
presence of arousal-inducing shock.
The use of Z-scores computed at the individual level may have obscured an effect of SCR on reactivated memories by standardizing values of participants that responded both weakly and strongly to the shock to the same scale.To explore a potential role for individual differences in absolute SCRs in the enhancement of subsequent memory, we ran a regression with individual differences in mean responses to the shock relative to trials reactivated without shock as a predictor of a difference score computed by subtracting the hit-rate for the reactivation condition with shock from that for reactivation without shock.This yielded no significant effect of SCR (β = -0.002,CI 95 = [-0.030,0.026], p = 0.892).This means that individual differences in skin conductance responses to shock do not relate to any potential enhancement of reactivation + shock versus reactivated memories without shock.Finally, we employed a similar analysis strategy to test whether the presence of shocks during the second session has affected all reactivated trials, rather than just those that were immediately followed by shock.Participants' average responses to shock trials, relative to reactivated trials, were used to predict difference scores calculated by subtracting hit-rates for not reactivated trials from that of reactivated trials without shock.If the latter were systematically enhanced by heightened arousal from shocks presented after shock trials, participants who responded more strongly to the shock should show a larger effect.This regression also revealed no significant effects of SCR (β = -0.003,CI 95 = [-0.035,0.028], p = 0.832).We thus find no evidence for the hypothesis that arousal induced by shocks had an enhancing effect on all reactivated trials.

Discussion
The present study was conducted to test whether the strength of declarative memories can be retroactively altered when reactivated in the presence of heightened noradrenergic arousal, but finds no evidence for this hypothesis.Recognition accuracy of objects that were reactivated followed by a shock was similar to objects reactivated without a shock, independent of the magnitude of the arousal response as indexed via skin conductance responses, and regardless of reactivation strength as indexed via reliving ratings.Since reactivation did enhance subsequent memory strength, and the arousal-inducing shocks caused increases in skin conductance throughout the second session, we conclude that our key manipulations were successful.Various explanations for this null-finding are considered below.
First and foremost, the absence of an enhancing effect of arousal may not be unique to reactivated memories: some previous studies have likewise failed to observe the enhancing effect of noradrenergic arousal on the consolidation of neutral stimuli, which is a core predicate of our hypothesis.One study that did find enhancement of consolidation by electric shock-induced arousal also reported a second experiment employing fMRI where the behavioral effect did not replicate (Schwarze et al., 2012).Using a highly similar design to investigate the necessary conditions for episodic memories to be enhanced by an arousing event, Dunsmoor and colleagues also did not observe enhancement of consolidation in three out of four experiments (Dunsmoor and colleagues, 2019).Notably, typical sample sizes both on the side of experiments that did find arousal-driven enhancements of consolidation and those that did not is rather low.This means it is hard to rule out the existence of this effect based on previous null-findings, but simultaneously suggests that its true effect size may be substantially below what is reported in published studies due to systemic biases in the publication process (Camerer et al., 2018).Here, we extrapolated the rationale behind consolidation studies beyond the strengthening of newly encoded events to reactivated memories, and did not find the predicted effect.One interpretation of this result is that reactivated memories are not susceptible to the same noradrenergic influences that newly encoded perceptions are.However, until a highly powered study replicates the arousal-driven enhancement of consolidation effect that inspired this study, it may also be that neither newly encoded perceptions nor reactivated memories are enhanced by arousal when it is isolated from other factors that affect memory.
Aside from not finding an effect of shock, a multilevel regression analysis showed no relation between the magnitude of arousal, as indexed using SCRs immediately following reactivation.Additionally, an exploratory analysis relating individual differences in average shock SCR to hit-rates showed no significant effect.Although contrary to our behavioral predictions, these findings do not necessarily contradict the hypothesis that NE released specifically in the basolateral amygdala drives the modulation of reactivated memories.One study that measured both skin conductance and amygdala activity found that only the latter was predictive of accurate memory for items that were followed by an arousal-inducing shock right after encoding (Hermans & Voogd, 2016).This suggests that heightened SCR responses to shock relative to no shock trials in the present study does not necessarily reflect heightened amygdala activity.Amygdala activation, however, seems to be a key driver of consolidation, as was convincingly demonstrated by Fig. 3. Performance on the object recognition by reactivation condition and testing group.Within the immediate testing group, reactivated objects were significantly less likely to be recognised than not reactivated objects.Subjects in the delayed testing group were more likely to recognise both reactivated and shock objects as compared to not reactivated objects.There was no main effect of testing group.(* = p < 0.05).Inman and colleagues (2018).In their study, epilepsy patients that were about to undergo surgery were presented neutral images that were sometimes immediately followed by direct electrical stimulation of the amygdala.This was found to enhance memory for images on a delayed test, but not on an immediate test.In our study, we unfortunately have no index of amygdala activation.However, the fMRI experiment by Schwarze et al. (2012) found no increases in amygdala activity following shock trials relative to shock trials, and -perhaps consequently -no effect of arousal on consolidation either.Since we used a highly similar design it is possible that the amygdala was unresponsive to shocks for similar reasons, but the exact conditions under which it does engage to strengthen reactivated memories for now remain elusive.
We may also look at the method used to induce and measure memory reactivation to explain the absence of the hypothesized strengthening effect by arousal on reactivation.During the second session, participants were instructed to indicate how strongly they relived all aspects of the encoded event corresponding to each presented context.Given that this also includes the details and narratives that were imagined during session one, it is possible that the objects themselves were not reactivated frequently or strongly enough to open the door for noradrenergic modulation.Yet, this is hard to rhyme with the fact that object recognition in the delayed testing group was enhanced for all reactivated trials, irrespective of shock.This indicates that at least some memories that were on the verge of forgetting got reactivated to the extent where they can be rescued by what can be considered as a testing effect (Chan & McDermott, 2007;Karpicke & Roediger, 2008), but these mechanisms are not further promoted by noradrenergic arousal.The multilevel regression analysis that included trial-level ratings of reliving further showed that noradrenergic arousal did not enhance subsequent memory at any level of subjective strength of reactivationa measure which itself was strongly associated with the likelihood of successful recognition.Finally, it may be that heightened noradrenaline following arousing stimuli lingers for longer than previous studies of memoryspecific consolidation suggest, in which case all reactivated trials should have benefited from the shocks throughout session two.This however leads to the prediction that participants who were more responsive to the shock should show a larger effect of reactivation on subsequent object recognition, which was also not supported by the data.In sum, given that we find effects of memory reactivation that are consistent with the existing literature, it seems unlikely that the present null finding is due to an insufficient reactivation procedure.
Various studies suggest that whether arousal enhances the consolidation of preceding neutral stimuli depends at least to some extent on cognitive factors (Dunsmoor et al., 2019;Knight & Mather, 2009), and these arguments may hold for enhancement of reactivated memories as well.According to the theory of arousal-biased competition, arousal specifically amplifies features and stimuli that are attended at the expense of those that are not (Mather & Sutherland, 2011).Given that during session two the contexts were actually presented whereas the objects were only reactivated, the contexts likely carried the largest attentional weights, meaning they may have been strengthened in the shock condition whereas memory for corresponding objects should consequently be weakened.Since we did not test memory for context images this hypothesis could not be explored in full, but the absence of an impairing effect of arousal on object memory argues against it.Beyond attention, Dunsmoor and colleagues showed that threat anticipation may be another necessary condition for arousal to enhance declarative memory (Dunsmoor et al., 2019).In a series of four experiments, recognition memory was only enhanced when objects were attended and encoded under the knowledge that a shock would be delivered immediately after.It may therefore be that heightened noradrenergic arousal only affects reactivated memories when they are attended under threat, which in the present study was not signalled.Finally, it may be that existing memories are only enhanced when they are directly relevant to the prediction of threat.One study showed enhanced episodic memory for stimuli belonging to a broader category that retroactively became predictive of electrical stimulation (Dunsmoor et al., 2015; but see Kalbe & Schwabe, 2021).Their design critically differs from ours in that the retroactive memory enhancement is not driven by the reactivation of specific memories, but does illustrate the potential importance of a meaningful connection between what was previously encoded and some biologically relevant outcome.This might explain why one study conducted in our lab found that neutral memories were enhanced after learning an overlapping emotional association, even though the original memory was not explicitly reactivated (de Vries et al., 2022).Notably however, the effect was not replicated in a highly similar study (de Vries et al., 2023).Whether retroactive effects of arousal on reactivated memories depend on the nature of their connection is an intriguing question to be addressed by future work.
The present study inhabits an intersection of two literatures that are both characterized by inconsistent findings: studies on the effects of noradrenergic arousal on memory for preceding neutral stimuli have reported mostly enhancements, but also impairments (Knight & Mather, 2009), and so have studies on the effects of more general emotional states on reactivated memories (Bos et al., 2014;Hupbach & Dorskind, 2014).As such, there is a plethora of competing explanations for this null finding that are difficult to weigh against each other, especially because pre-registered studies and replications remain extremely rare in emotional memory research.However, through the lens of evolution, our null-finding makes reasonable sense.The ultimate function of emotion-induced memory change, whether mediated by noradrenaline release in the amygdala or not, is to update one's models and schemas such that existing input-output relations are remapped to result in more favorable outcomes.Broadly this would require that there is something new to learn about what a reactivated memory predicts, or the decisions it affords.It may be argued that memory reactivation by means of partial reexposure inherently involves some degree of prediction error.Yet, if that were the case in the present study, it did not facilitate an updating effect of arousal.Alternative experimental designs that place the adaptive function of memory first and foremost may be better poised to generate novel insights into the dynamic mechanisms of memory.In situations such as the present experiment, in which the circumstances of arousal cannot be used to guide future behaviour, we argue that it does not affect neutral memories.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Design outline showing the three conditions and three phases of the experiment.During session one, participants encoded 120 memories, each consisting of a background image (context) and a foreground image (target object).In session two, the 40 target objects in the shock condition were reactivated by presenting their corresponding context images which were immediately followed by uncomfortable electrical stimulation administered to the wrist.Memories of objects in the reactivation condition were similarly reactivated without the shock, and the contexts corresponding to objects in the no reactivation condition were not presented.The third session consisted of a recognition test in which all target objects were shown again mixed in with an equal amount of lure objects.