The Effect of Mood on Shaping Belief and Recollection Following False Feedback

The current study examined how mood affects the impact of false feedback on belief and recollection. In a three-session experiment, participants �rst watched 40 neutral mini videos, which were accompanied by music to induce either a positive or negative mood, or no music. Following a recognition test, they received false feedback to reduce belief in the occurrence of the events displayed in some of the videos (Session 2). This was followed by an immediate memory test and a delayed memory assessment one week later (Session 3). The results revealed that participants in negative mood reported higher belief scores compared to those in positive moods, despite an overall decline in belief scores for all groups following the false feedback. Notably, individuals in negative moods exhibited less reduction in their belief scores after encountering challenges, thereby maintaining a higher accuracy in their testimonies. Over time, a reduction in the clarity of participants’ memory recall was observed, which correspondingly reduced their testimony accuracy. This study thus indicates that mood states play a role in shaping belief and memory recall under the in�uence of false feedback.


Introduction
Testimonies are vital in court proceedings, often being the key evidence, especially in the absence of objective evidence like DNA (Davis & O'Donohue, 2004; Kassin, 2008;Howe & Knott, 2015).Understanding how various factors, like the emotional state of witnesses and social feedback, affect memory accuracy is critical.Misleading feedback has been shown to signi cantly distort event recollections, resulting in testimonial inaccuracies (Douglass et Wright et al., 2000).Furthermore, the mood of a witness has been found to substantially in uence cognitive processes, including memory, thereby impacting the reliability of testimonies (Bell et al., 2019;Chai & Wu, 2022;Clausi et al., 2017;Fartoukh et al., 2014;Karjalainen, 2006).With this understanding of the intimate link between mood, memory, and the accuracy of testimonies, the current experiment aimed to examine the impact of positive and negative moods on testimony accuracy following false social feedback.
Building on this understanding of memory's fallibility, the in uence of social feedback has been recognized as pivotal in shaping memory recall and reporting, especially in eyewitness testimonies ( (Loftus, 2005) highlights how exposure to misleading post-event information, such as false social feedback, may lead to the integration of false information into one's recollection of the original event.For instance, in Zaragoza et al.'s (2001) study, participants rst watched a video and then were introduced to fabricate details about events not depicted in the video.Additionally, they experienced con rmatory feedback, receiving positive reinforcement for these invented responses.This study found that participants often integrated these false details into their memory.Con rmatory feedback, on the other hand, increased participants' con dence in their inaccurate recollections, leading them to believe more strongly in their erroneous identi cations of details in the video.
Failing to report information or memory omissions can frequently occur during memory retrieval processes (Armstrong et al., 2000;Healy & Nairne, 1985) and can in uence the trustworthiness of testimonies (Douglass et Vo et al., 2022).Recent empirical research has not only focused on the impact of feedback on omission errors but also examined its effect on the belief in occurrence of events (e.g., Azad et al., 2022).
Belief in the occurrence of events, in the context of autobiographical memory, has been posited as a primary driver for behavioral change (Scoboria et al., 2004(Scoboria et al., , 2015)).When memory is distorted by false feedback, it can shift foundational beliefs about an event, altering behavior and affecting one's ability to accurately recall and testify.Therefore, understanding the complex relationship between beliefs and misleading feedback is essential for evaluating the accuracy of testimonies.

The Power of Autobiographical Belief
Autobiographical memory is comprised of two primary elements: Belief about the occurrence of an event and the recollection of the event's speci c details (Scoboria et al., 2014).Belief pertains to the truth values associated with an event's occurrence (Scoboria et al., 2014).In contrast, recollection involves the retrieval of speci c details, emotions, and sensory experiences linked to the event (Scoboria et al., 2014).
Research suggests that belief about the occurrence of an event exerts a more signi cant in uence on behavior change than recollection (Scoboria et al., 2004;Scoboria et al., 2015; but see Burnell et al., 2022).
Social feedback has been demonstrated as a potent method to reduce belief in the occurrence of experiences, such as being informed by someone that the recalled event never occurred (Li et al., 2020(Li et al., , 2023a, b;, b;Otgaar et al., 2013bOtgaar et al., , 2017;;Scoboria et al., 2018;Wang et al., 2017Wang et al., , 2019)).For example, Otgaar et al. (2016) conducted a study in which participants were asked to perform, imagine, or hear about actions, such as breaking a toothpick.Then participants repeatedly imagined these actions.For the actions they identi ed as "performed", a random selection was subsequently challenged by social feedback, implying that these actions were not originally performed.Belief or recollection was relinquished in 38.7% of adult participants and 51.1% in children.Similarly, Li et al. (2020) used an imagination in ation procedure involving both familiar (e.g., drinking from a bottle) and bizarre actions (e.g., licking a cup).They found that 73.8% (n = 79) of participants gave up their belief for the remembered actions following false social feedback.
Prior research predominantly explored how individuals integrate incorrect information into their memories due to false feedback (Horry & Palmer, 2012).However, there has been limited exploration into the speci c effects of how such feedback alters beliefs about events and how this subsequently in uences the precision of the information reported.This study aims to bridge this gap by delving into how false feedback not only affects belief adjustments but also examines its broader impact on the accuracy of memory reporting and the prevalence of omission errors, offering new insights into the nuanced ways false feedback can in uence memory beyond mere incorporation of incorrect information.
Positive moods (e.g., pleasure, joy, and excitement), for example, are associated with heuristic processing, making individuals more susceptible to external in uences and misinformation due to a general focus on information (Bless & Fiedler, 2012;Forgas & George, 2001).Conversely, negative moods (e.g., sadness, anger, anxiety) prompt a more systematic processing strategy, where individuals pay more attention to detail and exhibit enhanced memory accuracy and detail-oriented recall (Bless et al., 1996;Gasper, 2004;Schwarz, 2000;Tiedens & Linton, 2001).Thus, mood states affect how information is processed, leading to variations in testimony accuracy, as evidenced by various studies ( When extrapolating these ndings to eyewitness testimony, it becomes evident that mood can signi cantly in uence its quality (Christianson, 1992 (Christianson, 1992;Christianson & Hübinette, 1993).Witnesses in negative moods are likely to provide more precise and accurate descriptions, enhancing testimony reliability, while those in positive moods might offer more generalized accounts, potentially affecting testimony accuracy (Christianson, 1992;Christianson & Hübinette, 1993;Goodman et al., 2014;Forgas, 1995;Bless & Fiedler, 2012).
Understanding the impact of mood on individuals' responses to false social feedback is essential, especially regarding memory recall and belief formation.Research indicates that an individual's response to false social feedback is signi cantly in uenced by their current mood state (Forgas, 1995).For example, individuals in positive moods are more susceptible to false feedback, likely leading to greater belief change and omitting key details by integrating misleading information (Bless & Fiedler, 2012;Forgas, 1995).In contrast, persons in negative moods, like sadness or anxiety, show resistance to false feedback due to their detail-oriented processing, leading to more accurate memories and less likelihood of overlooking important information (Bookbinder & Brainerd, 2016).

The Current Study
Drawing from the extensive literature on memory accuracy in legal contexts ( To address our research objectives, we designed an experiment segmented into three sessions.Participants were randomly assigned to either a positive music, negative music, or no music (control) group, with mood states induced via background music for the rst two groups.The rst session involved watching neutral mini videos and undertaking a recognition task, focusing on recalling details of the speci c events displayed in the videos.This was followed by a second session during which false feedback was given that aimed at challenging participants' beliefs about speci c memories of the videos.A subsequent memory test was conducted to determine if each description was remembered, believed to have occurred, or identi ed as not being part of the video content.One week later, their recollection of the video content was re-evaluated through a free recall task and a repeated memory test.
Guided by prior research, our hypotheses were as follows.Research indicates that individuals experience signi cant belief reduction when confronted with contradictory social feedback (Li et al., 2020(Li et al., , 2023a .Therefore, we hypothesized that exposure to false feedback would lead to a notable reduction in belief.This reduction, we anticipated, would in uence the accuracy of testimonies and result in an increase of omission errors.Additionally, in line with the Affect-Infusion Model (Forgas, 1995), we hypothesized that mood states would in uence cognitive processing strategies.Speci cally, we expected that participants in positive moods to be more vulnerable to false feedback compared to those in negative moods, leading to more signi cant belief reduction, increased detail omission, and reduced testimony accuracy.Conversely, we hypothesized that participants in negative moods, owing to their more detailed focus, would demonstrate greater resistance to false feedback than participants in positive or neutral moods.This resistance was expected to result in a smaller reduction in belief and, potentially, an increase in the precision and reliability of their recollections.

Method Participants
The participants were recruited from KU Leuven and Maastricht University, and online via social media.
To be eligible, all participants had to be at least 18 years old.Previous studies using music-based mood inductions have reported effect sizes with Cohen's d values ranging from 0.47 to 0.99 (Chung et  For our power analysis, we opted for a conservative medium effect size of .25.This ensured that our study remained adequately powered even if the actual effect size would turn out to be somewhat lower than what was observed in previous research.The a priori power analysis using G*Power (Faul et al., 2007; i.e., test family: F-test, xed effects, omnibus, one-way, f = .25,α = .05,power = .80,three groups) indicated that a sample of 159 participants (53 participants per group) was needed.
At the outset of the study, there were 197 participants.We excluded 24 participants due to their inability to complete the second session, and an additional 6 participants were removed because they failed to comprehend the experimental instructions, as evidenced by their non-compliant responses during the free recall phase.Thus, our nal sample consisted of 166 participants, 55 males and 111 females, with ages ranging between 18 and 39 years (M age = 23.04,SD = 3.93; see Table 1 for more details).Participants received 10 euros vouchers for their participation.Ethical approval for the project (G-2022-5371) was granted through the Social and Societal Ethics Committee (SMEC) and Privacy and Ethics (PRET), KU Leuven.This project was preregistered on the OSF, https://osf.io/f8t3b,and all the materials and data are available at https://osf.io/dv5hm/.Neutral Video.We selected 40 mini videos from the Chieti Affective Action Videos database, a publicly accessible resource originally developed by Di Crosta et al. (2020).Each selected video, lmed from a rst-person perspective, captures a simple action such as folding a chair, lasting for about 15 seconds.These videos were previously calibrated by Di Crosta et al. (2020) using 9-point scales for measuring arousal and valence.In our research, we also obtained arousal and valence ratings of the selected videos to establish their emotional intensity and affective value.On average, the arousal score of the videos was low (M = 2.44, SD = 0.71), indicating that they are predominantly neutral and unlikely to elicit strong emotional responses.In line with this, the average valence score was around the midpoint at 5.08 (SD = 0.63).To mitigate primacy and recency effects, we selected eight of the 40 mini videos as llers: four were positioned at the beginning and four at the end, while the remaining 32 videos were shown in random order.Filler videos did not appear in the recognition task.We preferred to use neutral videos so that it became possible to study the independent and combined effects of different mood states and social feedback, while minimizing potential confounding impact associated with emotional or traumatic experiences.
Mood Ratings.We used the three dimensions used by Zhang et al. (2017) for mood measurement to evaluate participants' mood states at different stages: (1) Happiness, (2) Sadness, and (3) Excitement or Calmness.These mood dimensions were assessed using three 100-mm visual analog scales (0 = Not at all, 100 = Extremely).
Recognition Test.The recognition tests consisted of 32 detailed descriptions of the videos that were watched by the participants.These included 16 true descriptions accurately re ecting the content of the videos (e.g., I put a brown bracelet on my left wrist) and 16 false descriptions misrepresenting certain details (e.g., I opened a black umbrella, while in video the umbrella remained closed).To ensure the objectivity and accuracy of these descriptions, they were independently evaluated by research assistants.The inter-rater reliability of these descriptions was assessed using Cohen's Kappa coe cient, a statistical measure for determining agreement between evaluators.The obtained Kappa value was 0.71, indicating substantial agreement and a rming the reliability of the descriptions in accurately re ecting the content of the videos.
Belief and Recollection Ratings.The experimenter explained to participants the difference between recollection and belief before the recognition test: "Recollection refers to the mental re-experiencing of an event.Belief, on the other hand, refers to the extent to which people believe the event really occurred in the way they remembered it".Thus, participants were asked to complete two separate questions in order to rate belief, "Do you believe that the event described here has appeared in the video collage (regardless of whether you remember or not)?" (1 = De nitely did not happen, 8 = De nitely did happen) and the recollection, "Do you actually remember that the event described here has appeared in the video collage?" (1 = No memory of the event at all, 8 = Clear and complete memory; Scoboria al., 2004).
Memory Test.In line with the methodology used by Azad et al. (2022), our memory test used a similar response structure.Participants were presented with each test item and asked to select one of three options: (a) Remember it was in the video (i.e., they could clearly remember seeing the test item in the video), (b) Believe it was in the video (i.e., they believed the item was in the video but had no clear recollection of it), or (c) Not in the video (i.e., they neither remembered nor believed the test item was in the video).

Procedure
Initially, participants were informed that they would be involved in viewing a series of mini videos.Following the informed consent procedure, participants provided their demographic data.The study was organized into three distinct sessions (see Fig. 1).Session One.Participants were randomly assigned to one of three different experimental conditions: (1) positive music, (2) negative music, or (3) no music (control group).For Conditions 1 and 2, music excerpts were played continuously in a loop at a comfortable volume for the participant.Before starting the experiment, participants were asked to rate their current mood.During listening to the music, participants were asked to complete a personal information questionnaire to prevent them from being aware of the purpose of the manipulation.To test the success of the mood induction, participants were asked to assess their current mood again after listening to 30 seconds of background music (Ferrer et al., 2013).Then, they were asked to watch 40 mini videos attentively, either with the assigned music continuing to play in the background or no music for the control group.They were instructed to imagine themselves performing the actions depicted in the videos and to try to remember as many details as possible.After this, participants were required to reassess their current mood state once more.To ensure that participants attended to these videos, we asked them to rate their attention to these videos using a sliding scale to measure their responses to three questions, "How closely did you pay attention to the video?", "How emotionally engaged were you while watching the video?", and "Were you able to maintain your focus on the video throughout its duration?"The scale ranged from 0 to 100, with 0 indicating no attention or engagement and 100 indicating complete attention and engagement.Finally, a distractor task was administered to remove the video visuals from working memory.The distractor task was a "easy" paper-and-pencil Sudoku puzzle, taken from an online database (Tadmor et al., 2016).Participants were asked to complete as much of the puzzle as possible within 90 seconds.
Session Two.After a 20-minute break, participants completed the recognition task individually.Participants only received the descriptions of the videos to avoid additional exposure to the videos.All the descriptions were presented in a predetermined random order.Participants were asked to identify if the description on the screen was a true/false description of the video by clicking on a "true" or "false" button.If a description was identi ed as "true", participants had to provide recollection and belief ratings.
After the recognition task, participants were informed that they would receive feedback on their incorrect identi cations of descriptions that did not appear in the videos.We challenged the rst ve correct identi cations of true descriptions made by each participant by informing them that these descriptions did not actually appear in the presented videos.For the remaining descriptions, participants received accurate feedback.Subsequently, participants were asked to provide belief and recollection ratings for all the feedback items.Following this, participants took part in a memory test, containing the same descriptions as those used in the recognition task.
Session Three.One week later, participants had a free recall of the content of the mini videos.
Participants were asked to write down any details in the video that they could remember.Then, participants completed the memory test again.

Recollection and Belief Changes Across Different Groups
Our hypothesis was that participants in the negative mood group would show a less pronounced decrease in belief scores following false social feedback, compared to those in the positive and no music groups.This deviated from our pre-registered analysis, as our focus shifted to examining belief score changes before and after the challenge.Therefore, we separately analysed belief and recollection scores.
The main effect analysis of time revealed a noticeable drop in belief scores across all groups from prechallenge to post-challenge, t(165) = 7.79, p < .001,95% CI [0.60, 1.00], Cohen's d = 0.71.Speci cally, the mean belief score before the challenge was 6.89 (SD = 0.77), which decreased to 6.09 after the challenge (SD = 1.35).Subsequent multiple comparisons revealed that the belief scores in the negative group were signi cantly higher than those in the positive group, with a difference of 0.45 (SD = 0.16), p = .020,95% CI [0.05, 1.14].However, the difference between the negative group and the no music group was not statistically signi cant, showing a difference of 0.38 (SD = 0.16), p = .066,95% CI [-0.02, 0.77].Furthermore, the variation in belief scores between the positive and no music groups was also not signi cant, with a difference of -0.07 (SD = 0.16), p = 1.00, 95% CI [-0.47, 0.33] (see Table 3 for detailed scores of each group).
To further understand the relationship between participants' belief scores and their mood states, we carried out a Pearson correlation analysis.This analysis was exploratory in nature and was not included in our preregistered plans, allowing us to explore these two factors in more depth.Our analysis revealed that prior to the challenge, belief scores were signi cantly and positively correlated with mood scores, r(166) = 0.23, p = .003.After the challenge, a similar pattern was observed.The belief scores remained signi cantly and positively associated with mood scores, r(166) = 0.33, p < .001.

Differences in Testimony Accuracy Across Groups
In our study, testimony accuracy was operationalized de ned as the ability of participants to accurately identify details as they occurred in the video.We hypothesized that exposure to false feedback would lead to belief reduction, resulting in more "Not in the video" responses, which in turn would signify less accurate testimonies.This effect was expected to be more pronounced in participants with positive moods, who were hypothesized to be more susceptible to such feedback, while those in negative moods were expected to show greater resilience and, consequently, better accuracy in their testimonies.The focus on the rst ve true memories challenged for each participant allowed for a targeted analysis of "Not in the video" choices, where a higher frequency indicated a stronger decrease in the accuracy of the testimonies.
To explore this mediating effect, we adopted the approach recommended by MacKinnon and Lockwood (2004) and utilized the SPSS macro developed by Preacher and Hayes (2008) for Bootstrap mediation effect testing.This methodology diverged from our pre-registered analysis approach as an exploratory addition.By using belief as a mediating variable, this analysis method allowed us to further elucidate the intricate relationships among mood states, belief, and the accuracy of testimonies.This nuanced approach provided deeper insights into how these elements interact, offering a more comprehensive understanding of their interconnected dynamics.Our analysis revealed that mood states signi cantly in uenced belief reduction (a path coe cient = 0.51, p = .027,95% CI [0.06, 0.96]), and belief reduction in turn signi cantly impacted testimony accuracy (b path coe cient = 0.28, p = .003,95% CI [0.10, 0.45]).Further examination showed that the indirect effect of mood states on testimony accuracy through belief reduction was statistically signi cant (a*b path coe cient = 0.14, SE = 0.05, z = 3.10, p = .002,95% CI [-0.01, 0.17]).These ndings clearly indicate that belief reduction plays a signi cant mediating role between mood states and memory accuracy.In the subsequent analysis, we extended our examination to assess the enduring impact of belief reduction as a mediating variable between mood states and testimony accuracy over a period of one week.We used the same Bootstrap approach for mediation effect assessment.Our ndings revealed that the in uence of mood states on belief reduction remained signi cant over a week (a path coe cient = 0.51, p = .027,95% CI [0.06, 0.96]).Additionally, belief reduction continued to signi cantly impact testimony accuracy (b path coe cient = 0.30, p = .015,95% CI [0.06, 0.54]).The indirect effect of mood states on testimony accuracy through belief reduction was found to be statistically signi cant (a*b path coe cient = 0.15, SE = 0.04, z = 3.74, p < 0.001, 95% CI [-0.01, 0.15]), underscoring the substantial mediating role of belief reduction.Our study also explored the role of belief reduction as a mediator between mood states and omission errors in a free recall task.The mediation effect of belief reduction on omission errors was not statistically signi cant (a*b path coe cient = -0.06,SE = 0.03, z = -2.07,p = 0.039, 95% CI [-0.10, 0.01]) The signi cant effect of mood states on belief reduction (a path coe cient = 0.51, p = 0.027, 95% CI [0.06, 0.96]) was again noted.However, belief reduction did not signi cantly affect the likelihood of participants committing omission errors in the free recall task (b path coe cient = -0.11,p = 0.119, 95% CI [-0.25, 0.03]).(More details about omission errors are provided in the supplementary materials which are available at https://osf.io/dv5hm/).

Longitudinal Effects on Testimony Accuracy
hypothesize that testimony accuracy might decrease over time, as the participants' memories become less precise.This phenomenon, often referred to as memory decay, might manifest as an increase in "Not in the video" responses over the week-long interval.To assess the long-term effects on testimony accuracy, a mixed-model repeated measures ANOVA was conducted.a 3 (Group: negative, positive vs. no music) x 2 (Time: immediate vs. one week later) mixed ANOVA was conducted.This analysis approach was part of our pre-registered plan.The frequency of "Not in the video" responses was the dependent variable.The results revealed no signi cant interaction between Group and Time, F(2, 163) = 1.64, p = .198,partial η 2 = .02.Nonetheless, a signi cant main effect of Time was observed, F(1, 163) = 11.98,p = .001,partial η 2 = .07.Additionally, a signi cant main effect for the Group was found, F(2, 163) = 5.53, p = .005,partial η 2 = .06.The frequency of "Not in the video" responses showed a signi cant difference between the immediate memory test and one week later, t(165) = -3.16,p = .002,95% CI [-0.33, -0.08], Cohen's d = -0.15.Speci cally, the average frequency of "Not in the video" responses during the immediate memory test was 1.01 (SD = 1.13), which increased to 1.21 one week later (SD = 1.42) (see Fig. 3).More details about the differences in the "Not in the video" response frequencies among different groups during the immediate and delayed memory test are provided in the supplementary materials.

Discussion
In the current experiment, we explored how different mood states (positive, negative, or neutral) in uenced belief and testimonial accuracy in response to false feedback about neutral video material.
The study revealed that music-induced moods signi cantly affected belief changes: participants in positive moods showed higher susceptibility and greater decline in belief when challenged, while those in negative moods demonstrated less belief reduction.Our ndings emphasize the crucial role of belief reduction as a mediator between mood states and the choice of "not in the video", a key measure of testimonial accuracy.This underscores the signi cance of belief in processing memory under misinformation.Notably, participants in negative moods displayed resistance to false feedback, leading to more accurate testimonies compared to those in positive or no music conditions.However, mood states and belief reduction did not signi cantly in uence omission errors in free recall tasks.
The Impact of False Social Feedback on Belief Across Different Mood Our study rmed social feedback as an effective method to reduce belief strength (Li et al., 2020;Otgaar et al., 2017).After receiving false feedback, all groups experienced a signi cant decrease in belief scores, yet no similar decline was observed in recollection scores.This suggests a potential separation between belief and detailed memory recall, supporting Scoboria et al.'s (2014) ndings that speci c recollections can persist even as belief weakens.This implies that while belief may uctuate due to external information, the speci c details of memories remain relatively stable.This stability might stem from the inherent resilience of memory structures, capable of withstanding misleading external information in the short term (Lorenz, 2010).In contrast, beliefs, as more malleable psychological constructs, are more easily in uenced by external factors, leading to adjustments in how events are perceived without necessarily altering the concrete details of the memories.
In addition, our ndings reinforce the Affect-Infusion Model (Forgas, 1995), which suggests that individuals in different mood states employ varied cognitive processing strategies, illustrating that mood states signi cantly sway belief adjustments.We observed that participants in negative mood states exhibited higher levels of belief scores compared to those in positive moods.This could be attributed to negative emotions prompting individuals to pay more attention to details, resulting in enhanced memory accuracy and a detail-oriented recall process (Bless et al., 1996;Gasper, 2004;Schwarz, 2000;Tiedens & Linton, 2001), thus fostering greater con dence in their own memories.Furthermore, a more in-depth analysis of belief reduction scores across groups revealed that participants in negative moods exhibited less reduction in belief scores compared to those in positive moods or the control group.This trend likely stems from the more analytical and systematic approach to information adopted by individuals in negative moods, enhancing their resilience against misleading cues and leading to a smaller reduction in belief (Forgas, 1995).Controversely, individuals in positive moods, who tend towards heuristic processing, were more prone to accepting external misinformation.This led to a notable decrease in their belief scores, indicating a higher susceptibility to false feedback.
However, our study observed no notable difference in belief reduction between the positive mood and no music groups when encountering false feedback, suggesting that the intensity of music-induced mood may not signi cantly impact cognitive processing.This lack of disparity may imply that the effect of positive mood on cognitive processing and openness to misinformation may vary among individuals.It also highlights the potential for music, irrespective of its emotional content, to serve as a cognitive distraction, in uencing participants' response to misinformation.This nding points to the need for further research on the in uence of emotional stimulus intensity and individual differences in cognitive processing.Future studies should consider neutral auditory settings, such as white noise, and vary music characteristics to more accurately assess its effect on cognition and susceptibility to false feedback.

Dynamics of Mood and Belief Reduction Shaping Testimonial Accuracy
In this study, we examined how mood states in uence testimony accuracy, focusing on the role of belief reduction as a mediating factor.Moods profoundly affect our perception and interpretation of the world ).We measured testimony accuracy through participants' choices of the "Not in the video" option, similar to legal inquiry methods where unreported or omitted real event details lead to testimonial inaccuracies, thereby affecting their overall reliability.In our study, frequent selection of this option indicated that participants repeatedly overlooked or forgot real details in the video, re ecting a decline in the accuracy of their testimonies.This approach not only served as an intuitive measure of the participants' memory regarding the video content but also effectively highlighted how they responded to and were in uenced by the misleading feedback.
Our ndings reveal that mood states signi cantly impact testimonial accuracy by in uencing belief reduction.Participants in positive moods were more prone to accepting misleading feedback, often choosing the "Not in the video" option, indicating lower testimonial accuracy.This might be attributed to their openness to new information, a characteristic of positive moods (Bless & Fiedler, 2012;Forgas & George, 2001).Conversely, those in negative moods demonstrated greater resilience against such feedback, likely due to a vigilant, critical approach to information processing, as noted by Storbeck and Clore (2005) and LeBlanc et al. (2015).Our results showed that this led to fewer selections of "Not in the video", signifying more accurate testimonies.This might be because negative moods enhanced the participants' critical analysis of information, making them more cautious in memory judgments and reducing the likelihood of misinterpreting actual events.
However, the mediating role of belief reduction between mood states and testimony accuracy was not as pronounced one week later.This can be attributed to the natural progression of memory processes, where initial strong beliefs, shaped by mood and feedback, are subject to retention and forgetting.Over time, beliefs may be internally, leading to adjustments in memory recall, independent of the original mood in uence.Furthermore, the initial impact of mood on belief formation gradually diminishes, giving way to the stronger in uence of external feedback content on memory.This suggests that while immediate responses to stimuli are crucial, the long-term dynamics of memory and belief involve a complex interplay of cognitive reassessment and external information.Additionally, we noted an increase in the selection of the "Not in the video" option over time, re ecting a natural decay in memory traces in line with classic memory theories such as Ebbinghaus's forgetting curve (Bahrick et al., 1993;Ebbinghaus, 1885).This suggests that as memories become less vivid, eyewitnesses may increasingly rely on external cues, potentially negatively affecting the accuracy of their recollections.However, our ndings did not reveal a signi cant interaction between mood states and time on memory test outcomes.
This indicates that the in uence of mood on memory may wane over time, particularly when external feedback is introduced.These ndings point towards the importance of long-term memory stability and inherent memory processes in shaping testimony accuracy, rather than initial mood states.This underlines the need for further research to explore how various elements, including mood, time, and external feedback, collectively in uence memory accuracy in high-stakes contexts like legal testimonies.
Lastly, in our study, we examined belief reduction as a potential mediator between mood states and omission errors, but contrary to our hypothesis this mediation effect was not signi cant.This supports the idea that while mood states and social feedback can in uence memory formation, their impact on established memories may be limited, echoing Brainerd et al. (2002) who found that formed memories tend to remain stable.This indicates a need for a more detailed analysis of how these factors interact in real-world scenarios, and the potential role of other variables like prior experiences or mood regulation strategies in memory processes.

Limitations
This study, while providing valuable insights, has certain limitations.Primarily, it focused on short-term mood and memory changes, which restricts the understanding of long-term mood effects on memory stability.Consequently, the implications for how long-term mood states in uence memory and testimony accuracy might not be fully captured.Additionally, the reliance on free recall tasks, which are subject to participants' subjective interpretations, raises concerns about potential biases and generalizations in memory recall.The use of music for mood induction also represents a limitation, as it may not re ect the complex spectrum of real-life mood variations and their impacts on memory.Moreover, our memory test approach, modelled after Azad et al. (2022), did not clearly distinguish between recollection without belief and absence of both recollection and belief.This distinction is crucial for understanding memory processes, especially in scenarios where recollection and belief in accuracy may not align.Future research should address these limitations to deepen our understanding of the intricate relationship between mood, belief, and memory recall.

Conclusion
Our research investigated the impact of mood states on testimony accuracy in contexts with false social feedback, revealing nuanced interactions between mood, belief, and memory recall.Negative moods were linked to more robust belief retention and greater recall accuracy, even when confronted with misleading information.In contrast, positive moods tended to lead to reduced belief stability and lower recall accuracy.These trends suggest that mood states subtly shape belief systems, which in turn in uence how accurately memories are recalled.Our research sheds light on the intricate cognitive process involved in memory accuracy, particularly in contexts requiring precise recall.The insights gained extend beyond memory and belief dynamics, presenting intriguing avenues for future research in cognitive psychology and related elds.

Declarations Ethical Approval
Ethical approval for this project was obtained from the Social and Societal Ethics Committee (SMEC) and Privacy and Ethics (PRET) at KU Leuven.The reference number for the ethical approval is G-2022-5371.In accordance with the guidelines provided by the ethical committees, all participants in the study were informed about the purpose and procedures of the research, and their consent to participate was obtained prior to their involvement.Con dentiality of the participants' personal information and data was strictly maintained throughout the study.All data collected were anonymized and stored securely, with access restricted to the research team.Any identifying information that could potentially breach con dentiality was removed or pseudonymized.

Funding
The research presented in this paper was support of the China Scholarship Council (grant number: 202007650008), awarded to Chunlin Li, and the FWO Research project grant (grant number: G0D3621N), which was awarded to Henry Otgaar.

Disclosure statement
No potential con of interest was reported by the authors.Figure 3

Figures Figure 1
Figures

Table 1
The number and age of participants in the negative, positive, and no music groups MaterialsEmotion-inducing Musical Stimuli.We used two musical stimuli, each with a duration of approximately 1 minute and 20 seconds.These were excerpts from Proko ev's 'Alexander Nevsky: Russia Under the Mongolian Yoke', which was used to induce negative mood, and Bach's 'Brandenburg Concerto No. 3', which was employed to elicit positive mood.These speci c pieces have a proven track record of successfully evoking different mood states in human participants, as substantiated by prior research(Fresco etal., 2007; Rowe et al., 2007; Stopa & Waters, 2005; Zhang et al., 2017; Zimasa et al., 2017).

Table 2
Mood scores for negative, positive and no music groups before and after mood manipulation.

Table 3
Recollection and belief scores for negative, positive, and no music groups before and after the provision of challenging feedback *** p < .001

Table 4
Results of analysis testing belief reduction as a mediator between mood states and testimony accuracy.

Table 5
Results of analysis testing belief reduction as a mediator between mood states and testimony accuracy for one week later.