The paradigm of CIE was originally developed by Wilkes and Leatherbarrow (1988), has been continuously used (Susmann and Wegener, 2023), in which participants first read a story with misinformation, such as a warehouse fire, and were led to believe that there might be flammable material in the wardrobe, which was the cause of the fire when the material was presented. Participants in the correction group received corrections for the misinformation they had read (i.e., they were informed that the wardrobe was empty), and no further corrections were provided to the non-correction group. Finally, the participants were asked to complete a Discrepancy Detection Test, in which they recalled the content of the material they read and inference about the cause of the warehouse fire.

Kendeou et al. (2014) proposed the Knowledge Revision Theory to explore the mechanisms inherent in the continued influence effect of misinformation, including five principles: encoding, passive activation, coactivation, integration, and competing activation (Kendeou et al., 2013). The encoding principle states that once information is present (correct or incorrect information) it is first encoded into memory; the passive activation principle states that all information is passively activated, independent of whether the activated information is correct or relevant, and the coactivation principle is an outgrowth of both the encoding and passive activation principles and is how new information is interconnected with the previously acquired information. The integration principle is the basis for knowledge correction, where new information will begin to dominate the entire information network as the amount of newly acquired correct information continues to increase; finally, the activation of corrected information competes with the activation of previously incorrect information for the final output (competitive activation principle). Several works have demonstrated that CIE is partly driven by information familiarity (Ecker et al., 2017). Throughout the corrective process, misinformation is frequently repeated, which increases its encoding strength and familiarity. As a result, CIE is improved since familiar claims are more likely to be believed (Bailey et al., 2021). Consistent with this logic, repeatedly presenting corrected information enhances its encoding and improves individuals’ memory of the correction, making it seem more familiar and more likely to be accepted (Ecker et al., 2014). Consequently, the activation of the corrected information competes with that of the original misinformation, as the repetition influences the strength of encoding.

When two or more group members work together to recognize, retain, and recall what they have learned, the process is called collaborative memory (Basden et al., 2000; Wissman, 2020). A typical paradigm of collaborative memory consists of three phases (Wissman, 2020): the learning phase, the interference phase, and the recall phase. In the learning phase, all participants individually learned the study material, followed by the interference phase. Finally, there was a recall phase where participants were divided into collaborative and nominal groups. The participants in the collaborative group were divided into groups of three and worked together to recall the material they had learned. The members of the collaborative group were encouraged to freely discuss with each other and recall as many items as possible. The participants in the nominal group were tested individually for recall. The effect of collaboration on memory was investigated by comparing the number of items recalled in the recall test between the collaborative group and the nominal group. Many studies have subsequently incorporated one or more individual tests following the initial collaborative memory procedure to examine if collaboration has a lasting effect on subsequent individual memory (Congleton and Rajaram, 2011; Jalbert et al., 2021).

There is a large body of studies suggesting that collaborative recall tends to result in higher memory accuracy compared to nominal groups, known as post-collaborative memory facilitation (Rossi-Arnaud et al., 2020). Post-collaborative memory facilitation is thought to primarily involve two cognitive processes: re-exposure and cross-cuing (Congleton and Rajaram, 2011). Re-exposure refers to the opportunity for participants to relearn during collaborative recall when they hear someone else report items that they were unable to recall when alone (Mundorf et al., 2021). In addition, cross-cuing occurs when a person hears other group members recalling information that they have forgotten and uses this as a cue to recall additional information (Guynn, 2024).

Prior researches have categorized collaboration conditions as turn-taking collaboration and free collaboration (Abel and Bäuml, 2020; Wright and Klumpp, 2004). Turn-taking collaborative recall involves each person taking turn reporting one piece of information at a time, moving to the next person when a person reports no recall, and so on (Nie and Guo, 2023). The free collaborative recall involves participants as a whole recalling as many of the items they have just learned as possible, with free consultation, discussion, and reminders among group members. Some studies have found significant differences in the effects of various collaborative conditions on memory (Rajaram et al., 2020). However, the results of the current study are inconsistent. For example, Thorley and Dewhurst (2007) found that turn-taking collaborative groups demonstrated improved memory after collaboration, and there was no significant difference in recall outcomes between groups that free collaboration group and nominal group. Weihai et al. (2015) found that both turn-taking collaboration and free collaboration resulted in improved memory after collaboration. Based on previous contradictory findings, various collaboration conditions have varying effects on memory.

At the same time, we found that the time of collaborative recall, as set by previous studies under different collaboration conditions, is inconsistent. The time of recall directly affects the encoding strength of memory (Mundorf et al., 2021), which may be a factor influencing the varying degree to which different collaboration conditions affect memory. Most studies on collaborative memory do not have an explicit time limit. For example, Abel and Bäuml’s (2020) study, where the recall phase was not explicitly time-limited free collaboration was used, and social interaction was found to enhance memory. Numerous studies have also demonstrated this phenomenon (Harris et al., 2012; Bärthel et al., 2017). In contrast, another one utilized a free collaboration condition, allowing participants 2 min to collaboratively recall a scene. The study found that the recall scores of the free collaboration group were not significantly different from those of the individual group in an immediate recall test (Huff et al., 2016). In addition, Abel and Bäuml (2020) explored the effects of turn-taking collaboration on memory and found that the turn-taking collaboration group did not show post-collaboration memory facilitation when there were no time constraints (Wright and Klumpp, 2004). Weihai et al. (2015) set the recall time to 8 min for free and turn-taking collaboration groups and 4 min for individual groups and found that both turn-taking and free recall produced post-collaboration memory facilitation, and the amount of post-collaboration memory facilitation for turn-taking recall was higher than that for free recall. From previous studies, we know that the recall time in different collaborative conditions affects memory results. Therefore, it is necessary to investigate the impact of recall time on collaborative memory.

This study intends to investigate the effect of collaboration on CIE. To explore this phenomenon, not only because it is common in daily life, but also because it is conducive to answering two important theoretical questions: (1) Does collaboration have an impact on CIE? Is this influence collaborative or facilitate? (2) What are the potential mechanisms by which collaboration affects CIE? This study intends to answer the above questions through three experiments. Experiment 1 provides two collaboration conditions, the nominal group and the collaborative group, to explore whether and how collaboration affects an individual’s CIE. Experiment 2 specifically investigated the effect of different collaboration styles and recall time on CIE. On the basis of experiment 2, experiment 3 explores the reasons for the different effects of different collaboration methods on CIE.

According to knowledge revision theory, the strength of memory encoding influences CIE, with the encoding activation of corrected information competing with the encoding activation of original misinformation for the final output (Newman et al., 2022). Numerous studies have shown that collaboration enhances individuals’ memory, as demonstrated by post-collaboration memory facilitation (Congleton and Rajaram, 2011; Wissman, 2020). When misinformation is corrected, collaboration may strengthen an individual’s memory encoding of the corrected information, thus enhancing memory and inference, and inhibiting CIE (Sanderson et al., 2022). Therefore, Hypothesis 1 of the present study is proposed: Collaboration inhibits CIE. Previous studies have divided cooperation into free cooperation and turn-taking cooperation, and different cooperation conditions have different effects on memory (Weihai et al., 2015). Moreover, previous studies did not consider the recall time, a variable that affects the intensity of memory coding. The influence of different cooperation conditions on CIE may be regulated by recall time. Therefore, Hypothesis 2 of the present study proposes that free collaboration and turn-taking collaboration inhibit CIE differently at different recall time. The variation in the degree of inhibition of CIE by collaboration conditions under different recall time may be attributed to differences in re-exposure, cross-cuing, and information forgetting during collaboration. If group members refer to the corrected information and its encoding strength increases, this leads to a greater acceptance of the correction by the individual and ultimately inhibits CIE. In addition, if group members are influenced by cross-cuing and refer to the corrected information, they can enhance the individual’s fluency in processing the corrected information and use it as a valid cue for extracting the information. This enhances the memory representation of the corrected information and leads to CIE inhibitory. Forgetting information, on the other hand, reduces the extent of memory encoding and facilitates CIE. Therefore, Hypothesis 3 of this study proposes that the various collaboration conditions exert varying degrees of influence on CIE at different time due to re-exposure, cross-cuing, and forgetting during the collaboration process. To test these hypotheses, the current study investigated the impact of collaboration on the continued influence effect (CIE) through a series of three experiments.

The experimental design was a 2 × 2 mixed design, with one between-participants factor and one within-participants factor. The between-participants factor was collaboration condition, which included two levels: nominal group and free collaboration group. The within-participants factor was correction condition, with two levels: corrected and uncorrected.

An a priori power analysis in G*Power (Faul et al., 2007) suggested that we would need at least 36 participants to detect a high effect using ANOVA (f = 0.25; power = 0.95; α = 0.05). We therefore recruited 52 participants from an academic institution. As pre-registered, we removed those who failed the attention check (n = 4). This left a final sample of 48 participants (29 female, 19 males; Mage = 20.33, SDage = 1.47). All experiments conducted in this study were approved by the Ethics Committee of University of Jinan.

This Experiment combines the collaborative memory paradigm with the CIE paradigm. Participants were randomly assigned to either collaborative or nominal groups of three.

The data that support the findings of this study are available on this link: https://osf.io/cxf8v/?view_only=e53405491228414bad47c55542b92112

The statistical software SPSS 22.0 was utilized to perform descriptive statistics and conduct repeated measures analysis of variance (ANOVA) (Figure 1). All analytical procedures were based on a priori research hypotheses, and statistical significance was determined using a threshold of p < 0.05. All findings that reached statistical significance were duly reported. In Table 1, we compare the CIE scores and standard deviations for the free and nominal groups in Experiment 1 in the corrected condition versus the uncorrected condition, with the CIE scores being the General Memory Score, the Critical Memory Score, the Corrected Memory Score, and the Inference Scores, respectively. Separate repeated measures ANOVAs were conducted for each of the CIE scores, and the results were as follows:

The results of Experiment 1 found that free collaboration inhibited CIE. The free collaboration group remembered the corrected information significantly better than the nominal group, the inference of the free collaboration group relied more on later correction information rather than on the original misinformation.

Although experiment 1 supported the hypothesis of this study to a certain extent, the collaboration group in experiment 1 was limited to free collaboration, and it did not specifically explore whether different collaboration condition would have different impacts on CIE, and whether other factors in collaboration would have different impacts on CIE. Based on the above analysis, experiment 2, based on previous studies, divided collaboration modes into free collaboration and turn-taking group, and introduced the variable of recall time. According to the average recall time of subjects in experiment 1 was 12 min, the long recall time was set to 12 min, and the short recall time was set to 6 min, so as to explore whether different collaboration condition under different times would have an impact on CIE.

The Experiment design was a 2 (correction condition: corrected, uncorrected) × 2 (recall time: 6 min, 12 min) × 3 (collaboration condition: free group, turn-taking group, nominal group) mixed experimental design. The collaboration condition and recall time were between-participants factors, while the correction condition was a within-participants factor.

An a priori power analysis in G*Power (Faul et al., 2007) suggested that we would need at least 54 participants to detect a high effect using ANOVA (f = 0.25; power = 0.95; α = 0.05). We therefore recruited 120 participants from an academic institution. As pre-registered, we removed those who failed the attention check (n = 4). This left a final sample of 116 participants (64 female, 52 males; Mage = 20.16, SDage = 1.58).

The Experiment materials were the same as those used in Experiment 1.

The study procedure was approximately the same as in Experiment 1, with the following differences (Figures 2, 3). Participants were randomly assigned to one of six study groups, with a recall time of 12 min for the group with the longer recall time and 6 min for the group with the shorter recall time. Each of the three members of the group took turn reporting one piece of information at a time. The information was both spoken and written on a piece of paper. If a participant could not recall the information within 30 s, they would move on to the next participant, and so on in rotation.

The statistical software SPSS 22.0 was utilized to perform descriptive statistics and conduct repeated measures analysis of variance (ANOVA) (Figure 4). All analytical procedures were based on a priori research hypotheses, and statistical significance was determined using a threshold of p < 0.05. All findings that reached statistical significance were duly reported. In Tables 2, 3, we compare the CIE scores and standard deviations of the free, turn-taking, and nominal groups in the corrected versus uncorrected condition for the conditions in Experiment 2 where the recall time was 6 and 12 min. Separate repeated-measures analyses of variance (ANOVAs) were conducted for each of the CIE scores, and the significant results were as follows:

At a recall interval of 6 min, the turn-taking group relied more on the corrected information for their inferences compared to both the free collaboration and nominal groups. No significant difference was observed between the free collaboration and nominal groups. At a recall interval of 12 min, both the turn-taking and free collaboration groups demonstrated greater reliance on the corrected information compared to the nominal group. However, no significant distinction was found between the turn-taking and free collaboration groups. These results indicate that the turn-taking group’s ability to inhibit the continued influence effect (CIE) was consistently effective, whereas the free collaboration group’s ability to inhibit CIE was influenced by recall time and lacked consistency.

The experimental design was a one-factor between-subjects design, with co-operation conditions divided into turn-taking and free collaboration groups. The dependent variables were re-exposure, cross-cuing, and forgetting scores were measured.

An a priori power analysis in G*Power (Faul et al., 2007) suggested that we would need at least 34 participants to detect a high effect using ANOVA (f = 0.25; power = 0.95; α = 0.05). We therefore recruited 42 participants from an academic institution. As pre-registered, we removed those who failed the attention check (n = 2). This left a final sample of 40 participants (22 female, 18 males; Mage = 21.04, SDage = 1.28).

The statistical software SPSS 22.0 was utilized to perform descriptive statistics and Independent t-test. All analytical procedures were based on a priori research hypotheses, and statistical significance was determined using a threshold of p < 0.05. All findings that reached statistical significance were duly reported. In Table 4, we compare the memory and inference scores of the free and turn-taking groups with their standard deviations for the Memory Scores, which are the Re-exposure Score, the Cross-cuing Score, and the Forgetting Score. Separate independent samples t-tests were conducted for each dependent variable, and the results were as follows:

The memory of the rotating group was affected by both the representation of others and the cross-cue, so the cooperative group showed stronger memory promotion and reasoning enhancement after collaboration. The memory of the free group was affected by the representation of others, and the content that was recalled separately for the first time was forgotten more, and the encoding intensity and familiarity of the corrected information were increased less, so the inhibition effect on CIE was poor.

The Experiment results provide empirical support for the Knowledge Revision Theory. The results of this study found that collaboration led to an increase in the intensity of individuals’ encoding of corrective information. However, the intensity of encoding of original misinformation was not significantly different from that of the nominal group. This suggests that in the collaborative condition, corrective information gained the final output in the competition for activation with original misinformation. It also indicates that corrective information could be activated, effectively inhibiting the continued influence effect of misinformation. The reason why corrective information was able to dominate the entire information network under the collaboration condition was due to the re-exposure and cross-cuing that occurred during the collaboration process. Cross-cuing elicits additional or repeated recall, while re-exposure facilitates repeated learning. This process provides individuals with effective cues for recalling information, thereby enhancing the strength of memory encoding (Basden et al., 2000). Collaboration allows group members to reinforce corrections through re-exposure and cross-cuing. As individuals become more familiar with the corrected information, their encoding strength improves, leading to better memory retention of the corrected information and increased influence of the corrected information on inference-making.

The findings of the present Experiment demonstrate that collaborative conditions influence CIE, turn-taking collaboration stably inhibits CIE at different recall time, and the inhibitory effect of free collaboration on CIE is affected by recall time; when the recall time is longer, free collaboration inhibits CIE. However, when the recall time is shorter, free collaboration does not inhibit CIE. The effect of different collaboration conditions on the continued influence effect of misinformation is moderated by recall time, with the intrinsic mechanisms being the re-exposure, cross-cueing, and forgetting during the collaboration process. Specifically, the group engaged in turn-taking collaboration shows that their memories were influenced by re-exposure and cross-cuing, whereas the group engaged in free collaboration showed that their memories were influenced only by re-exposure.

According to Knowledge Revision Theory, the magnitude of CIE partly depends on the strength of memory, which is primarily influenced by the familiarity of the information and the frequency of its repetition. A previous study has demonstrated that cross-cuing elicits additional or repeated recall, while re-exposure elicits repeated learning. It has also been found that repeated recall produces more long-term benefits for subsequent individual recall than repeated learning (Eghbaria-Ghanamah et al., 2021), so turn-taking collaboration stably inhibit CIE across recall time. Additionally, the present study found that free collaboration was primarily influenced by re-exposure. Re-exposure counteracted the collaborative recall interruptions caused by retrieval strategies (Blumen et al., 2014). This means that members of a free collaboration can overcome past retrieval strategy interruptions after the collaboration process, and their memory potential is restored to a normal level (Marion and Thorley, 2016; Nie et al., 2023). Furthermore, when collaborative recollections are prolonged, the reappearance of others has a longer-lasting impact on memory, offsetting more costs. The potential benefits of retrieval strategy disruption include the ability to overcome past disruptions, restore memory to normal levels, and more repetitions of the corrected information result in subjects remembering the corrected information more intensely, thus free collaboration inhibit CIE. When collaborative recall was short, others reproduced for a shorter period, resulting in less cost incurred to offset retrieval strategy interruptions. Additionally, the memory potential had not been fully restored, and the number of repetitions of the corrected information is less, and free collaboration is not able to inhibit CIE.

It was found that the alternating group inhibited CIE stably, while the free group was affected by recall time. When the recall time is 6 min, the reasoning of the rotating cooperative group is more dependent on the later corrected information than that of the free cooperative group and the nominal cooperative group. There is no significant difference between the nominal cooperative group and the free cooperative group. However, when the recall time is 12 min, the reasoning of the rotating cooperative group and the free cooperative group is more dependent on the later corrected information than that of the nominal group. There was no significant difference between the rotating collaboration group and the free collaboration group, which indicated that the phenomenon of rotating collaboration inhibiting the continuous influence of false information was stable, but the free collaboration was not stable, and only when individuals fully cooperated with others and shared memories, the free collaboration could inhibit CIE. The reasons for the different results of free collaboration and alternate collaboration under different recall time conditions are as follows: based on the previous theories of others’ reproduction and cross-cues, and the discovery that collaboration can destroy individual memory strategies to a certain extent and lead to individual forgetting, experiment 3 of this study was carried out.

This study has several limitations. First, while exploring the effect of collaboration on the continued influence effect (CIE), individual differences among participants can lead to variability in the collaboration process. Controlling for all potential variables in the collaborative setting is challenging, and these uncontrolled factors may influence the results. Additionally, the study’s duration was quite lengthy, which may have led to participant fatigue. This fatigue could have impacted the outcomes, particularly during the final memory test session. Consequently, participants might have provided fewer responses to the free recall questions in the final test due to decreased engagement.

In the future, we can further investigate the impact of collaboration on CIE. Specifically, we can examine how collaboration affects CIE under various influencing factors. Previous studies have found that collaboration on memory is influenced by various factors such as group sizes, intervals, the nature of reading materials, and individual differences in participants. Future studies can further investigate the impact of collaboration on collective intelligence enhancement, considering these aforementioned factors. The research design of this study combines the collaborative memory paradigm with the CIE paradigm. In future studies, researchers can further innovate the study design, streamline the procedures, minimize the impact of participant fatigue and guessing on the results, and effectively manipulate the collaborative variables to investigate their effects on CIE. In addition, future research can further explore the neural mechanisms through which collaboration affects CIE. In recent years, several studies have started to investigate the neural mechanisms behind the persistent effects of misinformation (Hua et al., 2022). One study found that the activation of the left middle temporal gyrus was significantly weaker in the correction condition during the encoding phase compared to the control condition. Additionally, the activation of the frontal gyrus and the anterior cingulate gyrus was weaker in the correction condition during the extraction phase. Furthermore, the functional connectivity between the frontal gyrus and the precentral gyrus was stronger. However, it remains unknown whether collaboration influences the activated brain regions of the CIE. However, it is not yet known whether collaboration affects the activation of brain regions in CIE. Future studies may explore the neural mechanisms underlying the effects of collaboration on CIE.