In Sync: Memory Similarity and Recall Order for the 2017 Eurovision Song Contest

Abstract

When recalling a widely shared public event, individuals tend to retrieve similar details in their narratives. However, the extent to which memory representations converge across individuals and whether recollections follow a systematic order remain unclear. In this study, we examined the extent to which memory representations of Portugal’s victory in the 2017 Eurovision Song Contest are shared across individuals, whether recollections follow systematic order patterns, and which determinants are associated with event memory accuracy and inter-subject memory similarity. Participants (N = 58) reported their memories of the event approximately 6.5 years after it occurred. Representational similarity analyses showed substantial overlap in recalled details across participants, indicating convergence in collective memory. Regression analyses showed that greater media exposure was associated with higher event memory accuracy, but we found no clear evidence that the reported memory determinants were associated with similarity across participants’ narratives. Analyses of recall order revealed a consistent sequential pattern, with the song title immediately following the singer’s name. The observed order may reflect the frequent co-presentation of these details in media coverage. Together, these findings suggest that collective memories for public events are characterised by some shared content, but also by an organised recall pattern that reflects shared cultural knowledge and associative memory processes.

1. Introduction

On the evening of 13 May 2017, millions of viewers across Europe and beyond witnessed Portugal win the Eurovision Song Contest, which took place in Kyiv, Ukraine. This was Portugal’s first victory in 53 years of participation. The song “Amar pelos dois”, performed by Salvador Sobral and written by his sister Luísa Sobral, became iconic, not only for its minimalist staging and emotional delivery but also for the collective excitement it sparked across the Portuguese public. For many, the moment was deeply emotional, widely celebrated, and extensively covered in national media. Events like this, which combine emotional salience, media attention, and national pride, offer a unique opportunity to study how people remember collectively. To what extent individual memories of a public event converge into a shared narrative? Which factors influence memory similarity? Is there a consistent temporal order in which the event’s details are retrieved across participants?

Collective memory refers to socially constructed representations of past events shared within a community that contribute to members’ social identity (Hirst & Fineberg, 2012; Hirst & Manier, 2008). Originating in the work of Halbwachs (1992), the study of collective memory emphasizes that memories are not merely individual mental records, but are also shaped, maintained, and transmitted through social interactions.

Over the last decades, researchers have applied this lens to numerous public events, from the Challenger space shuttle disaster (Neisser & Harsch, 1992) to terrorist attacks (i.e., 9/11; Hirst et al., 2009, 2015) to wars (Luminet & Spijkerman, 2017) to major sporting events (Bavassi et al., 2020; Merck et al., 2020), showing that certain events become landmarks of collective memory, deeply embedded in cultural narratives. These collective memories are reinforced by media coverage, education, conversations, and commemorative practices, all of which contribute to their persistence and homogeneity within a community (Halbwachs, 1992; Olick, 2007; Roediger & Abel, 2015; Wertsch, 2002).

From this perspective, what is collective about memory lies in the patterns of overlap across individuals’ representations of the same event. Another important concept, mnemonic convergence, refers to the process by which individual memories of the event become more similar over time through conversations, shared media exposure, and repeated discussion (Coman et al., 2016). Research has shown that for highly consequential public events, such as the 9/11 attacks, people from different countries tend to emphasize similar central aspects of the event, such as the attacks on the World Trade Center and the exceptional magnitude of the attacks, information that was widely covered by the media and discussed across countries (Luminet & Curci, 2009). In line with this, Hirst et al. (2015) found that, through repeated discussions and media exposure, people’s accounts of 9/11 converged over time. As such, collective memory can manifest as shared or converging memories across individuals, sometimes referred to as inter-subject memory similarity. This concept quantifies the degree of overlap between individuals’ recalled details of a given event (in the present study, inter-subject memory similarity will be used as an operational measure of shared memory content).

A related phenomenon in the realm of memory for public events is flashbulb memory, a type of autobiographical memory characterized by remarkably vivid and highly confident recollection of the personal circumstances in which one has learned about a significant or surprising event (R. Brown & Kulik, 1977; Hirst & Phelps, 2016). Flashbulb memory typically includes personal contextual details, such as with whom or where one was when first learned about the event. Thus, while flashbulb memories capture personal circumstances in which people knew about the public event, convergence in memory is observed primarily at the level of factual event representations associated with flashbulb-inducing events.

Despite extensive literature on collective memory, methodological approaches have often been limited to simple measures, such as computing the proportion of participants who recall specific factual details of an event (Hirst & Echterhoff, 2012). While this approach is valuable, it treats memory as a static collection of facts rather than a dynamic, socially embedded phenomenon, overlooking how recollections become aligned across individuals and how memories are shared at the group level (Hirst & Echterhoff, 2012). Assessing the degree of similarity between individuals’ recollections provides a theoretically grounded way to quantify memory convergence at the group level. Recent work by Cheriet et al. (2021) introduced a method to assess shared event memory for public events. In their study, young and older adults were asked to freely recall a widely shared public event (i.e., the deadly collapse of a bridge in Italy), and the content of these narratives was coded into discrete memory details. Similarity in the recalled narratives was quantified by comparing the overlap in details between all possible pairs of participants. This perspective allows for a better understanding of collective remembering, capturing not only the prevalence of the most common recalled details but also associative patterns in recall (i.e., which elements tend to co-occur across individuals’ memories).

While this method has improved our understanding of content similarity, little attention has been paid to how these shared event memories are organized during recall, particularly with respect to the sequential or narrative unfolding of the remembered details. This leaves a gap in the literature concerning the narrative organization of collective memories for public events—whether people reconstruct shared public events using similar sequences or story-like frameworks (i.e., the order in which details are recalled).

Studies on autobiographical memory have shown that individuals tend to organize life events according to cultural life scripts, that is, important life events that are culturally shared (e.g., graduating, getting married). Such cultural life scripts shape personal memories and narratives, suggesting that recall of autobiographical events is structured around salient landmarks (Rubin & Berntsen, 2003; Thomsen & Berntsen, 2008). In addition to structuring retrieval, life scripts also help to maintain autobiographical memories by repeatedly structuring retrieval over time (Rubin & Berntsen, 2003). Previous research on narrative organization also offers important insights into how memories are structured and retrieved in sequence. It has been proposed that more important or salient events function as effective retrieval cues, such that recalling these central events increases the likelihood of activating other causally or thematically related events from the same event cluster (N. R. Brown, 2005). In this way, certain events play an organizing role in memory retrieval. Relatedly, Sap et al. (2022) proposed a computational measure of narrative sequentiality, defined as the predictability of sentence order within a story according to a language model. They demonstrate that retold autobiographical narratives several months later have higher sequentiality than fresh recollections, suggesting that, with time, personal narratives rely more on the preceding sentence/context. This finding aligns with the observation that, over time and with memory consolidation, retelling past events relies less on recall from episodic memory and increasingly draws on common semantic knowledge (e.g., Bartlett, 1932).

Extending these insights to the domain of collective memory raises the question of whether shared public events are remembered not only with overlapping content but also with similar narrative sequences across individuals. Prior research suggests that both semantic schemas and scripts, as well as social and cultural influences—such as shared narrative conventions, media exposure, and culturally learned storytelling norms—shape how memories are organized and communicated (N. R. Brown, 2005; Cohn et al., 2014; Fivush & Nelson, 2004; Kaidesoja, 2024). These studies suggest that salient elements of an event frequently serve as anchors or cues in recall, and that shared cultural knowledge can lead to convergence in narrative structure across individuals. Building on these findings, our study aims to explore whether collective memory exhibits similar patterns of structured recall and inter-subject convergence.

For that, we investigated collective memory for Portugal’s 2017 Eurovision victory by examining the extent to which individuals share memory representations and the order in which event details are recalled. We measured inter-subject memory similarity and described the order in which participants recalled details to assess if people’s memories of this significant public event converge in content and in narrative structure. In addition to examining event memory and inter-subject similarity, we descriptively characterized flashbulb memories concerning the personal circumstances in which participants learned about the victory. We also inspected potential determinants of memory, including emotional intensity, media exposure, social discussion, and perceived importance, as these factors have been shown to influence both recall accuracy and convergence of memories across individuals (Coman et al., 2009; Hirst & Phelps, 2016; Luminet & Curci, 2009; Talarico & Rubin, 2003). We expect participants to recall common details due to the national and social importance of the event, its extensive media coverage, and frequent social discussions, factors that promote convergence in collective memory (Coman et al., 2009; Luminet & Curci, 2009; Roediger & Abel, 2015). In the context of the Eurovision Song Contest, the performer and the song title refer to directly related aspects of the winning performance, making them a natural pair for examining whether participants report details in a consistent order. Finally, we anticipate that emotional intensity, importance attributed to the event, frequency of media exposure, and frequency of social discussion will be positively associated with event memory accuracy and inter-subject memory similarity. This pattern would support the view that repeated social rehearsal and the cultural relevance of public events contribute to the alignment of individual memory representations within a community (Hirst et al., 2015; Kensinger, 2009; Talarico & Rubin, 2003). Accordingly, this study addressed three questions: how much overlap was present in the event details participants recalled; what sequences of details recurred in their narratives; and the extent to which the reported memory determinants were associated with event-memory accuracy and inter-subject memory similarity.

2. Methods

2.1. Participants

Sixty-eight university students participated in the study. Of these, 8 participants were excluded from the analysis because they did not know the event or failed to answer entire sections of the survey. Two additional participants were excluded from all analyses because they had incomplete responses to the event-memory questions (response rate lower than 70%). The final pool consisted of 58 participants aged 18 to 30 (M = 21.91, SD = 3.79), 48 of whom were female. At the time of the event, participants’ ages ranged from 11 to 23 years (M = 15.50, SD = 3.80). Most participants (67.2%, n = 39) were aged 11–16 years, while the remaining 32.8% (n = 19) were aged 18–23 years. Regarding education, 61.7% had completed high school, 21.7% held an undergraduate degree, and 16.7% had a master’s degree. Participants were tested approximately 6.5 years after the 2017 Eurovision Song Contest. While this represents a relatively long retention interval compared to typical laboratory memory studies, such time frames are common in research on collective and autobiographical memory of public events (see Berntsen & Thomsen, 2005; Hirst & Manier, 2008; Hirst et al., 2009). This interval allows for the examination of how memories of a widely shared public event are represented and organized at a long-term delay. All participants received course credit for their participation and were informed that the questionnaire was anonymous and that the data would be used solely for research purposes. The local Ethics Committee approved the study.

2.2. Procedure

Participants read the informed consent form and were briefed on confidentiality and the study’s main goal. They then proceeded to the different sections of the questionnaire, which was presented on Qualtrics (Qualtrics, 2026). Survey completion took, on average, 8.42 min (SD = 3.61).

First, participants were asked to provide demographic information, including their age, gender, nationality, city of residence, and educational degree. Additionally, they assessed their perceived ability to recall details of past events using a 5-point scale, ranging from 1 (below average) to 5 (above average).

In the second section, participants indicated if they were familiar with Portugal’s victory at the Eurovision Song Contest (no further details about the event were provided). If participants answered “yes”, the questionnaire continued, otherwise, it ended. Those familiar with the event were asked to recall it and write a description of what they remember, with as many factual details as possible, without word or time limits. We did not specify any time range or specific features of the event. As such, participants could focus on the moment of the victory, the performance details, the aftermath, or more general features of the occasion.

The third section of the survey assessed memory for the event using ten factual questions that covered a range of topics related to Portugal’s 2017 Eurovision victory. The questions included key details of the event, such as the year of the victory, the performer’s name, the title of the song, the host city, Portugal’s total score, and the runner-up. Other questions included peripheral or contextual details, such as the running order of Portugal’s performance, the colour of the singer’s clothing, who joined him on stage during the reprise, and the health issue that delayed his arrival at the contest. All questions referred strictly to the 2017 final itself and did not target broader Eurovision knowledge or events extending beyond the immediate context of that year’s competition. After each question, participants rated their confidence in their response on a 7-point scale (1 = not at all confident to 7 = extremely confident).

Next, a set of 6 questions assessed participants’ personal memories of the event. This section was presented only to participants who were at least 18 years old at the time of the event (n = 19). This criterion was intended to reduce potential developmental variation in the retrieval of the personal context in which participants learned about the event, as context and associative memory (involved in the encoding and retrieval of flashbulb memory details) can continue to develop during adolescence. The questions focused on the canonical characteristics described in the flashbulb memory literature (see R. Brown & Kulik, 1977): where they were, how they learned about the victory, what they were doing, how they felt, how others around them reacted, and what they did immediately after. For each answer, participants rated their confidence in their memories on a 7-point scale.

Lastly, we evaluated potential memory determinants. Using 7-point scales, participants indicated how frequently they followed news about the Eurovision Song Contest in the media (1—not at all frequent; 7—very frequent), how often they talked to others about the Eurovision Song Contest (1—not at all frequent; 7—very frequent), how surprised they felt when they learned about Portugal’s victory (1—not at all surprised; 7—extremely surprised), how well they could visualize that moment (1—not well; 7—extremely well), how strong is their physiological reaction when thinking about the moment they learned about the event (1-I don’t feel any reaction; 7-I feel a strong reaction), how intense their emotions were when they learned about the victory (1—not at all intense; 7—extremely intense), and how important this event was for them and the international community (1—not at all important; 7—extremely important) and the level of emotion felt (i.e., valence; 1—extremely negative; 7—extremely positive).

2.3. Measures and Coding

2.3.1. Analyses of Flashbulb Memory, Event Memory, and Their Determinants

Responses to the six flashbulb memory questions (e.g., Where were you when you learned about the event?) were scored 2 if the participant responded by providing details (e.g., “in my bedroom”), 0 if she/he did not respond, and 1 for intermediate cases (e.g., “I was outside”). The proportion of detailed responses was computed as the sum of scores divided by the maximum possible score (for each of the 6 questions, a maximum of 2 points was given, resulting in a maximum of 12 points). Additionally, the average confidence score for the given answers in flashbulb memory was used as each participant’s confidence value. The coding (0, 1, or 2) of each answer was computed by two independent judges—the first author (A.R.) and a research assistant (S.M.). The inter-rater reliability analysis for the coding, computed on the judges’ independent scores, indicates excellent reliability, with a Cohen’s κ of 0.91 (95% CI [0.76, 1.00]).

Responses to the 10 questions used to assess event memory (i.e., recall of factual details) were coded 1 if correct and 0 if incorrect or if no response was provided. Event memory accuracy was thus measured by the proportion of correct responses to the 10 questions (maximum of 10 points). The average confidence score for correct answers was taken as the confidence level for each participant. The two independent judges mentioned above assigned a coding value (0 or 1) to each answer. Cohen’s κ was calculated from their independent codes before any discrepancies were discussed (κ = 0.96, 95% CI [0.90, 1.00]). Then, the two judges’ coding was compared, and the differing values were discussed until agreement was reached.

The determinants of memory were assessed using self-report Likert-type scales and were treated as continuous variables in the analyses. They resulted in six constructs: (a) media exposure (how frequently participants followed Eurovision-related news); (b) rehearsal (how often they talked to others about the event); (c) surprise (how unexpected the victory felt); (d) emotional intensity (how strong was the emotional reaction); (e) vividness (composite index computed as the mean of visualization of the moment and physiological reaction); and (f) importance (composite index computed as the mean of personal importance and international importance).

2.3.2. Narrative Analysis

Participants were first invited to freely describe what they remembered about Portugal’s 2017 Eurovision victory. Of the 58 participants in the final sample, 54 provided a written description of the event. The other four participants completed the remaining parts of the survey but did not provide a written description and were therefore excluded from the narrative analyses. Participants were asked to focus on factual information of the event, but the instructions did not specify any particular time frames, prompts, or features of the event, allowing participants to report whatever came to mind spontaneously. As a result, the narratives varied in length and content, encompassing a range of information types. Some participants focused on key details, such as identifying Salvador Sobral as the winner, recalling the performance of “Amar pelos dois”, or mentioning the host city and the announcement of the final score. Others included peripheral or contextual information, such as the singer’s clothing, his health problems, audience reactions, or general impressions of the broadcast. Narratives also differed in specificity: while some reported precise details (e.g., who joined him on stage during the reprise), others described broader reactions or interpretations of the event (e.g., that the victory was emotional or unexpected).

To analyse the content of these narratives in a systematic way, we constructed a template narrative based on information gathered from social media and television news prior to data collection. This procedure follows Cheriet et al. (2021), is conceptually grounded in Wertsch’s (2002) notion of schematic narrative templates, and builds on earlier memory research that evaluates individual recollections against publicly established accounts of an event (Neisser & Harsch, 1992). This template included 9 key elements of the event, including the date (1), the name of the singer (2), the title of the song (3), the authorship of the song (4), the location of the contest (5), important reactions and events following the victory (6), characteristics of the singer (7), the singer’s health condition (8), and visual aspects of the performance (9). These elements were selected as key details to be coded; they were not defined a priori as more or less central to the event, and centrality was not measured in this study. This template served as a reference for coding the presence or absence of each detail in participants’ recollections.

The details mentioned in each participant’s memory description were compared to the key details outlined in the narrative template. Specifically, we counted the presence or absence of each detail, allowing us to compute the total number of details recalled by each participant. If the participant mentioned a detail, a score of 1 was given; if a detail was not mentioned, a score of 0 was given. To assess inter-subjects’ memory similarity, we used the same procedure as Cheriet et al. (2021), in which the details recalled by a participant are compared to those recalled by every other participant (i.e., participant A is compared to participant B, to participant C, and every other participant until the last participant). For each pair of participants, we computed the number of common details they recalled. Then, this number was divided by the total number of details mentioned at least once by either participant. For example, if participant A recalls details {1, 2, 5} and participant B recalls {1, 2, 4}, they share two details {1, 2} out of four {1, 2, 4, 5} total distinct details, yielding a similarity score of 0.50. This safeguarded that the similarity value between two participants was independent of the absolute number of memory details recalled by the two participants (i.e., two participants would show higher similarity when a large proportion of the details they each recalled overlap). Conversely, even if two participants recalled many details overall, their similarity would be low if only a small subset of those details were the same. This procedure was conducted for every pair of participants. The similarity value for each participant was the average of all comparisons. The inter-rater reliability, based on a 50% sample of the narratives, demonstrated considerable agreement for category presence (Cohen’s K = 0.956) and for details sequence (similarity index = 0.994).

To assess whether the observed mean pairwise similarity exceeded the level expected under random selection of details, we used a Monte Carlo randomization test with an equal-probability model. In each of 10,000 simulated datasets, we kept the number of details recalled by each participant unchanged and randomly selected that same number of distinct details from the nine template details, giving each detail an equal probability of selection. We recalculated the mean pairwise Jaccard similarity for each dataset using the same procedure as for the observed data. The one-tailed p-value was calculated as the proportion of simulated values equal to or greater than the observed value, using the plus-one correction.

For the recall-order analysis, we also recorded the sequence in which details appeared in each narrative. We tested whether the song title (detail 3) immediately followed the singer’s name (detail 2) more often than expected by chance using a within-participant permutation test of recall order. The order of the details within each participant was randomly shuffled 10,000 times, preserving the details each participant recalled, and the number of narratives with detail 3 immediately after detail 2 was recomputed in each permutation. The one-tailed p-value was computed as (b + 1)/(B + 1), where b is the number of permutations with a count at least as large as the observed count and B = 10,000. We ran the same test for the reverse sequence and for detail 2 appearing anywhere before detail 3. This analysis describes the order of details but does not explain why they appeared in that order.

Statistical analyses were conducted using IBM SPSS Statistics (Version 30, IBM Corp, 2023), except for the inter-subject memory similarity analysis and recall-order analyses, which were computed in R (Version 4.3.0; R Core Team, 2023). Recall frequencies and flashbulb-memory characteristics were summarized descriptively. We used a randomization test to assess inter-subject similarity and a permutation test to assess whether the focal recall-order pattern occurred more often than expected under random ordering. Regression analyses were used to examine associations between memory determinants and event-memory accuracy and inter-subject memory similarity.

3. Results

3.1. Flashbulb Memory Characteristics

Only participants who were 18 years or older at the time of the event responded to the flashbulb memory questions (n = 19). On average, these participants answered 85% (SD = 23%) of the questions, indicating that their personal recollections of Portugal’s victory at the Eurovision Song Contest were notably detailed. Almost all participants (95%) were able to report their location, how they learned about the event (95%), their own feelings (89%), and the feelings of others (95%). More than half (58%) responded to all flashbulb memory questions. Overall, participants reported moderately vivid memories of the moment they learned about Portugal’s victory (M = 3.66; SD = 1.61) and reported high confidence in the evoked memories (M = 6.53, SD = 1.07), even 6.5 years after the event (see Table 1).

Table 1. Descriptive Statistics of the Memory Measures.

Variable	M	SD
Flashbulb memory
Proportion of responses provided (0–1)	0.85	0.23
Confidence (1–7)	6.53	1.07
Event memory
Proportion of correct responses (0–1)	0.48	0.18
Confidence (1–7)	5.48	1.13
Inter–subject memory similarity
Proportion of similarity (0–1)	0.42	0.14

Note. Flashbulb memory, event memory and similarity represent proportions of recalled details and range from 0 to 1. Confidence ranged from 1 to 7 (not at all confident to extremely confident). n = 19 (flashbulb memory), N = 58 (event memory) and n = 54 (inter-subject memory similarity).

3.2. Event Memory Accuracy and Confidence

Regarding event memory, assessed through responses to the ten factual questions, participants (N = 58) correctly recalled 48% (SD = 18%) of the key details asked, with overall high confidence (M = 5.48, SD = 1.13). Several details were correctly remembered by more than 50% of the participants, and hence can be considered as part of collective memory (Cheriet et al., 2021; Merck et al., 2020). Most participants recalled the name of the singer (95%), the title of the song (86%), the person who joined the singer on stage after victory (71%), the colour of the singer’s clothes (71%) and the heart problem of the singer (69%).

3.3. Inter-Subject Memory Similarity

The analysis of inter-subject memory similarity—based on participants’ free-recall narratives (n = 54) and defined as the proportion of overlapping event details recalled by pairs of participants—showed that the observed mean pairwise similarity was 42% (SD = 14%). Under the equal-probability randomization model, the mean similarity was 19.91% (95% simulation interval [19.18%, 21.06%]). The observed similarity was significantly higher than expected under this model (p < 0.001).

Descriptive statistics for these event memory measures are presented in Table 1.

3.4. Determinants of Event Memory Accuracy and Similarity

We first conducted exploratory Pearson correlations to investigate whether age (at the time of the event) was not significantly associated with event memory accuracy, r(56) = 0.195, p = 0.143, or with inter-subject memory similarity, r(56) = 0.125, p = 0.350. Both correlations were positive but weak in magnitude (Cohen, 1988), indicating that age accounted for less than 4% and 2% of the variance in accuracy and similarity, respectively.

Concerning the determinants typically associated with memory for public events, participants reported following Eurovision-related news (i.e., media exposure; M = 4.16, SD = 1.97) and discussing the victory with others (i.e., rehearsal; M = 4.48, SD = 1.66) to a moderate extent. They experienced relatively high levels of surprise when learning about the outcome (M = 5.57, SD = 1.17) and of emotional intensity associated with the event (M = 4.72, SD = 1.76). Regarding the phenomenological characteristics of the memory, participants showed moderate overall vividness (M = 3.66, SD = 1.61, both in terms of visualization of the moment (M = 3.93, SD = 1.87) and physiological reaction (M = 3.38, SD = 1.80). Regarding importance, participants attributed moderate personal importance to the event (M = 3.88, SD = 1.81), but considered it highly important at the international level (M = 5.60, SD = 1.40), yielding an overall importance score of M = 4.74 (SD = 1.24). Finally, valence ratings indicated that the memory was experienced as highly positive (M = 6.03, SD = 1.09). Descriptive statistics for these determinants are presented in Table 2.

Table 2. Descriptive Statistics of Potential Determinants of Event Memory.

Variable	M	SD
Media exposure (1–7)	4.16	1.97
Rehearsal (1–7)	4.48	1.66
Surprise (1–7)	5.57	1.17
Emotional intensity (1–7)	4.72	1.76
Visualization of the moment (1–7)	3.93	1.87
Physiological reaction (1–7)	3.38	1.80
Vividness (1–7)	3.66	1.61
Personal importance (1–7)	3.88	1.81
International importance (1–7)	5.60	1.40
Importance (1–7)	4.74	1.24
Valence—negative to positive (1–7)	6.03	1.09

Note. all variables ranged from 1 to 7.

To explore the extent to which these factors predicted (a) event memory accuracy and (b) inter-subject memory similarity, two multiple linear regressions were conducted, each including as predictors: media exposure, rehearsal, surprise, emotional intensity, vividness, and importance (see Table 3 and Table 4). Before interpreting the regression coefficients, we examined multicollinearity diagnostics. Tolerance values were all above 0.20, and VIF values remained below 5 across models, indicating acceptable levels of multicollinearity. Although the highest condition indices approached conventional thresholds (≈25–30), no predictor showed variance proportions concentrated in a single component, suggesting that collinearity did not meaningfully compromise the stability of the regression estimates.

Table 3. Multiple Linear Regression Predicting Event Memory Accuracy.

Predictor	B	SE	β	t	p
Media exposure	0.046	0.023	0.489	1.997	0.051
Rehearsal	0.009	0.024	0.078	0.369	0.714
Surprise	−0.021	0.021	−0.132	−0.987	0.328
Emotional intensity	−0.015	0.021	−0.144	−0.733	0.467
Vividness	0.039	0.021	0.344	1.885	0.065
Importance	−0.015	0.023	−0.102	−0.677	0.501

Note. R² = 0.39; adjusted R² = 0.31; F(6, 51) = 5.34, p < 0.001; N = 58.

Table 4. Multiple Linear Regression Predicting Inter-Subject Memory Similarity.

Predictor	B	SE	β	t	p
Media exposure	0.009	0.023	0.126	0.404	0.689
Rehearsal	0.014	0.024	0.164	0.589	0.560
Surprise	0.031	0.021	0.256	1.497	0.143
Emotional intensity	−0.066	0.022	−0.775	−3.049	0.004
Vividness	0.015	0.021	0.177	0.721	0.476
Importance	0.043	0.021	0.381	2.024	0.050

Note. R² = 0.25; adjusted R² = 0.13; F(6, 38) = 2.14, p = 0.071; n = 45.

To examine which variables predicted accurate event memory, we conducted a multiple regression analysis with event memory accuracy as the dependent variable. The model included the previously identified memory determinants as predictors. The overall regression model was significant, F(6, 51) = 5.34, p < 0.001, explaining 39% of the variance in event memory accuracy (adjusted R² = 0.31). Among the predictors, only the frequency with which participants followed news coverage of the event (i.e., media exposure) emerged as a marginally significant positive predictor (β = 0.49, p = 0.051), suggesting that greater exposure to media reports was associated with more accurate memory for the event’s details. No other factor significantly contributed to the model, with rehearsal, surprise, emotional intensity, vividness, and importance failing to predict event memory above and beyond their shared variance with media exposure.

Second, to examine whether the six memory determinants predicted the degree of inter-subject similarity, the same multiple regression analysis was conducted, with the proportion of shared details as the dependent variable. Because the model required complete data on inter-subject memory similarity and all six predictors, cases with missing data were excluded listwise. The model therefore included 45 participants.

The overall model predicting inter-subject memory similarity did not reach statistical significance, R² = 0.25, F(6, 38) = 2.14, p = 0.071. At the level of individual coefficients, emotional intensity had a statistically significant negative estimate (β = −0.78, t = −3.05, p = 0.004), whereas importance had a positive estimate at the 0.05 threshold (β = 0.38, t = 2.02, p = 0.050). The remaining predictors were not statistically significant (all with p > 0.14). These estimates are exploratory given that the overall model did not reach statistical significance.

3.5. Recall Order

Next, we examined the type of details most commonly reported in participants’ narratives and the order in which they were recalled. Among the 54 participants who recalled at least one detail, the most frequently mentioned details were the name of the singer (detail 2; 98%) and the song’s title (detail 3; 78%), and 41 participants (76%) mentioned both. Table 5 shows the most frequent combinations of details across narratives, regardless of order. Other frequent combinations included the singer’s name with the singer’s characteristics (detail 7; 29.6%), the singer’s name with the song’s authorship (detail 4; 29.6%), and the event date (detail 1) with the singer’s name (31.5%). The most frequent three-detail combinations were {1, 2, 3} and {2, 3, 4}, each mentioned by 27.8% of participants.

Table 5. Most Frequent Combinations of Details Across Narratives.

Details	Frequency	N
{2}	0.981	53
{3}	0.778	42
{2, 3}	0.759	41
{7}	0.315	17
{1}	0.315	17
{1, 2}	0.315	17
{4}	0.296	16
{2, 7}	0.296	16
{2, 4}	0.296	16
{3, 7}	0.278	15
{1, 3}	0.278	15
{3, 4}	0.278	15
{1, 2, 3}	0.278	15
{2, 3, 4}	0.278	15
{2, 3, 7}	0.259	14
{9}	0.204	11
{2, 9}	0.204	11
{3, 9}	0.167	9
{2, 3, 9}	0.167	9
{4, 7}	0.148	8
{1, 4}	0.148	8
{2, 4, 7}	0.148	8
{1, 3, 4}	0.148	8
{1, 2, 4}	0.148	8
{1, 2, 3, 4}	0.148	8

Note. Combinations are sets of details mentioned by the same participant; their order of recall is not represented. Frequencies are the proportions of participants (n = 54) who mentioned all details in the set.

Using each participant’s ordered sequence of details, the song title immediately followed the singer’s name in 24 of 54 narratives (44%; 59% of participants who mentioned both). Because both details were mentioned very frequently, we compared this with chance using a permutation test. Under random ordering, the expected count was 13.1 (95% interval 8–19); the observed count was significantly higher (p < 0.001). The reverse sequence occurred in 13 narratives, in line with chance (expected 13.2; p = 0.585). The singer’s name appeared somewhere before the title in 26 of the 41 narratives that included both, which did not differ statistically from chance (expected 20.4; p = 0.060). The observed and expected counts for these recall-order patterns are presented in Table 6.

Table 6. Observed and Expected Frequencies of Recall-Order Patterns for the Singer’s Name and Song Title.

Recall-Order Pattern	N	Expected Under Random Ordering, n [95% Interval]	p
Detail 2 immediately before detail 3	24	13.1 [8–19]	<0.001
Detail 3 immediately before detail 2	13	13.2 [8–19]	0.585
Detail 2 anywhere before detail 3	26	20.4 [14–27]	0.060

Note. n = 54 narratives; 41 included both details. Expected counts and 95% permutation intervals are based on 10,000 random shuffles of the order of details within each participant’s narrative, preserving the details recalled by each participant. One-tailed p-values were calculated as (b + 1)/(B + 1), where b is the number of permutations with a count at least as large as the observed count and B = 10,000.

4. Discussion

Our study aimed to investigate the extent to which individuals share similar memory representations of Portugal’s victory in the Eurovision Song Contest and to explore whether detail recollections follow a specific order pattern. By integrating content-based similarity with an examination of sequential recall patterns, our study contributes to the literature on collective memory by highlighting the extent to which memory representations are shared across individuals and if recall unfolds in structured ways.

The flashbulb memory findings provide a descriptive characterization of participants’ autobiographical memories for the circumstances in which they learned about the victory. Participants who were adults at the time of the event answered, on average, to 85% (SD = 23%) of the flashbulb memory questions, reporting high confidence (M = 6.53), and moderate vividness (M = 3.66) on their responses. This suggests that memory for the Eurovision Song Contest 2017 has characteristics of a flashbulb memory. Yet, these findings concern personal context and do not necessarily relate to event-memory accuracy, inter-subject similarity, or recall order.

Participants correctly recalled, on average, 48% of the information prompted by our event memory questions. To examine which determinants best accounted for individual differences in event memory accuracy, we conducted a regression analysis. Results revealed that media exposure showed the strongest association with event memory accuracy, although this effect only approached statistical significance (p = 0.051). This finding is consistent with previous work on public event memory, which has reported that greater exposure to information sources is associated with increased accuracy for factual details (Hirst et al., 2009; Tinti et al., 2014). Indeed, media exposure may reinforce the memory traces originally encoded, but it can also modify existing memory traces. Through reconsolidation, new information (acquired after the event) may be integrated in memory, allowing us to correct errors or incongruencies, and thus contributing to greater factual accuracy (Hirst et al., 2015; Hirst & Phelps, 2016).

Turning to the degree to which memory content was shared across participants, our findings reveal some level of convergence between participants’ recollections, i.e., a 42% overlap in the event-related details included in participants’ narratives. Although the overall model predicting inter-subject similarity did not reach statistical significance, the individual coefficient estimates may be of interest for future research. Increased emotional intensity associated with the event was related to lower similarity values, suggesting that strong personal reactions may amplify idiosyncratic, person-specific aspects of memory. In contrast, greater importance attributed to the event was associated with greater similarity across participants’ recollections, suggesting that socially meaningful public events promote alignment around shared narrative structures. These exploratory patterns should be interpreted cautiously and do not provide firm evidence that either factor predicts similarity. Of note, the regression predicting inter-subject similarity included 45 participants and six predictors, making the analysis underpowered. Its coefficient estimates may therefore be imprecise and should be interpreted cautiously and examined in larger samples.

Contrary to our hypothesis, we did not find an association between media exposure and inter-subject memory similarity. We should note that mnemonic convergence for public events often emerges primarily for a subset of key event details, rather than across the full range of remembered information, and can occur independently of individual differences in media exposure (Hirst et al., 2009, 2018). Similarly, conversational rehearsal may not uniformly increase memory similarity. When conversations selectively highlight particular aspects of an event, they can induce socially shared retrieval-induced forgetting, whereby unmentioned details are suppressed, potentially leading to divergence rather than convergence across individuals (Coman et al., 2009). In the case of the 2017 Eurovision victory, this was a highly publicized national event, for which a dominant narrative was firmly established. In such a context, increased media exposure may primarily reinforce an existing shared but restricted core rather than introducing additional common details that would increase similarity across individuals. Similarly, increased conversational rehearsal may not consistently reinforce the same set of event elements, limiting its impact on overall memory overlap. Indeed, the observed convergence on 42% of the details reported indicates partial alignment around a shared narrative core, while still allowing for individual variability.

Beyond shared content, we also examined whether recollections followed systematic patterns in the order in which event details were recalled. Participants frequently recalled the same key details—notably, the artist’s name and the song title—and some recall sequences appeared in the same order across individuals. These recurring patterns indicate that convergence in collective memory may extend beyond which details are remembered to include how those details are organized during retrieval. In particular, the song title immediately followed the singer’s name more often than expected under random ordering, whereas the reverse sequence occurred at the expected rate. This describes the order of recall but does not show that recalling the singer’s name prompted recall of the title. One possible explanation for this pattern is the fact that the artist’s identity and the song were widely publicized and associated with the victory. Yet, the present data do not allow firm conclusions about whether this pattern was shaped by media coverage, conversational rehearsal, or broader cultural narratives. We postulate that the observed recall structure may reflect general cognitive organization processes, such as associative linking, but may also result from the influence of specific social channels. Furthermore, the frequent recall of a small number of details is consistent with the idea that public accounts provide common reference points for remembering major events. The present study cannot determine which sources contributed to their recurrence, but it suggests that collective remembering can be concentrated around a limited set of widely available details.

Overall, our study offers valuable insights about the extent to which memory representations of public events are common across individuals. The results of the inter-subject memory similarity analysis show that the details of a public event are similar across individuals, underscoring the collective dimension of memory while allowing for individual variation. Moreover, the identification of a recurring recall sequence underscores the role of shared cultural knowledge and the idea that public narratives provide common reference points when people remember major events.

Some limitations of the current study should be mentioned, which also point to important directions for future research. First, our measure of memory similarity is based on proportional overlap between recalled details and does not consider the absolute amount of information recalled. This means that participants may appear highly similar even if each recall only a single detail, if that detail is the same. In such cases, high similarity reflects overlap in a minimal set of memories rather than broadly shared or detailed memory representations. However, in the present sample, such cases were infrequent, as only two participants recalled a single, identical detail, which suggests that this limitation is unlikely to have substantially influenced the overall pattern of results. In addition, some participants may have recalled idiosyncratic details that fell outside the predefined coding categories and were therefore not captured by the similarity measure. More generally, these factors may limit the extent to which similarity scores can be interpreted as reflecting the overall richness of the participant’s memories. Future studies could address this by combining similarity measures with indices of recall quantity and by allowing for more flexible coding schemes that better capture idiosyncratic or less commonly reported details.

Second, event memory accuracy may depend on additional factors not assessed here, such as prior knowledge. Both Ribeiro et al. (2023) and Tinti et al. (2014), studying memories for Football Cups, found that individuals with greater knowledge about football recalled specific matches more accurately. In the context of the Eurovision Song Contest, prior knowledge may include familiarity with the competition format, past editions, or the artist, which can facilitate the integration and retention of specific contest details. Hence, event memory accuracy may depend to some extent on contextual factors, such as media exposure, while also being influenced by more stable knowledge schemas. Future studies should therefore consider assessing prior knowledge, especially for popular events with strong fan communities.

Third, this study assessed memory at a single time point, precluding conclusions about how collective memories converge along time and whether similar sequential patterns would emerge at different moments. Longitudinal designs following the same participants across multiple time points would allow a more direct examination of how collective memories change or stabilise over time. Similarly, the current design does not allow firm conclusions about the mechanisms underlying recall order. Future studies could extend this work by analysing recall order across different types of public events and by testing the processes involved more directly.

Finally, even though evidence for gender differences in event and flashbulb memory is limited, with most studies not examining these relationships, we note that our sample included substantially more women than men, which may limit the generalizability of the findings across genders.