Negative emotional contagion and online rumor sharing: The mediating role of perceived rumor credibility

Emotional contagion in digital networks and rumor spread

At the individual level, the affect-as-information framework holds that under uncertainty and cognitive load, people often treat their emotional responses as heuristic cues for judging risk or credibility (Schwarz & Clore, 1983; Schwarz, 2012). Accordingly, negative emotional contagion (NEC) —the extent to which negative emotion induced by content resonates with individuals—may serve as a proximal psychological mechanism shaping belief and sharing behavior. Emotional contagion is not limited to offline settings: longitudinal network studies and platform-level experiments demonstrate that emotions spread across interpersonal layers of social networks (Fowler & Christakis, 2008) and can be evoked even in the absence of face-to-face cues (Kramer et al., 2014).

During crises, emotional reactions become more persistent and contagious, producing multi-peak emotional dynamics and conformity-driven diffusion (Pröllochs et al., 2021; Zhou et al., 2023). At the content level, rumor threads and cascades often exhibit stronger anger, fear, and pessimistic tones and are linked to deeper propagation structures (Pröllochs et al., 2021;Vosoughi et al., 2018). Experimental and large-sample behavioral research further shows that high-arousal negative states (e.g., anxiety) can generally amplify people's belief in and willingness to share both false and true information, a pattern consistent with motivation-boosting and heuristic-processing mechanisms (Freiling et al., 2023).

Taken together, these findings suggest that, compared to debunking/true information, rumors are more likely to evoke stronger negative emotional contagion and trigger stronger sharing impulses.

Perceived credibility as a proximal mechanism

NEC may increase sharing intentions directly or indirectly by raising individuals' perceived credibility of information. In the context of high-speed social media browsing, perceived credibility consistently emerges as a proximal determinant of sharing, although its effect varies across platforms, content formats, and audience segments (Ecker et al., 2022; Mang et al., 2024).

Experimental research shows that credibility labels and fact-checking tags generally reduce sharing of false content, but the magnitude of these effects varies substantially across types of warnings and across user groups (Yaqub et al., 2020). At the same time, platform-level findings reveal an “implied truth effect”: when only some false items receive warning labels, unlabeled items are perceived as more accurate (Pennycook & Rand, 2019). Lightweight “accuracy-prompt” interventions robustly improve sharing discernment across platforms and demographic groups (Pennycook et al., 2021; Pennycook & Rand, 2022).

Psychological inoculation and prebunking, which prepare individuals by introducing common deceptive tactics and logical flaws ahead of exposure, have been validated as scalable solutions in real-world digital environments (Roozenbeek et al., 2022; Roozenbeek & van Der Linden, 2019). Meta-analytic evidence confirms their broad effectiveness (Lu et al., 2023). The timing of corrections also influences sustainability of effects (Brashier et al., 2021).

Together, it indicates a credibility-mediation pathway: NEC increases perceived credibility, which in turn promotes sharing.

Present study

To elucidate the interplay among trait emotional susceptibility, NEC, perceived credibility, and sharing behavior, this study advances three hypotheses grounded in prior literature: (H1) Individuals with higher emotional susceptibility will exhibit stronger tendencies to share rumors. (H₂) Exposure to rumor information (as opposed to debunking or verified content) will evoke greater NEC and stronger intentions to share. (H3) NEC will positively predict online sharing intentions, with perceived credibility serving as a mediating factor.

To test these hypotheses, participants' trait emotional susceptibility, NEC in online contexts, perceived credibility of rumor-related information, and sharing intentions were assessed. Each participant was exposed to both rumor and debunking materials for comparison. The analysis proceeded in three stages: first, examining whether trait emotional susceptibility predicts rumor-sharing tendencies (H1); second, comparing emotional and behavioral responses to rumor versus debunking content (H₂); and third, under the rumor condition, testing whether NEC influences sharing through perceived credibility (H3), interpreted in light of prior research on credibility interventions.

Collectively, this study seeks to deepen understanding of how dispositional traits, emotional reactions, and cognitive evaluations jointly shape rumor dissemination in digital environments.

Design and materials

Real-world online rumors (and matched debunking texts) were collected across common knowledge, current affairs, and unexpected incident categories. Five researchers pre-screened 15 rumor materials; a separate panel (n = 5) rated negative emotionality (PANAS-negative items; 5-point scale). Six materials (three rumor texts; three debunking texts) showing high and comparable negative emotional valence across categories were selected.

Participants

A total of 131 questionnaires participated online for this study. After excluding 14 questionnaires with incorrect responses on validity check items and one outlier questionnaire based on response time (less than 3 min or more than 20 min), a sample of 116 frequent social media users (49.1% male; 23.12 ± 4.16 year) were retained.

Measures

Trait emotional susceptibility (TES). Participants' dispositional sensitivity to emotional contagion was measured using the Chinese Emotional Contagion Questionnaire, revised by Zhang et al. (2017) from Doherty (1997)'s Emotional Contagion Scale to better fit the Chinese sociocultural context. The scale consists of 25 items with a 5-point scale (1–5) and demonstrated a good reliability in the present study (Cronbach's α =. 852).

NEC. State-level negative emotional responses evoked by each stimulus were assessed using the negative affect subscale of the Positive and Negative Affect Schedule (PANAS; Watson et al., 1988). Participants rated the extent to which they experienced negative emotions after each material on a 5-point scale (1–5), with higher scores indicating stronger negative emotional contagion. The negative subscale demonstrated a good reliability in the present study (Cronbach's α =. 914).

Perceived credibility (PC). Perceived credibility was measured by asking participants to rate the perceived authenticity of each piece of information on a 5-point Likert scale (1 = very low credibility, 5 = very high credibility). Higher scores reflected stronger beliefs in the credibility of the material.

Sharing tendency (ST). Participants' propensity to spread information was assessed by ratings of their likelihood to like, comment on, or repost the content, using a 5-point scale. Higher scores indicated a greater tendency to share. Sharing behaviors encompassed both online actions (e.g., reposting, commenting, and liking on WeChat, QQ, Weibo, Douyin, and Kuaishou) and offline behaviors (e.g., retelling or discussing information with others in real-life settings).

Procedure

Participants first completed the measure of trait emotional susceptibility. They were then presented with three rumor texts. After reading each text, they sequentially completed the three state measures in the following fixed order: NEC, PC, and online ST. Subsequently, participants were exposed to three debunking texts and completed the same sequence of assessments following each text. The order of materials within each block (rumor vs. debunking) was randomized to minimize ordering effects.

Descriptive statistics

Table 1 indicated that trait emotional susceptibility was moderate (3.83 ± 0.55). In the rumor condition, participants reported higher levels of negative emotional contagion (3.07 ± 0.90), perceived rumor credibility (3.64 ± 0.78), and online sharing tendency (3.17 ± 1.08). In contrast, the debunking condition yielded lower negative emotional contagion (2.32 ± 1.08), lower perceived credibility (3.26 ± 0.87), and lower sharing tendency (2.88 ± 1.16).

Trait susceptibility and sharing

Participants were divided into high (4.26 ± 0.30) and low (3.39 ± 0.35) groups based on the median score (3.82) of trait emotional susceptibility measure. An independent samples t-test revealed that participants with high trait emotional susceptibility exhibited significantly greater rumor sharing tendencies (3.52 ± 1.02) than those with low susceptibility (2.81 ± 1.03), t (114) = 3.54, P < 0.001, Cohen's d= 1.03. Debunking information yielded similar pattern, 3.33 ± 1.14 vs. 2.44 ± 1.01, t (114) = 4.46, P < 0.001, Cohen's d= 1.08. The correlation of trait susceptibility and sharing tendency were. 324 and. 408 for rumor and debunking information, respectively.

The difference between rumor versus debunking

Given the interaction of high vs. low trait emotional susceptibility and rumor vs. debunking information type were not significant (Fs [1, 114] < 1.47, Ps > 0.22), the comparisons between rumor and debunking information type were based on the whole sample. Paired-samples comparisons showed that negative emotional contagion was significantly higher following exposure to rumor than debunking (3.07 vs. 2.32, t [115] = 8.31, P < 0.001, Cohen's d = 0.97), Perceived credibility was lower for rumor than debunking information (3.64 vs. 3.26, t [115]= 4.29, P < 0.001, Cohen's d = 0.96), sharing tendencies were significantly higher for rumor than debunking information (3.17 vs. 2.88, t [115]= 3.75, P < 0. 001, Cohen's d = 0.82).

Predictive relationship between the three measures

Regression analyses demonstrated that negative emotional contagion significantly predicted online sharing tendency (β = 0.422, t [114] = 4.97, P < 0.001; R² = 0.171). negative emotional contagion also significantly predicted perceived credibility (β = 0.286, t [114] = 3.19, P =. 002; R² =. 074), and perceived credibility strongly predicted online sharing tendency (β = 0.591, P < 0.001, R² = 0.344). These findings indicate that both emotional and cognitive appraisals contribute to sharing behavior.

Mediation analysis

A bootstrapped mediation analysis (5, 000 resamples) was conducted to test the mediation effect, with online rumor sharing tendency as the dependent variable, individual negative emotional contagion intensity as the independent variable, and perceived rumor credibility as the mediator. The results indicated a significant indirect effect of negative emotional contagion on online sharing tendency through perceived credibility (indirect effect = 0.147, 95% CI [0.034, 0.237]). The direct effect (0.422) remained significant when the mediator was included, yielding a reduced but still significant direct effect (0.275, 95% CI [0.127, 0.423]). Perceived credibility, thus, partially mediated the relationship between negative emotional contagion and online sharing tendency, accounting for approximately 34.8% of the total effect (Table 2).

Effects of trait emotional susceptibility on sharing (H1)

Consistent with H1, individuals high in trait emotional susceptibility showed significantly higher tendencies to share rumors compared to low-susceptibility individuals. This aligns with the affect-as-information framework, which posits that people often use their emotional reactions as cues when processing uncertain information (Schwarz, 2012; Schwarz & Clore, 1983). Individuals who are dispositionally more sensitive to emotional cues may experience stronger affective resonance when encountering ambiguous or threat-related content, making them more vulnerable to rumor-driven emotional triggers.

This result is also consistent with research showing that certain personality-affective traits—such as emotional contagion proneness or negative affectivity—can increase susceptibility to misinformation and emotionally charged content (Doherty, 1997; Han et al., 2023). Given that rumors often evoke fear, anger, or surprise (Pröllochs et al., 2021; Vosoughi et al., 2018), individuals high in emotional susceptibility may be more reactive and more likely to propagate unverified information.

Rumor vs. debunking information: emotional and behavioral differences

The comparison between the rumor and debunking conditions revealed clear emotional and behavioral divergences, lending support toH₂. Exposure to rumors elicited significantly higher NEC and higher sharing intentions, consistent with evidence that false or unverified content typically triggers stronger high-arousal negative reactions and spreads more widely than verified information (Avram et al., 2020; Vosoughi et al., 2018). Rumors often leverage emotional intensity to drive engagement, which may explain why participants reported stronger emotional resonance and a greater desire to share.

Interestingly, perceived credibility was also higher for rumors than for debunking information, suggesting that emotional responses may enhance the perceived plausibility of rumor content. Prior studies show that affective arousal can bias credibility judgments by increasing reliance on heuristic processing (Ecker et al., 2022; Freiling et al., 2023). This is consistent with crisis communication research demonstrating that spikes in collective negative emotion often precede increases in rumor prevalence (Dong et al., 2020; Zhou et al., 2023), indicating that emotionally intense contexts may amplify the perceived validity and urgency of rumor content.

The absence of a significant interaction between trait susceptibility and information type suggests that the rumor advantage (higher NEC and sharing) is robust across individuals regardless of dispositional differences. In other words, emotional reactions to rumor information appear to exert a general effect on users, echoing findings that emotional arousal can amplify sharing tendencies across broad populations (Freiling et al., 2023).

Emotional and cognitive predictors of sharing

The regression analyses lent support to H3 by demonstrating that both emotional and cognitive factors contribute to rumor-sharing behavior. The emotional effect is consistent with prior research showing that negative affect—especially anxiety, anger, or fear—can increase belief in and sharing of both true and false information by heightening motivation and reducing analytic scrutiny (Freiling et al., 2023; Lühring et al., 2024).

Meanwhile, the strong predictive effect of perceived credibility echoes extensive literature identifying credibility judgments as one of the most consistent proximal determinants of sharing behavior (Ecker et al., 2022; Mang et al., 2024). Regardless of its accuracy, when users perceive information as credible, they are more likely to engage in downstream behavioral actions such as liking, reposting, or recommending the content.

Mediation through perceived credibility

The mediation analysis further revealed that perceived credibility partially mediated the effect of NEC on sharing, accounting for roughly 34.8% of the total effect. This indicates that emotional reactions shape sharing behavior via two complementary pathways: a direct emotional pathway, where NEC increases impulsive sharing tendencies; an indirect cognitive pathway, where NEC increases credibility evaluations, which in turn promote sharing.

This dual-pathway model aligns with research showing that emotionally charged content can appear more plausible due to heightened involvement and reduced critical scrutiny (Pennycook & Rand, 2019; Pennycook & Rand, 2022). It also fits findings that fact-checking labels and pre-bunking interventions can meaningfully reshape credibility judgments and reduce misinformation spread (Lu et al., 2023; Roozenbeek et al., 2022; Yaqub et al., 2020;).

At the same time, the substantial direct effect suggests that even when users recognize that information may be questionable, emotional resonance alone may still drive sharing—a pattern widely documented in work on emotional contagion and cascade virality (Fowler & Christakis, 2008; Kramer et al., 2014; Pröllochs et al., 2021).

Theoretical and practical implications

Taken together, the findings of this study extend theoretical frameworks that emphasize the central role of affect in information processing under conditions of uncertainty. By demonstrating that negative emotional contagion influences rumor sharing through both direct emotional pathways and indirect cognitive pathways, the results provide empirical support for dual-process accounts of misinformation diffusion. Specifically, the findings reinforce perspectives that position emotion as a primary driver of information spread in uncertain contexts (Ecker et al., 2022; Schwarz & Clore, 1983), while simultaneously highlighting the malleable nature of credibility judgments as a critical cognitive mechanism through which emotional responses shape behavior.

Importantly, the results carry clear practical implications for misinformation intervention strategy. The partial mediation effect indicates that effective interventions should address both affective and cognitive processes. Emotion-targeted strategies, such as calming or de-escalating messaging during crisis situations, may help curb the initial emotional amplification that fuels rumor diffusion. In parallel, credibility-focused interventions, including prebunking techniques and accuracy prompts, can disrupt the indirect pathway linking negative emotional contagion to sharing by recalibrating users' credibility assessments. Moreover, individuals with high emotional susceptibility may represent a particularly vulnerable group, suggesting that tailored or intensified interventions may be necessary to counteract emotionally amplified rumor sharing in this population.

Limitations and future directions

Despite its contributions, several limitations warrant consideration and outline opportunities for further research.

Firstly, the sample size, although adequate for detecting medium-to-large effects, was limited and may not fully capture population-level variability in emotional susceptibility or media literacy. Future studies should recruit more diverse and larger samples across cultural, demographic, and platform-specific contexts to enhance generalizability.

Secondly, the study employed controlled rumor and debunking stimuli. While this increases internal validity, real-world online environments involve dynamic and multimodal content (images, videos, comments, algorithmic recommendations). Ecologically valid designs, such as field experiments on live platforms, would better capture naturalistic rumor dynamics.

Thirdly, the present model focuses on emotional and credibility processes. However, other factors, such as analytical thinking, political ideology, trust in institutions, social influence, and platform affordances, may interact with emotional susceptibility. Testing moderating effects would clarify the boundary conditions of the observed mechanisms.