Extended Abstract
Background and Purpose
Coaches occupy a uniquely powerful position in organized sport, extending far beyond technical instruction to shape the psychological and social environment of their teams. Through their control over participation opportunities, performance feedback, recognition, sanctions, and the establishment of behavioral norms, coaches fundamentally influence athletes' developmental experiences, motivation, and well-being. This influence, when exercised constructively, can foster growth, resilience, and team cohesion. However, when characterized by persistent humiliation, authoritarianism, narcissism, self-promotion, and unpredictability, coaching behavior can undermine athletes' dignity, psychological safety, and commitment to healthy team norms, creating a climate of fear and insecurity. Despite extensive examination of destructive leadership in conventional workplace settings, its specific behavioral pathways within the unique context of coach–athlete relationships—characterized by emotional dependency, power imbalance, and high performance pressure—remain insufficiently understood. Drawing upon the complementary theoretical lenses of social learning theory and social exchange theory, the present study proposes that athletes may observe and internalize a coach's hostile conduct as normative when it goes uncorrected, and may also respond to perceived disrespect and injustice by disengaging from team norms and reciprocating negatively. Deviant behaviors—voluntary actions that violate important group norms—are therefore hypothesized to operate as an intermediate behavioral pathway through which toxic coaching leadership becomes associated with more severe interpersonal outcomes, namely bullying and aggression. The study examines both the direct associations of coaches' toxic leadership with athletes' deviant behaviors, bullying, and aggression, and the mediating role of deviant behaviors in these relationships. By doing so, it addresses a critical gap in the sport psychology literature, where most prior research has focused on individual-level outcomes such as burnout or dropout, rather than the collective and relational damage that spreads through team dynamics.
Methods
The present study employed a quantitative, cross-sectional, correlational design grounded in a positivist research philosophy. This design enabled the simultaneous examination of relationships among latent constructs without experimental manipulation, reflecting the naturalistic setting of sport teams. The target population comprised all active male and female athletes participating in team and individual sports in Chaharmahal and Bakhtiari Province, Iran, who had competed for at least one consecutive season in provincial or national leagues. As the precise sampling frame was unavailable due to geographical dispersion and the absence of a comprehensive registry, the sample size was estimated using Cochran's formula for infinite populations (95% confidence level, 5% margin of error), yielding a minimum of 384 participants. Given accessibility constraints and concurrent geopolitical circumstances, a non-probability convenience sampling method was employed. Data were collected via self-administered electronic questionnaires designed on the Porsline platform, with the survey link distributed through sports-specific social media channels. After initial screening and removal of incomplete or invalid responses, 384 valid questionnaires were retained for analysis. Toxic leadership was assessed using Schmidt's 28-item Toxic Leadership Scale, comprising five dimensions: abusive supervision, authoritarian leadership, narcissism, self-promotion, and unpredictability, rated on a five-point Likert scale. Deviant behaviors were measured using Bennett and Robinson's eight-item workplace deviance scale, capturing interpersonal and organizational deviance. Bullying was evaluated using a 45-item questionnaire based on Leymann's classification, covering behaviors threatening communication, social relations, personal reputation, occupational position, and physical health. Aggression was measured using the 29-item Buss–Perry Aggression Questionnaire, encompassing physical aggression, verbal aggression, anger, and hostility. All instruments demonstrated satisfactory psychometric properties in the current sample, with Cronbach's alpha values ranging from 0.904 to 0.952 for full scales. Higher-order reflective-reflective models were specified for the multidimensional constructs of toxic leadership, bullying, and aggression to preserve theoretical dimensionality while reducing estimation complexity. Content and face validity were confirmed by three experts in sport management and sport psychology.
Data analysis proceeded in two stages. Descriptive statistics—means, standard deviations, frequencies, and percentages—were computed using IBM SPSS version 27. Given the violation of multivariate normality assumptions (Kolmogorov–Smirnov test significant), bivariate relationships were examined using Spearman's rank correlation coefficient. For simultaneous testing of the measurement and structural models and estimation of direct and indirect effects, variance-based structural equation modeling using partial least squares (PLS-SEM) was conducted with SmartPLS version 3. The selection of PLS-SEM over covariance-based SEM was justified on three methodological grounds: (a) the predictive orientation of the research in explaining variance in endogenous variables (bullying and aggression), (b) the non-fulfillment of multivariate normality assumptions, and (c) the presence of higher-order constructs, which are more efficiently estimated within the variance-based framework (Hair et al., 2022). Following the two-step procedure recommended by Anderson and Gerbing (1988) and contemporary guidelines (Hair et al., 2022), the measurement model was first assessed through factor loadings (acceptable > 0.70), Cronbach's alpha (> 0.70), composite reliability (> 0.70), and average variance extracted (AVE > 0.50). Discriminant validity was evaluated using both the Fornell–Larcker criterion (Fornell & Larcker, 1981) and the HTMT ratio (Henseler et al., 2015). The structural model was then evaluated through standardized path coefficients (β), t-values derived from bootstrapping with 5,000 subsamples, coefficients of determination (R²), and predictive relevance (Q²), with all paths interpreted as statistical associations due to the cross-sectional design.
Results
The final sample comprised 205 men (53.4%) and 179 women (46.6%), with the largest age group being 35–44 years (29.2%) and 41.7% holding a Master's degree. Most participants (71.4%) reported between 5 and 24 years of sport experience. The mean scores were 3.016 for toxic leadership, 3.091 for deviant behaviors, 3.020 for bullying, and 2.938 for aggression. Among toxic leadership dimensions, authoritarian leadership scored highest; among bullying dimensions, threatening communication was most prominent; and verbal aggression had the highest mean among aggression dimensions. All factor loadings ranged from 0.568 to 0.865 and were significant at p < 0.001. Composite reliability values exceeded 0.85 for all constructs, and AVE values were above 0.50, confirming convergent validity. The Fornell–Larcker criterion supported discriminant validity, with the square root of AVE for each construct exceeding its correlations with other constructs.
Structural path analysis revealed that toxic leadership was positively associated with deviant behaviors (β = 0.386, SE = 0.043, t = 8.977, p < 0.001). Deviant behaviors were positively associated with bullying (β = 0.443, SE = 0.038, t = 11.791, p < 0.001) and aggression (β = 0.515, SE = 0.037, t = 13.828, p < 0.001). Toxic leadership also had significant direct associations with bullying (β = 0.258, SE = 0.046, t = 5.621, p < 0.001) and aggression (β = 0.245, SE = 0.041, t = 5.974, p < 0.001). The indirect effect of toxic leadership on bullying through deviant behaviors was significant (β = 0.171, t = 7.171, p < 0.001), as was the indirect effect on aggression (β = 0.199, t = 7.599, p < 0.001). As direct effects remained significant and all effects were positive, both mediation effects were complementary and partial, with the indirect pathway accounting for approximately 39.9% of the total association with bullying and 44.8% of the total association with aggression. The model explained 14.9% of the variance in deviant behaviors, 35.2% in bullying, and 42.2% in aggression. Q² values of 0.082, 0.184, and 0.272, respectively, indicated predictive relevance for all endogenous constructs. The path from deviant behaviors to aggression was the strongest in the model, highlighting the critical role of norm-violating conduct in the emergence of destructive interpersonal outcomes.
Conclusion
The findings confirm that deviant behaviors constitute a significant complementary pathway through which toxic coaching leadership is associated with bullying and aggression among athletes. Uncorrected humiliation, authoritarianism, and unpredictability from coaches signal to athletes that harmful conduct is acceptable, while perceived disrespect and injustice weaken their commitment to team norms, enabling the normalization of rule-breaking. The substantial indirect effects—accounting for 40–45% of the total relationships—support a two-stage process wherein toxic leadership first alters athletes' perceptions of acceptable norms and weakens normative commitment, which then manifests as bullying and aggression. The persistence of direct effects after controlling for the mediator suggests that additional mechanisms—including anger, perceived injustice, psychological insecurity, a violence-supportive climate, and the quality of the coach–athlete relationship—operate independently as complementary pathways. Theoretically, these findings extend the destructive leadership literature from conventional organizational settings to the dynamic, affectively charged context of sport teams, reconceptualizing deviant behaviors not merely as outcomes but as behavioral conduits that transmit individual-level harm to the collective relational network. Practically, this study underscores the necessity of broadening coach evaluation beyond competitive outcomes and technical competence to include interpersonal conduct, psychological safety monitoring, and team climate assessment. Sports clubs and federations should implement multi-source (360-degree) evaluations of coaching behavior, establish explicit codes of conduct with confidential reporting channels, and include constructive communication, emotion regulation, conflict management, and bullying prevention in coach education and license-renewal programs. Early identification and intervention in minor deviant acts may interrupt the pathway from destructive leadership to more severe interpersonal harm. Limitations include the cross-sectional design, which precludes causal inference; convenience sampling, which limits generalizability; and reliance on self-report measures and instruments originally developed for workplace contexts, necessitating caution in interpretation. Future longitudinal and multi-source studies are recommended to examine temporal ordering, compare sport types and competition levels, and explore moderating roles of gender, age, and experience to refine the model for Iran's specific sporting culture.

Article Message
Toxic coaching leadership should not be treated solely as an individual style problem. When humiliating, authoritarian, and unpredictable conduct is tolerated, it reshapes the informal norms of a team and increases the likelihood of deviant, bullying, and aggressive behaviors among athletes. Prevention therefore requires action at both the coach and organizational levels. Clubs should assess interpersonal conduct alongside competitive performance, establish clear behavioral standards, protect confidential reporting, and include constructive communication, emotion regulation, conflict management, and bullying prevention in coach education and license renewal. Early attention to deviant conduct may interrupt the pathway from destructive leadership to more serious interpersonal harm. Ignoring either the coach's behavior or the team's normative climate will leave the cycle of harm intact.

Ethical Considerations
This study was conducted in full compliance with the ethical standards for research in behavioral and sport sciences. Prior to data collection, all participants were informed about the study's objectives and nature, and it was emphasized that participation was entirely voluntary, with no coercion or financial incentives involved. Questionnaires were administered electronically via the Porsline platform without recording identifying information (e.g., names, phone numbers, or national IDs), ensuring complete anonymity and confidentiality of responses throughout all stages of analysis and reporting. Participants retained the right to withdraw from the study at any time without facing any disciplinary or evaluative consequences.
Authors’ Contributions
All four authors contributed equally and substantially to all stages of the research, including study design, data collection and analysis, drafting the initial manuscript, critical revision of the scientific content, and final approval of the submitted version.
Conflict of Interest
The authors explicitly declare that no financial, organizational, or personal conflicts of interest exist that could have influenced the research findings or the review process. Furthermore, this research received no external funding from commercial, governmental, or charitable entities.
Acknowledgments
The authors extend their sincerest gratitude to each and every athlete from Chaharmahal and Bakhtiari Province who participated in this research, trusting the research process, dedicating their valuable time, and providing honest and accurate responses to the electronic questionnaires. We also warmly thank all those who assisted in distributing the questionnaire link through sports social networks and facilitating access to the target population; their cooperation, though unnamed here, is deeply appreciated and unforgettable.