Introduction: Peer victimization is the experience of being a target of intentional negative or aggressive actions by peers (Card & Hodges, 2008). Peer victimization at school is a growing concern because of its high prevalence rates (Finkelhor et al., 2013) and potential harmful impacts on student mental health and developmental outcomes (e.g., Singham et al., 2017). Meta-analytic studies have found that peer victimization is significantly and positively associated with externalizing behavior problems such as aggressive or disruptive behaviors (Reijntjes et al., 2010) and internalizing behavior problems such as anxiety, depressive symptoms, and withdrawal (Reijntjes et al., 2011). There are different theoretical models explaining the associations between peer victimization, externalizing, and internalizing problems. Attribution theory posits that peer victimization influences one's attributions (e.g., self-blame or peer blame) that in turn influences externalizing or internalizing behaviors (Graham & Juvonen, 2001; Perren et al., 2013). Interpersonal risk model views victimization as a major stressor that would influence interpersonal relationships (Kochel et al., 2012; Gong et al., 2021). These perspectives can be categorized as "victimization-driven models," suggesting that victimization experiences precede or predict later behavioral maladjustment. In contrast, "symptoms-driven models" propose an opposite direction of influences that externalizing or internalizing symptoms confer an increased risk for future victimization experience. Some suggest "mutually reinforcing or cyclical risk processes" between peer victimization and externalizing or internalizing problems (Davis et al., 2019; Kochel et al., 2012). Others view peer victimization as a mediating process through which externalizing or internalizing behavior problems lead to the other domain of behavior problems, often called cascading models. For example, children with externalizing problems may be at higher risk of experiencing peer victimization (Dodge et al., 1997; Ladd & Troop-Gordon, 2003; Reijntjes et al., 2011), which may in turn lead to an increased risk for internalizing behaviors by undermining self-esteem and inducing feelings of frustration (Bell-Dolan et al., 1995; Zwierzynska et al., 2013). Inversely, children with internalizing problems may be at higher risk of being bullied (Averdijk et al., 2016; Kochel et al., 2012; Reijntjes et al., 2010) and victims of peer rejection or bullying may express their feelings of frustration, worthless, and marginalization by being affiliated with academically and/or socially deviant peers, which can subsequently lead to an increased risk for the emergence or exacerbation of externalizing problem behaviors (Dishion et al., 2010; Monahan et al., 2009; Reijntjes et al., 2011). This study aims to test which of these theoretical models may explain the associations between externalizing, and internalizing problems in elementary school children by focusing on within-person directional relations among these measures. Methods: We analyzed the data drawn from the ECLS-K: 2011, a longitudinal study of a nationally representative sample of U.S. school-aged children. Analytic sample includes 6,945 children followed from Grade 3 to Grade 5. The study variables include: teacher-reported externalizing and internalizing problems and student self-reported peer victimization in verbal, social, reputational, and physical subdomains. We used random-intercept cross-lagged panel modeling (RI-CLPM: Hamaker et al., 2015) as an alternative to the standard CLPM. In contrast to CLPM, which is not structured to disaggregate within- and between-person levels of associations, RICLPM models within-person directional associations after partialling out stable between-person differences. As shown in Figure 1, total variance in each observed repeated measure (grey squares) is separated into between- and within-person components. The black ovals represent between-person components, i.e., latent stability of each construct that differs across students (often called random intercepts). Correlations among latent stability factors represent the between-person level of associations among the three constructs (upper panel of the diagram). "EX3" to "IN5" in the white ovals represent time-specific latent variables that capture each student's time-specific deviations from his or her typical levels of victimization or behavior problems. These time-specific latent variables are used to examine the within-person level of associations among three constructs (bottom panel of the diagram). Autoregressive paths from one variable at time t to the same variable at time t+1 represent the temporal stability of each construct. Cross-lagged paths from one variable at time t to the other variable at time t+1 indicate the predictive effects of one domain at each time point on the other domain at the immediate next time point. Figure 2 shows four different hypothesized models of directional associations between peer victimization, externalizing, and internalizing problems. Analyses were done separately for each victimization subdomain and overall victimization. Results and Next Step: About 49% to 66% of the variance in student self-reported victimization were attributable to the within-person differences that were stable over the upper elementary grades. At the between-person level, peer victimization, externalizing and internalizing problems were highly and positively correlated with one another. Correlation ranged from 0.42 to 0.60 between externalizing and victimization, 0.29 to 0.48 between internalizing and victimization, and 0.37 to 0.38 between externalizing and internalizing. At the within-person level, cross-lagged effects were generally non-significant. Only significant cross-lagged effect was: a negative path from teacher-reported physical victimization at Grade 4 to internalizing problems at Grade 5; and a negative path from teacher-reported child externalizing problems at Grade 3 to child-reported physical victimization at Grade 4. Therefore, the results do not provide direct support for any of hypothetical models described above. Instead, the results suggest that peer victimization, externalizing, and internalizing problems might result from early risk factors that are common to all these conditions. A recent RI-CLPM applied study of adolescents suggests gender differences in these dynamics, i.e., no relation between depression and victimization for males and an evidence of transactional processes for females. Given this, our next step is to examine gender differences using multi-group RI-CLPM. More results and implications of the results will be presented at the conference.