Within‐person longitudinal associations between depressive symptoms and interpersonal problems in adolescent girls Authors: Leehyun Yoon, Kelsey E Magee, Stephanie D Stepp, Erika E Forbes, Kate E Keenan, Alison E Hipwell, Amanda E Guyer Published: 2026 ABSTRACT AbstractBackgroundAdolescence, particularly for girls, is a period of heightened risk for depressive symptoms and increased interpersonal stress. Although depressive symptoms and interpersonal problems are known to co‐occur, prior work has predominantly examined between‐person associations, leaving within‐person processes unclear. This study addressed this gap by testing longitudinal within‐person associations between depressive symptoms and interpersonal problems in adolescent girls over 4 years, evaluating whether links were reciprocal or unidirectional.MethodsParticipants included 543 adolescent girls from the Pittsburgh Girls Study. Across four waves (mean ages = 16.5–19.3 years), girls self‐reported depressive symptoms and eight types of interpersonal problems: domineering, vindictive, cold, socially avoidant, nonassertive, exploitable, overly nurturant, and intrusive. For each interpersonal problem, we fit both constrained and unconstrained random intercept cross‐lagged panel models and selected the constrained model unless it demonstrated a significantly poorer fit than the unconstrained model, prioritizing parsimony.ResultsThe selected models demonstrated good fit for all eight interpersonal problems. Within‐person increases in depressive symptoms predicted subsequent within‐person increases in overly nurturant and socially avoidant problems, whereas these interpersonal problems did not predict subsequent within‐person increases in depressive symptoms. By contrast, depressive symptoms and domineering problems reciprocally predicted one another over time.ConclusionsThis study clarifies how depressive symptoms and interpersonal problems influence one another over time within individuals in adolescent girls. Findings highlight domineering problems as a potential intervention target for breaking cycles of mutual reinforcement with depressive symptoms. Moreover, addressing depressive symptoms may yield social benefits by reducing subsequent overly nurturant and socially avoidant problems.Keywords: Adolescence, girls, depressive symptoms, interpersonal problems, longitudinal modeling Depressive symptoms and interpersonal problems, defined as maladaptive patterns in relating to others, are strongly interrelated. According to the stress generation theory (Hammen, 1991, 2005) and the interpersonal theory of depression (Coyne, 1976; Joiner, 1999), depressed individuals are not merely passive recipients of stressful experiences but may also contribute to their stressors through maladaptive social behaviors. For example, excessive dependence on others for validation can provoke conflict or rejection. These negative social responses deprive individuals of the social connection they need, intensifying depressive symptoms. In turn, worsening symptoms further fuel maladaptive interpersonal behaviors, creating a self‐perpetuating cycle of social dysfunction and emotional distress. The link between depressive symptoms and interpersonal problems may be especially strong during adolescence due to heightened social sensitivity (Guyer, Silk, & Nelson, 2016; Nelson, Leibenluft, Mcclure, & Pine, 2005; Somerville, 2013). Particularly, adolescent girls place a higher value on social acceptance than boys (Guyer, Caouette, Lee, & Ruiz, 2014; Guyer, Mcclure‐Tone, Shiffrin, Pine, & Nelson, 2009; Rudolph, 2002) and report more frequent exposure to interpersonal stressors, such as peer conflict or rejection, which helps explain their higher rates of depressive symptoms compared to boys (Hamilton, Stange, Abramson, & Alloy, 2015; Hankin, Mermelstein, & Roesch, 2007; Shih, Eberhart, Hammen, & Brennan, 2006; Stange, Hamilton, Abramson, & Alloy, 2014). Mid‐ to late adolescence represents a particularly important developmental window, as depressive symptoms increase relative to earlier adolescence (Chen, Haas, Gillmore, & Kopak, 2011; Mak, Fosco, & Lanza, 2021) and coincide with the period approaching the peak age of onset of major depressive disorder (Solmi et al., 2022). In parallel, peer involvement increases across this period (Lam, Mchale, & Crouter, 2014), creating greater opportunities for interpersonal problems and their associations with depressive symptoms to emerge. Accordingly, understanding the dynamic association between interpersonal problems and depressive symptoms in girls during this developmental period is essential for designing interventions aimed at addressing social dysfunctions and depression. One useful framework for examining interpersonal problems in relation to depressive symptoms is the interpersonal circumplex model (Horowitz, Rosenberg, Baer, Ureño, & Villaseñor, 1988). This model organizes interpersonal functioning along two orthogonal dimensions: agency (ranging from dominance to submission) and communion (ranging from affiliation to detachment). Their intersection yields eight interpersonal problem domains: domineering (high agency), vindictive (high agency/low communion), cold (low communion), socially avoidant (low agency/low communion), nonassertive (low agency), exploitable (low agency/high communion), overly nurturant (high communion), and intrusive (high agency/high communion). These domains are arranged in a circular structure, with adjacent domains sharing similar interpersonal features, whereas opposing domains reflect contrasting interpersonal tendencies. The interpersonal circumplex is considered comprehensive because it was derived from interpersonal problems reported by a large group of psychotherapy patients (Horowitz et al., 1988). Prior research using the interpersonal circumplex has largely focused on adult clinical samples and shows that depression is associated with elevated overall interpersonal problems (i.e., higher total scores across domains), reflecting broad interpersonal dysfunction (Iovoli, Rubel, Steinbrenner, & Lauterbach, 2025; McFarquhar, Luyten, & Fonagy, 2018). At the domain level, depression has been linked to multiple interpersonal problem domains, although associations with intrusive tendencies have been less consistent (Locke, Sayegh, Weber, & Turecki, 2018; McEvoy, Burgess, Page, Nathan, & Fursland, 2013). Studies further indicate that interpersonal problems across multiple domains tend to improve over the course of psychotherapy for depression, with some studies reporting improvement across all domains (Holtforth, Lutz, & Grawe, 2006; Renner et al., 2012; Vittengl, Clark, & Jarrett, 2003). Although descriptive, multiple studies (Locke et al., 2018; Renner et al., 2012; Vittengl et al., 2003) have suggested a relative prominence of interpersonal problems in withdrawal‐related (i.e., social avoidance) and affiliative–submissive patterns (e.g., overly nurturant, nonassertive, and exploitable). Consistent with this, circumplex profile analyses (Locke et al., 2018) indicated peak elevations in low‐agency regions, and models accounting for shared variance across subdomains identified associations with social avoidance and overly nurturant behaviors (McEvoy et al., 2013), although it additionally identified an effect of dominance. Only a small number of studies using interpersonal circumplex measures have focused on adolescent samples, and existing evidence is limited to clinical samples. In one such study (Israel & Langeveld, 2021), internalizing symptoms were associated with most interpersonal domains except intrusive problems, with the strongest associations observed for socially avoidant, nonassertive, and overly nurturant problems. Similarly, another study (Viana & Stevens, 2013) found that nonassertive and overly nurturant problems remained significantly associated with depressive symptoms after accounting for overlap among interpersonal dimensions, with overly nurturant tendencies additionally mediating the association between anxiety and depressive symptoms. Taken together, these findings tentatively suggest that adolescent depression may be closely linked to withdrawal‐related (e.g., social avoidance) and affiliative‐submissive patterns (e.g., overly nurturant, nonassertive), although evidence in adolescents remains limited. Consistent with this interpretation, longitudinal research on interpersonal problems related to adolescent depression, although not framed within the interpersonal circumplex, has identified interpersonal tendencies that map onto affiliative–submissive patterns. For example, dependency, an overreliance on others for emotional support and validation, has been consistently linked to depression. Among youth, dependency predicted heightened interpersonal stress (Shih, Abela, & Starrs, 2009), negative life events and subsequent emotional distress (Shahar & Priel, 2003), and depressive symptoms following stressful experiences (Adams, Abela, Auerbach, & Skitch, 2009). Studies with college students found that dependency interacts with interpersonal stress to predict depression (Shahar, Joiner Jr, Zuroff, & Blatt, 2004) and that overly dependent and caring tendencies longitudinally predicted greater interpersonal stress (Shih & Eberhart, 2010). Studies with designs to test or at least infer bidirectional associations have suggested reciprocal relations exist between depressive symptoms and affiliative‐submissive and withdrawal‐related interpersonal problems. For example, sociotropy, a construct involving the sacrifice of personal needs to preserve social connections, has been linked to increased interpersonal stress and subsequent depressive symptoms, while also showing associations with a history of depression (Shih, 2006). Moreover, a study of adolescent girls found that reassurance‐seeking behaviors, repeatedly asking others for validation, predicted increases in depressive symptoms in the context of negative peer experiences, while depressive symptoms coupled with peer rejection also predicted elevated reassurance‐seeking (Prinstein, Borelli, Cheah, Simon, & Aikins, 2005). Moreover, social withdrawal or avoidance has been prospectively associated with increases in depressive symptoms (Gazelle, Shafer Lundin, Lei, Cao, & Litchfield, 2024; Katz, Conway, Hammen, Brennan, & Najman, 2011; Smith, Barstead, & Rubin, 2017). Evidence also suggests that depressive symptoms may also lead to greater avoidance of social interactions (Allen, Chango, Szwedo, & Schad, 2014), suggesting bidirectional relations. The present study investigated the longitudinal associations between depressive symptoms and a broad range of interpersonal problems in adolescent girls, as organized within the interpersonal circumplex. Prior research on this topic has largely focused on between‐person differences or has not disentangled between‐person and within‐person associations. We addressed this gap using a random intercept cross‐lagged panel model (RI‐CLPM) (Hamaker, Kuiper, & Grasman, 2015) with four annual waves of data. This approach separates stable between‐person differences from intraindividual fluctuations, testing whether deviations from an adolescent's typical level of depressive symptoms predict subsequent within‐person deviations in interpersonal problems, and vice versa. Capturing within‐person processes is particularly valuable in adolescence, a period of substantial social and emotional change (Crone & Dahl, 2012; Guyer et al., 2016). Even if not severe relative to one's peers, deviations from one's own normative patterns of interpersonal tendencies may be perceptible to others and subtly alter social interactions (Bian, Liang, Bian, & Zhao, 2026; Zhou, Wang, Zhu, & Gong, 2024). Similarly, within‐person increases in depressive symptoms may signal disruptions in emotional, cognitive, or physiological self‐regulation. These disruptions can place increased demands on the individual's regulatory capacity, which, if sustained over time, may contribute to cumulative physiological burden (i.e., allostatic load; McEwen, 1998) and compromise adaptive functioning. Identifying the antecedents and consequences of such within‐person fluctuations is therefore critical for early detection and prevention of socioemotional difficulties, as well as for informing personalized interventions. Based on longitudinal work reviewed above and prior research (e.g., Israel & Langeveld, 2021; Locke et al., 2018) indicating a relative prominence of the affiliative–submissive and withdrawal regions of the circumplex in relation to depression, we hypothesized that nonassertive, exploitable, and overly nurturant problems, as well as socially avoidant tendencies, would show bidirectional associations with depressive symptoms. Other interpersonal domains were examined exploratorily, consistent with prior evidence that depression is associated with broad interpersonal dysfunction across circumplex domains. Method Participants Participants were 543 girls from the youngest cohort (cohort 5) of the Pittsburgh Girls Study (PGS), a longitudinal study tracking psychosocial functioning from childhood through young adulthood. All were assigned female at birth. Consent/assent was obtained, and the study was approved by the University of Pittsburgh's Human Research Protection Office. The current study used data from Waves 12 to 15 (mean ages = 16.47–19.33 years; see Table S1 for age details), during which depressive symptoms and interpersonal problems were assessed. Of the original 588 participants, 45 were excluded due to missing all key variables. The final sample (N = 543) was 41.8% Caucasian, 52.3% African American, 1.5% Asian, and 4.4% multiracial. At Waves 12, 13, 14, and 15, respectively, 6.63%, 7.18%, 7.73%, and 9.94% of participants received public assistance; 81.77%, 78.45%, 84.16%, and 85.27% did not; and 11.60%, 14.36%, 8.10%, and 4.79% did not report. Of these 543 participants, 454 had complete data across all waves. For partial responses (e.g., 3 of 4 items), average scores were computed to generate a summary score. The proportion meeting symptom‐based criteria for major and minor depression ranged from 4.68% to 7.93% and 3.46% to 7.72% across waves, respectively (see Table S2). Measures Depressive symptoms At Wave 12, participants completed the Adolescent Symptom Inventory (ASI)‐4 (Gadow & Sprafkin, 1997), transitioning to the Adult Self‐Report Inventory (ASRI)‐4 (Gadow, Sprafkin, & Weiss, 2004) at Waves 13–15. Both included 10 items assessing symptoms such as depressed mood, lack of interest or pleasure, suicidal ideation, feelings of worthlessness or guilt, low energy, changes in appetite, changes in sleep habits, difficulty concentrating, lack of confidence, and a sense that things never work out. The ASI‐4 includes an additional item related to changes in activity level, which was excluded to ensure consistency across waves. For ASI‐4, most symptoms were rated on a 4‐point scale (i.e., 0 (never), 1 (sometimes), 2 (often), or 3 (very often)), but changes in appetite, sleep habits, and concentration were rated using binary options (i.e., 0.5 (no) or 2.5 (yes)), consistent with the original development and scoring of the ASI‐4. All ASRI items used a 4‐point scale. The possible score ranges were 1.5–28.5 at Wave 12 and 0–30 at Waves 13–15. Items were summed to create a symptom severity score. Internal consistency was excellent across waves (McDonald's ω = .85–.90). Interpersonal problems The Inventory of Interpersonal Problems‐Circumplex Scales‐Short Form (Soldz, Budman, Demby, & Merry, 1995) assessed eight types of interpersonal problems via 32 items (4 per subscale): Domineering (too aggressive; e.g., “I try to control other people too much”), Vindictive (suspicious and distrustful; e.g., “It is hard for me to be supportive of another person's goals”), Cold (unaffectionate; e.g., “I keep other people at a distance too much”), Socially Avoidant (socially anxious and shy; e.g., “It is hard for me to socialize with other people”), Nonassertive (failing to be forceful; e.g., “It is hard for me to confront people with problems that come up”), Exploitable (too trusting and permissive; e.g., “It is hard for me to let other people know when I am angry”), Overly Nurturant (too eager to please others; e.g., “I put other people's needs before my own too much”), and Intrusive (seeking attention inappropriately; e.g., “I tell personal things to other people too much”). Items were rated on a 0–4 scale. Subscale scores were averaged. Internal consistency was excellent for six subscales (Cold, Socially Avoidant, Nonassertive, Exploitable, Overly Nurturant, and Intrusive; ω = .76–.90) and acceptable for two (Domineering and Vindictive; ω = .65–.74). Data analysis strategy Little's test indicated data were not missing completely at random (χ 2 = 472, df = 251, p < .001). To assess whether data could be considered Missing at Random (MAR), we examined whether the missingness of each variable at a given wave (coded 1 = missing, 0 = not missing) was associated with the average of that variable across prior waves using Welch's t‐tests. All tests were nonsignificant (see Table S3), suggesting the missing data mechanism may correspond to MAR. Missing data were handled using full information maximum likelihood (FIML). Before the primary analysis, we calculated the complement of the intraclass correlations (i.e., 1‐ICC) to confirm substantial within‐person variability. Next, we estimated eight RI‐CLPMs to examine the within‐person longitudinal association between the eight subtypes of interpersonal problems and depressive symptoms. The RI‐CLPM approach decomposes interindividual (between‐person) and intraindividual (within‐person) associations between two variables (Hamaker et al., 2015). For the between‐person variables (random intercepts representing a person's typical level of depressive symptoms and interpersonal problems), two latent random intercepts were regressed on the observed data from four waves, with factor loadings set to 1. These random intercepts (one for interpersonal problems and one for depressive symptoms) were allowed to correlate. To specify within‐person changes in depressive symptoms and interpersonal problems (i.e., deviations from a person's typical level), four latent factors were regressed on the four‐wave observed variables, with factor loadings set to 1. The measurement error variances of the observed scores were set to 0, ensuring that all variances of observed scores were captured by the between‐person and within‐person latent factors. To estimate stability and cross‐lagged paths, all within‐person latent factors (except for the first wave) were regressed on the within‐person latent factors of previous waves. The stability paths test whether a person shows a similar deviation from their typical level at a subsequent wave. The cross‐lagged paths (our core interest) examine whether an increased deviation from a person's typical level in one variable predicts an increased deviation from their typical level in another variable at a subsequent wave. Concurrent within‐person latent variables were allowed to correlate, testing whether an increased deviation from a person's typical level in one variable is associated with an increased deviation from their typical level in another variable concurrently. Parameters were estimated with a robust maximum likelihood estimator (MLR) to handle non‐normality of data and FIML to address missing data. Model fit was assessed using the following criteria: CFI ≥0.95, RMSEA ≤0.06, and SRMR ≤0.08 (Hu & Bentler, 1999). Although the chi‐square statistic was calculated, we did not rely on it for model evaluation, as its significance level is highly dependent on sample size. It has been suggested that most models are rejected when the sample size exceeds 400 (Stone, 2021). We compared constrained and unconstrained models using a chi‐square difference test. In the constrained models, cross‐lagged paths, stability paths, and concurrent associations were constrained to be equal across all waves, whereas they were allowed to vary across waves in the unconstrained models. In the constrained model, the concurrent association at the first wave was freely estimated and not included in the constraints, as there is no prior wave to inform expectations of temporal stability in this residual correlation. Unless the constrained model demonstrated a significantly poorer fit than the unconstrained model, we selected the constrained model for parsimony. Analysis code for the RI‐CLPM models is available on the Open Science Framework: https://doi.org/10.17605/OSF.IO/V82Z9. Results Descriptive statistics Descriptive statistics for depressive symptoms and interpersonal problem variables are presented in Tables 1 and 2 (see Table S4 for exploratory interpersonal problem variables). All variables were positively skewed. For interpersonal problem variables, there was a substantial proportion of zero responses (ranging from 29% to 64%). A subset of participants reported zero across all available waves (7%–31% across interpersonal problem variables), indicating no within‐person variation. These participants did not contribute to the estimation of within‐person cross‐lagged or autoregressive effects involving interpersonal problems. Accordingly, the effective sample size for the within‐person effects of interest ranged from 331 to 483, depending on the interpersonal problem variable. Correlations among study variables are presented in Tables S5–S8. Table 1Descriptive statistics for depressive symptomsWave 12Wave 13Wave 14Wave 15 N 491492499517Missing52514426Mean5.536.086.46.44Standard deviation4.255.014.654.91Median4.55.566IQR6766Minimum1.5000Maximum20.5282425Skewness1.090.850.510.74Kurtosis0.640.79−0.070.49 N < LB a 0000 N > UB a 68410 N (%) at min b 145 (29%)84 (17%)63 (13%)73 (14%) a LB and UB denote lower and upper bounds defined as 1.5 × IQR below the first quartile (Q1) and above the third quartile (Q3), respectively. b N at min represents the number (percentage) of participants scoring the minimum possible value on the scale at each wave, with percentages calculated based on the number of participants with available data at that wave. Descriptive statistics for depressive symptoms LB and UB denote lower and upper bounds defined as 1.5 × IQR below the first quartile (Q1) and above the third quartile (Q3), respectively. N at min represents the number (percentage) of participants scoring the minimum possible value on the scale at each wave, with percentages calculated based on the number of participants with available data at that wave. Table 2Descriptive statistics for hypothesized interpersonal problem variablesWave 12Wave 13Wave 14Wave 15Socially Avoidant N 492492499517Missing51514426Mean0.350.410.430.42SD0.620.670.690.68Median0000IQR0.50.50.750.5Minimum0000Maximum4443.5Skewness2.592.242.052.04Kurtosis8.355.694.474.28 N < LB a 0000 N > UB a 29412850 N (%) at 0 b 288 (59%)280 (57%)271 (54%)289 (56%) N (%) all‐zero c 136 (26%)Nonassertive N 492492499517Missing51514426Mean0.300.380.420.39 SD 0.560.650.700.66Median0000IQR0.310.560.50.5Minimum0000Maximum3.2543.53.5Skewness2.282.252.062.14Kurtosis5.435.934.164.63 N < LB a 0000 N > UB a 71415043 N (%) at 0 b 316 (64%)300 (61%)283 (57%)305 (59%) N (%) all‐zero c 161 (31%)Exploitable N 492492499517Missing51514426Mean0.400.440.490.49 SD 0.580.620.660.68Median0.130.250.250.25IQR0.50.750.750.75Minimum0000Maximum33.533.75Skewness1.741.791.471.87Kurtosis2.63.41.593.85 N < LB a 0000 N > UB a 45203028 N (%) at 0 b 246 (50%)226 (46%)236 (47%)220 (43%) N (%) all‐zero c 96 (19%)Overly Nurturant N 492492499517Missing51514426Mean0.600.670.750.7 SD 0.70.790.860.84Median0.250.50.50.5IQR111.251.25Minimum0000Maximum43.7543.75Skewness1.521.381.261.23Kurtosis1.991.581.070.91 N < LB a 0000 N > UB a 121799 N (%) at 0 b 199 (40%)171 (35%)170 (34%)158 (31%) N (%) all‐zero c 53 (10%) a LB and UB denote lower and upper bounds defined as 1.5 × IQR below Q1 and above Q3. b N at 0 represents the number (percentage) of participants scoring 0 at each wave, based on available data at that wave. c N all‐zero represents the number (percentage) of participants with ≥2 observed waves, all equal to 0, based on participants with at least two non‐missing observations. Descriptive statistics for hypothesized interpersonal problem variables LB and UB denote lower and upper bounds defined as 1.5 × IQR below Q1 and above Q3. N at 0 represents the number (percentage) of participants scoring 0 at each wave, based on available data at that wave. N all‐zero represents the number (percentage) of participants with ≥2 observed waves, all equal to 0, based on participants with at least two non‐missing observations. Variances explained by within‐person fluctuations The complement of the intraclass correlations (1‐ICC) indicated that a substantial proportion of variance in all variables was attributable to within‐person fluctuations, rather than to stable individual differences: depressive symptoms (56%), domineering (51%), cold (49%), socially avoidant (54%), nonassertive (54%), exploitable (58%), overly nurturant (55%), and intrusive (57%) problems. These values (~50%) indicate a balance between within‐ and between‐person variance, appropriate for RI‐CLPM estimation (Hamaker et al., 2015). RI‐CLPM Model fit For six of the models, which included the domineering, vindictive, cold, socially avoidant, overly nurturant, and intrusive interpersonal problem subscales, constrained RI‐CLPM showed excellent fit comparable to unconstrained RI‐CLPM and was retained for parsimony. For the other two models, unconstrained RI‐CLPM provided a better fit and was retained. Model fit indices are presented in Table 3. Table 3Model fit indices for models including each of eight interpersonal problem subscalesSubscaleConstrainedUnconstrainedDifference testRI‐CLPM fitRI‐CLPM fitDomineering χ 2 (25) = 52.25, p = .001, CFI = 0.98, RMSEA = 0.05, SRMR = 0.05 χ 2(15) = 43.09, p < .001, CFI = 0.98, RMSEA = 0.06, SRMR = 0.04Δχ 2 = 11.57, Δdf = 10, p = .31Vindictive χ 2 (25) = 50.94, p = .002, CFI = 0.97, RMSEA = 0.04, SRMR = 0.04 χ 2(15) = 36.59, p = .002, CFI = 0.98, RMSEA = 0.05, SRMR = 0.04Δχ 2 = 15.18, Δdf = 10, p = .13Cold χ 2 (25) = 39.06, p = .04, CFI = 0.99, RMSEA = 0.03, SRMR = 0.04 χ 2(15) = 38.85, p < .001, CFI = 0.98, RMSEA = 0.05, SRMR = 0.03Δχ 2 = 4.18, Δdf = 10, p = .94Socially Avoidant χ 2 (25) = 43.10, p = .01, CFI = 0.98, RMSEA = 0.04, SRMR = 0.04 χ 2(15) = 27.95, p = .02, CFI = 0.99, RMSEA = 0.04, SRMR = 0.03Δχ 2 = 15.48, Δdf = 10, p = .12Nonassertive χ 2(25) = 56.41, p < .001, CFI = 0.97, RMSEA = 0.05, SRMR = 0.05 χ 2 (15) = 34.58, p = .0028, CFI = 0.98, RMSEA = 0.05, SRMR = 0.04 Δχ 2 = 21.75, Δdf = 10, p = .02Exploitable χ 2(25) = 57.07, p < .001, CFI = 0.97, RMSEA = 0.05, SRMR = 0.04 χ 2 (15) = 34.45, p = .003, CFI = 0.98, RMSEA = 0.05, SRMR = 0.03 Δχ 2 = 22.57, Δdf = 10, p = .02Overly Nurturant χ 2 (25) = 58.62, p < .001, CFI = 0.97, RMSEA = 0.05, SRMR = 0.04 χ 2(15) = 43.78, p < .001, CFI = 0.98, RMSEA = 0.06, SRMR = 0.04Δχ 2 = 16.05, Δdf = 10, p = .10Intrusive χ 2 (25) = 47.80, p = .004, CFI = 0.98, RMSEA = 0.04, SRMR = 0.04 χ 2(15) = 49.07, p < .001, CFI = 0.97, RMSEA = 0.07, SRMR = 0.04Δχ 2 = 5.86, Δdf = 10, p = .83Each row presents model fit indices for the constrained and unconstrained versions of the random intercept cross‐lagged panel models (RI‐CLPM) involving depressive symptoms and a specific interpersonal problem subscale. Bolded values indicate the model selected based on chi‐square difference testing. Model fit indices for models including each of eight interpersonal problem subscales Each row presents model fit indices for the constrained and unconstrained versions of the random intercept cross‐lagged panel models (RI‐CLPM) involving depressive symptoms and a specific interpersonal problem subscale. Bolded values indicate the model selected based on chi‐square difference testing. Within‐person cross‐lagged paths We found positive reciprocal cross‐lagged effects between domineering problems and depressive symptoms and unidirectional cross‐lagged effects from depressive symptoms to socially avoidant and overly nurturant problems (Figures 1, 2, 3). In five models, which included vindictive, cold, intrusive, nonassertive, and exploitable problems, no cross‐lagged effects were observed (Figures S1–S5). Random intercept cross‐lagged panel model examining associations between domineering problems and depressive symptoms. Dep, depressive symptoms; DOM, domineering problems; RI, random intercept. Standardized coefficients are presented. A constrained model was chosen. ***p < .001, **p < .01, *p < .05 Random intercept cross‐lagged panel model examining associations between socially avoidant problems and depressive symptoms. Dep, depressive symptoms; RI, random intercept; SA, socially avoidant problems. Standardized coefficients are presented. A constrained model was chosen. ***p < .001, **p < .01, *p < .05 Random intercept cross‐lagged panel model examining associations between overly nurturant problems and depressive symptoms. Dep, depressive symptoms; ON, overly nurturant problems; RI, random intercept. Standardized coefficients are presented. A constrained model was chosen. ***p < .001, **p < .01, *p < .05 Within‐person concurrent associations At Wave 12, significant within‐person associations with depressive symptoms were observed for socially avoidant, overly nurturant, and exploitable problems, but not for other domains (Figures 1, 2, 3, Figures S1–S5). Importantly, concurrent associations at Wave 12 represent baseline‐level deviations, uninfluenced by prior fluctuations. In contrast, associations at later waves reflect residual covariation embedded within ongoing temporal dynamics, as each variable's deviation has been shaped by autoregressive and cross‐lagged influences from prior waves. Across Waves 13–15, all interpersonal problem domains showed significant concurrent positive associations with depressive symptoms. Within‐person stability paths Positive carry‐over stability of depressive symptoms was observed across all models (Figures 1, 2, 3, Figures S1–S5). Stability was also observed for domineering, vindictive, and cold problems (Figure 1, Figures S1 and S2), and for nonassertive and exploitable problems from Wave 13 onward (Figures S4 and S5). In contrast, socially avoidant, overly nurturant, and intrusive problems showed no stability effects (Figures 2 and 3, Figure S3). Between‐person associations Significant between‐person associations were found in all models, indicating that girls with higher overall levels of interpersonal problems also reported greater depressive symptoms across time (Figures 1, 2, 3, Figures S1–S5). Discussion This study examined within‐person longitudinal associations between depressive symptoms and interpersonal problems across 4 years of adolescence in girls. Our hypotheses were partially supported. Specifically, we found unidirectional associations between depressive symptoms and subsequent interpersonal problems marked by social avoidance and a tendency to sacrifice one's own needs for others. Additionally, depressive symptoms showed bidirectional links with more domineering interpersonal tendencies, suggesting a potentially reinforcing cycle over time. These findings enhance understanding of intraindividual temporal dynamics between depressive symptoms and different facets of interpersonal problems, informing more precise behavioral targets for interventions aimed at reducing depression and interpersonal difficulties in adolescent girls, a population particularly sensitive to social rejection and at elevated risk for depression. Although the hypothesized bidirectional link was not observed, within‐person increases in depressive symptoms predicted subsequent increases in overly nurturant problems. This finding aligns with previous research showing that a history of depression is associated with high levels of sociotropy (Shih, 2006) and that depressive symptoms, combined with peer rejection, predict increases in reassurance‐seeking (Prinstein et al., 2005). One explanation is that depression‐related reductions in self‐esteem may heighten sensitivity to social evaluation, prompting efforts to please others for acceptance (Alarcón & Forbes, 2017). Relatedly, studies have found that depressed individuals exhibit greater prosocial behaviors, such as accepting unfair offers (Harlé, Allen, & Sanfey, 2010; McClure et al., 2007), ostensibly to keep in the good graces of others. Another explanation involves difficulties regulating negative affect in response to others' misfortune and maintaining interpersonal boundaries (Tone & Tully, 2014), consistent with evidence that affective empathy is positively associated with depression in adolescents (Yan, Zeng, Su, & Zhang, 2021). Depressive symptoms may also foster interpersonal guilt, a heightened sense of responsibility for others' suffering, which can lead individuals to prioritize others' needs over their own (Tone & Tully, 2014; O'Connor, Berry, Weiss, & Gilbert, 2002). Interestingly, overly nurturant problems did not predict subsequent increases in depressive symptoms, possibly because such tendencies may offer short‐term emotional relief by fulfilling affiliative needs. However, they may still carry emotional costs either by undermining long‐term well‐being not captured by our depressive symptom measure, or by exerting delayed effects beyond the 1‐year window. A similar pattern was observed for socially avoidant problems, with depressive symptoms predicting increases in socially avoidant problems, but not vice versa. This finding is consistent with prior work demonstrating prospective effects of depressive symptoms on later social avoidance during adolescence (Allen et al., 2014). However, it contrasts with studies that examined unidirectional prospective associations, in which earlier social avoidance prospectively predicted later depressive symptoms, and reported significant effects (Gazelle et al., 2024; Goodwin, Fergusson, & Horwood, 2004; Jakobsen, Horwood, & Fergusson, 2012; Smith et al., 2017). A potential explanation for this discrepancy lies in our focus on within‐person changes over time rather than between‐person prospective associations. Specifically, increases in social avoidance relative to an individual's own baseline may not significantly contribute to increases in depressive symptoms, whereas social avoidance relative to that of peers might play a larger role. Meanwhile, our within‐person analysis supported one side of the hypothesized bidirectional relation: increases in depressive symptoms relative to an individual's baseline predicted subsequent increases in socially avoidant problems. One explanation is that increases in depressive symptoms, accompanied by heightened social evaluative concerns (Alarcón et al., 2017; Woody et al., 2019; Rudolph & Conley, 2005; Yoon, Keenan, Hipwell, Forbes, & Guyer, 2023) and negative expectations about one's likability (Bone et al., 2021), may lead adolescent girls to become overly cautious in social interactions out of fear of rejection. In addition, depressive symptoms have been associated with reduced striatal activity in response to social acceptance (Olino, Silk, Osterritter, & Forbes, 2015) and subsequent social anhedonia (Stringaris et al., 2015). Such neurobiological changes may lead to lower motivation to engage socially. Notably, unlike most of the interpersonal problems examined, the stability paths for both overly nurturant and socially avoidant problems were not significant, indicating that these tendencies fluctuate within individuals across adolescence. This pattern suggests that such tendencies may be particularly malleable and sensitive to changes in emotional states. Relatedly, depressive symptoms demonstrated relatively strong autoregressive stability. Greater stability may reduce the residual variance available to be explained at subsequent time points, making it more difficult to detect cross‐lagged effects from interpersonal problems to depressive symptoms. This may have partially contributed to the absence of such effects for overly nurturant and socially avoidant problems. However, the presence of a significant effect from domineering problems to depressive symptoms suggests that stability alone does not fully account for the observed pattern of null cross‐lagged effects. Although not part of our a priori hypotheses, we found bidirectional associations between domineering problems (defined here as being too aggressive) and depressive symptoms. Prior studies on the longitudinal relations between aggression and depression using traditional cross‐lagged panel methods have yielded mixed results, with some reporting unidirectional effects from aggression to subsequent depressive symptoms (Blain‐Arcaro & Vaillancourt, 2017; Van Der Giessen et al., 2013) and others reporting reciprocal relations (Beyers & Loeber, 2003). Only one study (Yang et al., 2024) used a within‐person approach and reported null effects. Unlike this study, which focused on younger adolescents of both sexes and examined relational and physical aggression separately, our study focused on older female adolescents, for whom depression is typically elevated. We also used a broader measure of aggressive tendencies (e.g., “I am too aggressive toward other people”). These differences may have contributed to the observed cascading effects and highlight the need to examine moderators (e.g., gender, aggression type, social context). To explain the observed associations, domineering or aggressive behavior may lead to peer rejection (Agoston & Rudolph, 2013; Card & Little, 2006), especially when overt (Smith, Rose, & Schwartz‐Mette, 2010), which can worsen depressive symptoms. This “aggression‐rejection cycle” may be more pronounced in girls due to gender norms that expect girls to be more empathetic and prosocial (Rose & Rudolph, 2006), potentially amplifying risk for peer rejection. Additionally, depressive symptoms in adolescence are frequently accompanied by irritability and frustration (Wolff & Ollendick, 2006), emotional states that may increase the likelihood of hostile interactions, especially under provocation. Alternatively, domineering behavior may serve as a compensatory strategy in the context of depressive symptoms. Depressive symptoms are often characterized by reduced perceived control or feelings of helplessness (Nolen‐Hoeksema, Girgus, & Seligman, 1992), and domineering behavior may function as a means of regaining a sense of control. It may seem surprising that depressive symptoms predicted increases in both domineering and overly nurturant problems, given their characterization by high agency and high communion, respectively. We interpret this pattern as reflecting both contextual variability and heterogeneity in adolescents' interpersonal responses to depressive symptoms. For example, the same adolescent may respond to depressive symptoms in different ways depending on contextual moderators such as social context (e.g., the presence vs. absence of interpersonal provocation) or interaction partner (e.g., peers vs. family). At the same time, different adolescents may show distinct interpersonal responses to similar increases in depressive symptoms depending on between‐person moderators such as personality or temperament. Relatedly, domineering and overly nurturant problems were only modestly correlated in the present study (r = .20–.28; Tables S5–S8). Future research using multilevel frameworks, shorter assessment intervals, and measures of contextual and between‐person moderators may clarify whether these behaviors co‐occur within individuals across situations or reflect between‐person heterogeneity in responses to depressive symptoms, thereby informing more tailored, person‐centered intervention strategies. This study has several limitations. First, interpersonal problems were self‐reported, which may reflect negatively biased self‐perceptions. Adolescents with elevated depressive symptoms may perceive themselves and their relationships more negatively, which could have influenced their reports of interpersonal problems. Future studies could improve measurement validity by incorporating reports from parents, peers, or teachers, or by objective assessments such as observational methods or behavioral tasks. Second, the domineering subscale demonstrated only acceptable internal consistency (McDonald's ω = .65–.74), warranting replication with psychometrically stronger measures. Third, although depressive symptoms were scored following the original scale development, three items at Wave 12 used binary responses, whereas the same items were rated on a 4‐point scale at Waves 13–15. Although recoded to align with the 0–3 metric and capturing the same underlying depressive symptom constructs, this difference in response format at one of the four waves represents a limitation, and findings should be interpreted with this consideration in mind. Fourth, some participants reported zero interpersonal problems across all available waves, indicating no within‐person variability and thus not contributing to within‐person estimates. Accordingly, findings primarily reflect adolescents who exhibit within‐person variability in interpersonal problems and may not generalize to those with consistently no interpersonal problems. Future studies should extend these findings by examining within‐person temporal associations among depressive symptoms, interpersonal problems, and interpersonal dependent stress (e.g., peer rejection), which would provide a more direct test of stress generation and interpersonal theories of depression. In addition, because girls and boys may differ in interpersonal behaviors that contribute to depressive symptoms (e.g., behaviors that deviate from gender norms may elicit stronger negative social responses and subsequent distress) and in how they respond to depressive symptoms within social contexts, the findings may not generalize to boys. Future work should examine sex differences using multi‐group extensions of the RI‐CLPM. The current study has several strengths. Theoretically, it advances stress generation theory and the interpersonal theory of depression by clarifying the intraindividual temporal dynamics between depressive symptoms and interpersonal problems, and by identifying specific interpersonal dimensions most strongly implicated. Methodologically, the use of RI‐CLPM disentangles within‐person fluctuations from stable between‐person differences, enabling a focus on deviations from individuals' typical levels. These findings carry clinical implications. For instance, regularly monitoring changes in adolescents' depressive symptoms may help prevent increases in overly nurturant and socially avoidant tendencies that can subtly affect social interactions. In addition, the bidirectional link between depressive symptoms and domineering problems suggests a potential feedback loop, highlighting the value of early interventions targeting domineering behavior (e.g., interpersonal psychotherapy for adolescents) through communication and problem‐solving skills training (Young, Mufson, & Schueler, 2016). Conclusions In conclusion, this study provides evidence of intraindividual temporal associations between depressive symptoms and specific interpersonal problems in adolescent girls. Using a within‐person longitudinal approach, we identified that domineering problems show a reciprocal relation with depressive symptoms, while overly nurturant and socially avoidant problems tend to follow increases in depressive symptoms. These results suggest that monitoring increases in depressive symptoms relative to one's typical levels may provide opportunities for intervention to prevent downstream effects on social functioning, while early intervention targeting domineering behavior may help prevent self‐reinforcing cycles of depression and interpersonal problems. These findings underscore the value of separating within‐person fluctuations from between‐person differences when examining adolescent mental health. Ethical considerations This study involved secondary analyses of data from a larger longitudinal project approved by the University of Pittsburgh's Human Research Protection Office (IRB number: STUDY19040224; approved September 19, 2022). Written parental consent and adolescent assent were obtained. Key points What's known? Depressive symptoms and interpersonal problems are linked in adolescence, consistent with the interpersonal theory of depression and stress generation theory. Depressive symptoms and interpersonal problems are linked in adolescence, consistent with the interpersonal theory of depression and stress generation theory. What's new? This study models within‐person longitudinal associations between depressive symptoms and eight types of interpersonal problems over 4 years among adolescent girls, clarifying intraindividual temporal dynamics beyond stable between‐person differences. This study models within‐person longitudinal associations between depressive symptoms and eight types of interpersonal problems over 4 years among adolescent girls, clarifying intraindividual temporal dynamics beyond stable between‐person differences. Within‐person increases in depressive symptoms predicted subsequent increases in overly nurturant and socially avoidant problems, but not vice versa. Within‐person increases in depressive symptoms predicted subsequent increases in overly nurturant and socially avoidant problems, but not vice versa. Depressive symptoms and domineering problems showed reciprocal within‐person associations over time. Depressive symptoms and domineering problems showed reciprocal within‐person associations over time. What's relevant? Monitoring within‐person deviations in depressive symptoms or interpersonal problems—specifically domineering tendencies—may help detect early signs of socioemotional dysregulation and guide timely, personalized intervention. Monitoring within‐person deviations in depressive symptoms or interpersonal problems—specifically domineering tendencies—may help detect early signs of socioemotional dysregulation and guide timely, personalized intervention. Supporting information Table S1. Age at each wave. Table S2. Number and percentage of adolescents meeting symptom‐based criteria for major depressive disorder or minor depressive disorder at each wave. Table S3. Welch's t‐tests examining whether the missingness of a particular variable is associated with the average value of the same variable in previous waves. Table S4. Descriptive statistics for exploratory interpersonal problem variables. Table S5. Spearman correlations among study variables at Wave 12. Table S6. Spearman correlations among study variables at Wave 13. Table S7. Spearman correlations among study variables at Wave 14. Table S8. Spearman correlations among study variables at Wave 15. Figure S1. Random intercept cross‐lagged panel model examining associations between vindictive problems and depressive symptoms. Figure S2. Random intercept cross‐lagged panel model examining associations between cold problems and depressive symptoms. Figure S3. Random intercept cross‐lagged panel model examining associations between intrusive problems and depressive symptoms. Figure S4. Random intercept cross‐lagged panel model examining associations between nonassertive problems and depressive symptoms. Figure S5. Random intercept cross‐lagged panel model examining associations between exploitable problems and depressive symptoms. Acknowledgments This research was supported by NIH grants R01MH093650 (AEG, KK, EF), R01MH066167 (KK), R03MH116519 (AEG), and R01MH056630. The authors thank Marley Adler and Siri Boinapalli for assistance with data organization and intraclass correlation programming, respectively. The authors have declared that they have no competing or potential conflicts of interest. Conflict of interest statement: No conflicts declared. Data availability statement Data will be freely distributed following a data use request protocol to qualified academic investigators for noncommercial research. Data will be shared in a format that protects the anonymity and privacy of participants. Analysis code for the RI‐CLPM models is publicly available on the Open Science Framework at https://doi.org/10.17605/OSF.IO/V82Z9. References Associated Data Supplementary Materials Table S1. Age at each wave. Table S2. Number and percentage of adolescents meeting symptom‐based criteria for major depressive disorder or minor depressive disorder at each wave. Table S3. Welch's t‐tests examining whether the missingness of a particular variable is associated with the average value of the same variable in previous waves. Table S4. Descriptive statistics for exploratory interpersonal problem variables. Table S5. Spearman correlations among study variables at Wave 12. Table S6. Spearman correlations among study variables at Wave 13. Table S7. Spearman correlations among study variables at Wave 14. Table S8. Spearman correlations among study variables at Wave 15. Figure S1. Random intercept cross‐lagged panel model examining associations between vindictive problems and depressive symptoms. Figure S2. Random intercept cross‐lagged panel model examining associations between cold problems and depressive symptoms. Figure S3. Random intercept cross‐lagged panel model examining associations between intrusive problems and depressive symptoms. Figure S4. Random intercept cross‐lagged panel model examining associations between nonassertive problems and depressive symptoms. Figure S5. Random intercept cross‐lagged panel model examining associations between exploitable problems and depressive symptoms. Data Availability Statement Data will be freely distributed following a data use request protocol to qualified academic investigators for noncommercial research. Data will be shared in a format that protects the anonymity and privacy of participants. Analysis code for the RI‐CLPM models is publicly available on the Open Science Framework at https://doi.org/10.17605/OSF.IO/V82Z9.