Self-reported childhood family adversity is linked to an attenuated gain of trust during adolescence

The study investigates how trust develops from adolescence into young adulthood and whether early family experiences shape this developmental trajectory. Trust—defined as a willingness to be vulnerable to another’s actions—requires balancing risks and potential benefits. Adolescence is a key period for social development when other-oriented behaviors and the ability to form stable relationships mature. The authors hypothesize that self-reported childhood family experiences, especially parenting quality and adversity, influence intra-individual development of trust. They aim to (1) delineate normative longitudinal trajectories of trust behavior and underlying cognitive mechanisms, (2) explain inter-individual variability via family adversity, and (3) test whether laboratory-measured trust predicts subsequent real-life peer relationship quality.

Trust is commonly quantified via economic games, enabling characterization of risk–benefit trade-offs and cooperative behavior, and identification of neural correlates (e.g., anterior insula). Computational modeling of multi-round trust games implicates social preferences (risk and inequity aversion), planning, and theory of mind, and includes an ‘irritability’ parameter capturing retaliatory impulses to perceived unfairness. Prior cross-sectional studies on adolescent trust yield mixed findings: some report age-related increases, others no differences, and some decreases, likely reflecting inter-individual variability not captured by cross-sectional designs. Developmental psychopathology proposes that earlier family experiences shape later social relationships, suggesting early adversity may alter trust development. The literature also debates whether adolescent risk-taking differs from other ages in non-social contexts; thus, specificity of social risk in trust requires testing.

Design and sample: Longitudinal cohort from the NSPN project including n = 570 community participants (284 female, 286 male by self-report) aged 14.10–24.99 at baseline (mean 19.05, SD 2.96), with follow-up at mean 1.48 years later (range 0.99–2.6). A subsample (n = 55) completed an additional 6-month retest. Ethics approvals and informed consent obtained. Tasks: (1) Multi-round Trust Game (10 rounds): participants were investors receiving 20 coins per round, choosing an investment (tripled by experimenter) to a trustee (computer algorithm emulating human trustees) who repaid an amount; feedback shown each round. Measures: a priori trust (round 1 investment), mean trust (round-by-round investment across 10 rounds), and reciprocity (relative change in investment following trustee repayment changes). (2) Non-social risk task (baseline only): roulette-type gambling task estimating conventional risk preferences (variance and skewness). Family experiences: Retrospective self-report at baseline via Alabama Parenting Questionnaire (APQ) and Measure of Parenting Styles (MOPS). Principal component analysis on standardized subscale scores yielded a composite family experience factor (higher scores = greater family adversity). Questionnaire timing preceded T1 task on average by 0.43 years (SD 0.35). Peer relations: Cambridge Friendship Questionnaire (CFQ) administered three times (around T1 [T1q], ~1 year later [T2q; mean 1.13 years], and ~2.3 years later [T3q; mean 2.26 years]) to assess peer relationship quality. Statistical analyses: Linear mixed-effects models (R, afex) with Satterthwaite df approximations. Age decomposed into cross-sectional (between-subject) and longitudinal (within-subject) components. Models included fixed effects for cross-sectional age, longitudinal age, their interaction, and sex as covariate; random intercepts and subject-specific slopes. For round-by-round investments, trial number was included. Family adversity factor and its interactions added to assess moderation effects. For CFQ analyses, baseline trust metrics (T1) were used as predictors along with age components at questionnaire times, sex, and interactions; family adversity added as moderator. Multiple comparisons for computational parameters corrected via Bonferroni (7 parameters; alpha = 0.007). Computational modeling: Interactive partially observable Markov decision process (I-POMDP) for the multi-round trust task with seven parameters: inequality aversion (α), social risk aversion (ω), theory of mind sophistication (k), planning horizon (P), irritability (q(ζ)), irritation awareness (ζ), and choice stochasticity (β). Parameters estimated via full grid search minimizing negative log-likelihood of observed choices. Model comparison using Draper BIC across full model and reduced variants confirmed the full 7-parameter model as best (lowest BIC). Posterior predictive simulations conducted to validate parameter-behavior links. Controls: Additional analyses included non-social risk measures as covariates, and checks with ordinal mixed models for discretized parameters. IQ and socioeconomic status included in supplementary control analyses (no qualitative changes).

- Longitudinal increase in trust: - A priori trust (round 1 investment): significant cross-sectional age effect (F(1,567)=19.10, p<0.001, estimate=0.26, se=0.06) and longitudinal effect (F(1,568)=19.45, p<0.001, estimate=0.67, se=0.15); no interaction (p=0.572). - Mean trust (round-by-round investments): significant cross-sectional age (F(1,567)=24.32, p<0.001, estimate=0.58, se=0.06) and longitudinal development effects (F(1,10819)=109.81, p<0.001, estimate=0.27, se=0.05); significant interaction (F(1,10819)=15.14, p<0.001, estimate=-0.07, se=0.02) indicating steeper longitudinal increases in younger adolescents and flattening with age. - Retest subsample (n=55): time-point effect on investments (F(2,1418)=4.77, p=0.009); significant increases from 6 to 18 months (t=2.61, p=0.009) and from baseline to 18 months (t=2.74, p=0.006), but not from baseline to 6 months (t=0.13, p=0.90), supporting developmental rather than retest effects. - Reciprocity: no significant cross-sectional or longitudinal effects (all Fs<0.086, ps>0.05). - Computational mechanism: social risk aversion decreased with age and over time: - Social risk aversion showed significant cross-sectional age effect (F(1,566.92)=20.50, p<0.001, estimate=-0.21, se=0.004) and longitudinal effect (F(1,567.10)=20.49, p<0.001, estimate=-0.055, se=0.01); no interaction (p=0.125). - Higher social risk aversion predicted lower total earnings at T1 (t(785)=-25.65, p<0.001, r=-0.68) and follow-up (t(567)=-26.85, p<0.001, r=-0.75). - No age-related changes in non-social risk-taking; age effect on social risk aversion remained significant controlling for gambling measures (t=-5.180, p<0.001, estimate=-0.024, se=0.005). - Retest subsample: social risk aversion decreased over 18 months (F(2,106)=4.11, p=0.02; baseline vs 18 months t=2.86, p=0.005) but not over 6-month intervals. - Family adversity moderates trust development: - Significant interaction between family adversity and longitudinal development on mean trust (F(1,9696)=6.83, p=0.009, estimate=-0.03, se=0.01): greater adversity associated with a flatter increase in trust over time; no main effect of adversity (p=0.153) and no adversity×mean age interaction (p=0.211). - No adversity×longitudinal interaction for a priori trust (round 1) (F(1,507)=1.83, p=0.177). - Computational basis of adversity effect: - Family adversity positively associated with Irritability parameter (F(1,506)=7.92, p=0.005, estimate=0.006, se=0.002; Bonferroni-corrected), indicating greater propensity for unmentalized retaliation after below-expectation partner actions. - Behavioral marker of retaliation aligned with modeling: greater adversity linked to greater retaliation (F(1,540.15)=3.98, p=0.047, estimate=-0.06, se=0.03). - Uncorrected evidence for adversity×longitudinal development interaction on social risk aversion (F(1,507)=4.44, p=0.036), not surviving Bonferroni correction. - Trust predicts peer relationship quality and resilience: - Family adversity predicted lower CFQ at baseline (T1q: t(377)=-5.863, r=-0.289, p<0.001) and at ~2.3-year follow-up (T3q: t(377)=-3.305, r=-0.168, p=0.001). - CFQ increased over time (F(2,1221.62)=24.70, p<0.001; T1 to T3 contrast=1.06, se=0.15). - Higher a priori trust at T1 predicted better peer relations across time (F(1,757.74)=5.27, p=0.022, estimate=0.06, se=0.02); mean investments showed a trend (F(1,759.72)=3.10, p=0.069). - Significant interaction: a priori trust × adversity × longitudinal development (F(1,1050.70)=6.77, p=0.009, estimate=0.05, se=0.01): adolescents with higher adversity showed the strongest positive association between baseline a priori trust and gains in peer relation quality, suggesting a resilience role of trust.

The study confirms that trust increases from adolescence to young adulthood and, crucially, demonstrates within-person developmental gains over approximately 1.5 years. Computational modeling indicates that a decrease in social risk aversion underlies this maturation, aligning with the idea that adolescents become less averse to social risks in cooperative exchanges, thereby improving outcomes. However, inter-individual variability is substantial. Consistent with developmental psychopathology frameworks, retrospective self-reported childhood family adversity attenuated the longitudinal increase in trust. Modeling linked adversity to heightened irritability—a propensity for unplanned, punitive responses to perceived unfairness—suggesting a mechanism by which adverse early experiences may foster hostile interpretations and retaliation that undermine cooperative development. Furthermore, laboratory-measured unconditional trust predicted improvements in real-life peer relationship quality, particularly among those with greater adversity, implying that trust functions as a resilience factor facilitating adaptive social integration despite negative early experiences. These findings refine our understanding of adolescent social development by distinguishing social-specific risk processes from non-social risk-taking and by identifying individual difference factors shaping trajectories.

This work shows that trust, as measured in an iterated economic game, increases from adolescence into young adulthood, driven by a developmental decrease in social risk aversion that enhances cooperative outcomes. Early family adversity moderates this trajectory, attenuating trust gains and associating with a higher propensity for unmentalized retaliatory responses (irritability). Importantly, baseline unconditional trust predicts subsequent improvements in peer relations over up to three years, especially for youths with higher adversity, indicating a resilience role for trust. Future research should contrast trust toward familiar versus unfamiliar partners, examine mechanisms of hostile attribution in adolescents with adversity, and explore clinical implications (e.g., borderline personality features) and targeted interventions that reduce social risk aversion or irritability to support social relationship development.

Key limitations include reliance on retrospective, self-reported measures of childhood family experiences, which are subject to recall biases. The sample predominantly reported low levels of negative family experiences, potentially yielding modest effect sizes and limiting generalizability to more severely affected populations. Additionally, computational parameter estimation used discretized grids for tractability, which may limit precision; ongoing work aims to relax this constraint.