Emotional adjustment and peer relationships: The role of behavioural reputation and classroom social climate

PLOS One | https://doi.org/10.1371/journal.pone.0332297 September 18, 2025 5 / 13 fit, Χ2(200) = 1092.23, p < .001, CFI = .91, RMSEA = .06, SRMR = .04. Note that covariances were specified within two sets of context-specific items (feeling angry, worried, and sad at home; and feeling angry, worried and sad at school). Measurement invariance across gender was also assessed: configural invariance was indicated by satisfactory model fit, Χ2(400) = 1370.31, p < .001, CFI = .91, RMSEA = .06, SRMR = .04, with no significant deterioration when equal loadings were specified, △Χ2(19) = 14.13, p = .776. Given that our planned analysis involved use of composite scale scores for emotional adjustment, each child received a mean score across all items within each emotion subscale (possible range 0–3). In addition, a classroom social climate score was created for each class by subtracting the class’s mean score for negative/victimising peer experiences from the class’s mean score for positive/prosocial peer experience scores (possible range −3–3); higher scores thus indicate a greater relative frequency of positive over negative peer experiences. Preliminary analysis showed that the mean across all 60 classrooms was 1.15 (SD = .36, range .45 to 1.79), and that the scores were not significantly correlated with year group or with the percentage of girls in the class, rs < .13, ps > .05. For multigroup analyses to evaluate the moderating effect of classroom social climate, we compared the 20 classrooms (from 14 schools) displaying the highest difference

between reported positive and negative peer experiences (i.e., positive experiences were considerably more frequent than negative experiences, M = 1.54, SD = .12) with the 20 classrooms (from 12 schools) showing the least difference (M = .75, SD = .18, t(38) = 16.16, p < .001). Behavioural reputation and social acceptance. Based on established approaches to sociometric measurement [21], children also completed an online sociometric survey, involving nominations of up to three classmates for 7 items, including brief behavioural descriptors for the reputation items: most like to spend free time with; least like to spend free time with; cooperative (help other people, share, and take turns); disruptive (not very good at being in a group, and don’t listen to other people); shy (really quiet and always seem to be on their own); starts fights (say mean things to other people, or push or hit them); and leader (often seem to be in charge of things, and good at deciding what to do). The total number of nominations received by each child was tallied and standardised within classroom, i.e., converted into z scores with a mean of 0 and SD of 1 for each classroom. The difference in ‘most-like’ and ‘least-like’ nominations received by each child constituted that child’s social preference scores [21]. Procedure Following ethics approval from the University of Sussex, School of Life Sciences Research Governance Committee (approvals RB0109 and RB0310), head teachers provided written informed consent to take part in the research. Note

that parents received information letters and forms to return if they wished to withdraw their child from participating in the study. Fewer than 5% of participants were withdrawn from the survey administration. In addition, assent was gained from children to complete the two online surveys in their schools’ computer laboratories in 15-minute sessions supervised by school staff. The surveys were completed independently (with audio available for children with literacy needs) and ques- tions were presented one at a time. Children identified themselves and made peer nominations using code numbers. The dataset for our analyses was derived from projects in three UK local authorities where these data were collected in the school year from 1 September 2010–31 July 2011. Surveys were completed once during the first half of the school year (approximately 4 months after the start of the school year), and again at the end of the school year (approximately 6 months after the first time point). Results Table 1 shows the inter-correlations for all measures. Using the Lavaan library in R, with Full Information Maximum Likelihood for handling missing data in all analyses, an initial fully saturated structural equation model evaluated path- ways from all emotional adjustment and behavioural reputation measures at Time 1 to social preference at Time 2, after controlling for social preference at Time 1. Pathways were also included from all emotional adjustment measures,