Chinese university survey links supportive environments with student wellbeing

A cross-sectional survey of 815 higher education students in China found that perceptions of a supportive teaching environment were strongly associated with psychological wellbeing, alongside psychological safety, university belonging and learning engagement. The researchers estimated a standardized total association of 0.435 between the teaching environment and psychological wellbeing and a smaller serial indirect association of 0.036 for the prespecified sequence linking environment, safety, belonging, engagement and wellbeing. The serial estimate had a 95% bootstrap confidence interval from 0.026 to 0.047.

These results describe covariation in survey responses, not evidence that one factor causes another to develop or improve over time. The national online survey was conducted from October through December 2025 and covered higher education institutions in eastern, western, southern and northern regions. The sample included 595 undergraduates, 81 junior-college students, 118 master’s students and 21 doctoral students; 460 participants were female and 355 male.

Researchers used regional stratification to select 16 cities. Within each region, the selection probability for each city was calculated from a composite weight assigning equal weight to its share of the national resident population and its share of regular higher education institutions across the country. These same composite weights were then used to allocate sample-size quotas among the selected cities, with no city permitted to contribute more than 35% of the final sample.

Eligibility required participants to be at least 18, currently enrolled, to provide consent, to complete all 33 core questionnaire items and an attention check, and to give valid responses within the five-point agreement scale. The study separated exploration from confirmation by dividing the records with a fixed random seed: 408 were assigned to an exploratory factor-analysis group and 407 to an independent confirmatory factor-analysis group. Background characteristics were statistically similar between the two groups, with no significant differences across the eight tested distributions.

Exploratory principal-axis factoring with Promax rotation retained five factors corresponding to the teaching environment, psychological safety, university belonging, learning engagement and psychological wellbeing. Primary loadings ranged from 0.646 to 0.844, the largest cross-loading was 0.153, and the five factors explained 59.48% of common variance. The Kaiser-Meyer-Olkin measure was 0.949, while Bartlett’s test of sphericity was significant.

The instruments covered perceptions of teacher guidance, peer support, psychological services, institutional fairness and feedback channels for the teaching environment; safe expression and help-seeking for psychological safety; acceptance and membership for belonging; behavioral, emotional and cognitive participation for engagement; and self-acceptance, positive relations, autonomy, environmental mastery, purpose and personal growth for wellbeing. The five-factor measurement model fit the independent confirmation group well, with chi-square equal to 529.551 for 485 degrees of freedom, a comparative fit index of 0.994 and a root mean square error of approximation of 0.015. It performed better than models combining psychological safety with belonging, belonging with engagement, all three intermediate constructs, or all five variables into one factor.

Internal consistency was high for every scale, with Cronbach’s alpha and composite reliability above 0.88, while average variance extracted values exceeded 0.50. The maximum heterotrait-monotrait ratio was 0.617, supporting the argument that the constructs were related but statistically distinguishable. Configural, metric and scalar invariance tests also supported comparison across major demographic groups, although the wellbeing measure was a nine-item short form rather than a complete multidimensional scale for its six theoretical domains.

In the full-sample path model, the teaching environment had the strongest focal conditional association with psychological safety, with a standardized coefficient of 0.503 and a medium effect size of f² equal to 0.339. Students’ perceptions of teacher guidance, peer support, psychological services, institutional fairness and responsive feedback channels therefore covaried strongly with whether they felt able to ask questions, seek help, acknowledge difficulty or make mistakes without excessive concern about negative evaluation. Psychological safety was associated with university belonging at 0.397, belonging was associated with learning engagement at 0.455, and engagement was associated with psychological wellbeing at 0.395.

Each of these three paths was medium in magnitude according to the study’s effect-size criteria, but their cross-sectional nature prevents them from establishing a temporal sequence. Decomposition of the primary model yielded a total indirect association of 0.214, with a 95% bootstrap interval of 0.171 to 0.258, while the direct and indirect components were similar in size. The full four-step serial estimate of 0.036 represented about 8.26% of the total association, while all seven shorter and longer indirect estimates had bootstrap intervals excluding zero.

This pattern indicates that the survey data were compatible with multiple overlapping routes rather than one dominant chain. The largest specific indirect estimate was 0.050 for the route from the teaching environment through psychological safety to psychological wellbeing, followed by 0.038 through university belonging and learning engagement, 0.030 through psychological safety and learning engagement, and 0.028 through learning engagement alone. Students with positive environmental appraisals also tended to report more safety, stronger institutional membership, greater participation in learning and better overall psychological functioning, although some individual direct and indirect paths were small.

The researchers also compared all six possible orders for the three intermediate variables while holding the observed data, constraints and number of model parameters constant. The theoretically specified order—psychological safety followed by university belonging and then learning engagement—had the strongest relative fit, including a comparative fit index of 0.933 and the lowest Bayesian information criterion. Its advantage over the next-best order was a BIC difference of 210.352.

Nevertheless, its absolute fit remained inadequate, with chi-square divided by degrees of freedom of 17.741, a TLI of 0.866, an RMSEA of 0.143 and an SRMR of 0.111. All six candidate orders produced statistically significant full-chain product terms, ranging from 0.022 to 0.065, underscoring that a significant cross-sectional product cannot by itself identify the correct causal order. The theoretical order was therefore only relatively more compatible with the observed covariance structure, while the alternatives could not be ruled out from these data.

Results remained similar across demographic and academic subgroups, with full serial estimates ranging from 0.025 to 0.047 and every within-group bootstrap interval excluding zero. None of the eight tested between-group differences was significant, however, and all relevant difference intervals included zero. Measurement invariance supported the comparability of these analyses, but the study did not demonstrate that the association was stronger or weaker by gender, age, level of study, academic field, living arrangement, self-rated academic performance, academic stress or frequency of psychological distress.

Sensitivity analyses were also broadly stable: after adjustment for five demographic and educational variables, the total association was 0.429 and the serial estimate was 0.035; after an extended eight-variable adjustment, they fell to 0.366 and 0.022. Equal-weighted wellbeing scoring, rank transformation, a second bootstrap seed and exclusion of one quality-flagged record produced estimates close to the primary result. The extended adjustment also weakened two small paths that had reached the 0.05 significance level in the primary model.

The central limitation is the study’s cross-sectional design. Students reported their perceptions and psychological states during the same survey period, so the direction of influence cannot be determined, and reverse relationships remain possible: students with better psychological wellbeing may evaluate their institutions more favorably. The authors also noted that personality, socioeconomic conditions, prior achievement and pre-existing mental health could influence both environmental perceptions and wellbeing and were not fully accounted for.

Regional differences, examination periods, internship schedules and other contextual factors may also have affected responses, and voluntary participation did not guarantee complete population coverage. Future longitudinal follow-up, cross-lagged models and educational interventions would be needed to test whether changes in teaching conditions precede changes in safety, belonging, engagement and wellbeing. For now, the findings identify a measurable cluster of student experiences and a theoretically plausible structure for future testing, but they do not demonstrate that modifying the environment will necessarily produce the proposed psychological benefits.

Subject of Research: Cross-sectional associations among university teaching environments, psychological safety, belonging, learning engagement and wellbeing

Article Title: The predictive association between the teaching environment and psychological wellbeing: psychological safety, university belonging, and learning engagement

Article References: Jin, M., & Wang, Q. (2026). The predictive association between the teaching environment and psychological wellbeing: psychological safety, university belonging, and learning engagement. Frontiers in Psychology, 17, Article 1983869. https://doi.org/10.3389/fpsyg.2026.1983869

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1983869

Keywords: higher education, student wellbeing, psychological safety, university belonging, learning engagement, teaching environment, cross-sectional survey, China, structural equation modeling, serial association, predictive, association