By presenting the inaugural Kazakh pilot data on psychological safety in secondary classrooms and combining student narratives with essential statistics, this study offers (i) a preliminarily adapted set of local-language measures with initial internal-consistency evidence, (ii) baseline indicators for educational leaders, and (iii) practical recommendations aligned with AJUE’s scholarship agenda centered on student-focused learning. The findings will contribute to the extensive data collection planned for 2025–2026 and will inform the development of modules for teacher training required by the national framework program.

In classroom contexts, psychological safety refers to students’ shared perceptions that interpersonal risk-taking (e.g., asking questions, admitting confusion) will not lead to ridicule or retribution (Chen et al., 2022; Duffy et al., 2022). A later integrative review clarified its historical foundations, outlined its antecedents (supportive leadership, inclusive norms), and established links between this construct and learning behaviours in various contexts (Edmondson and Lei, 2014; Lackie et al., 2023). Although much of this research emerged from organisational studies, educational researchers have embraced the construct, acknowledging that classrooms also require collective participation and the readiness to admit mistakes to sustain engagement and learning.

The seven-item Team Psychological Safety Scale, introduced by Edmondson (1999), remains the standard instrument in this field and has been readily adapted for educational research, including studies involving K-12 staff and university undergraduates (Chin et al., 2024). Recent work also highlights the value of measuring adolescents’ perceived safety directly, including multi-item instruments tested in adolescent samples (Cogan et al., 2025; Cogan et al., 2024). Together, these developments support classroom research that links interpersonal climate to well-being while underscoring the need for careful adaptation when instruments are used in new languages and contexts.

Fleming et al. (2024) tracked 769 rural educators in the United States over a four-year period. They discovered that a “stable-high” safety profile was correlated with lower levels of burnout and greater self-efficacy compared to “dynamic-low” or “medium” profiles.

At the student level, emerging findings from Asia are significant. A Chinese survey involving 1,267 junior high students demonstrated that a positive school climate enhanced psychological capital, which in turn mitigated academic burnout, partially mediating the impact of school climate. In India, a comparative study of 636 higher secondary students identified significant variations in psychological well-being correlated with school type, emphasising the importance of contextual climate in adolescent mental health (Raghu and Sultana, 2024).

In the realm of higher education, a study published in AJUE regarding Malaysian undergraduates demonstrated that the fulfilment of basic psychological needs and academic motivation fully mediated the relationship between academic support and student engagement, thereby reinforcing the pathway from a supportive environment to positive outcomes (Chin et al., 2024). Collectively, these studies reveal consistent associations between safe environments and overall well-being. However, few specifically address upper secondary student populations. Russian-language evidence supports this mechanism: higher anxiety is linked to lower satisfaction with the educational environment and lower perceived psychological safety, expressed as fear of knowledge testing/social stress (girls) and teacher-related fears/self-expression difficulties (boys) (Ivanova and Stepanova, 2024). In a large vocational-student sample (n = 2,789; ages 12–25), higher psychological security was associated with higher social intelligence, especially understanding/predicting others and stronger social skills (Baeva et al., 2021).

Across jurisdictions, adolescent well-being and classroom climate have become explicit policy concerns, with school systems increasingly emphasising prevention, support structures, and teacher capacity-building to reduce student distress and sustain engagement (Adams, 2025; Tang, 2024). This policy turn reinforces the importance of generating context-specific evidence on classroom psychological safety, particularly in upper-secondary grades where academic demands intensify.

Kazakhstan’s State Programme for the Development of Education and Science 2020–2025 commits to establishing “safe and comfortable” learning conditions for every child while acknowledging that the national response system for students in challenging situations remains inadequately developed. A review of peer-reviewed literature indicates that this concept, although explored at various educational levels, has yet to be quantified in Kazakh secondary classrooms. Existing studies predominantly focus on elementary pupils (Mukhtarkyzy et al., 2025; Sarzhanova et al., 2023; Tusseyev et al., 2021) or the adaptation processes of university students (Litvinov et al., 2020; Tasmagambetova et al., 2024). Methodologically, mixed-method survey + interview designs are increasingly used in Kazakhstan to evaluate education reforms (e.g., a 24-student mixed-method study on dual education and a specialised teacher-training platform), supporting the feasibility of explanatory mixed-method approaches in local settings (Seitaliyeva et al., 2025). Furthermore, no published research has been identified that assesses psychological safety among Kazakh Grade 10–11 cohorts or evaluates its association with well-being, thus revealing a significant empirical gap that the present pilot study seeks to address.

Across various regions and educational levels, psychological safety consistently demonstrates a correlation with increased engagement, reduced burnout, and improved mental health. However, the extant literature tends to exhibit a bias towards higher education and samples involving educators, with Central Asian secondary schools being nearly absent from the discourse. By conducting a pilot mixed-methods study in four schools within Semey City, the present research addresses three significant gaps: (i) it offers local-language measures that have been preliminarily adapted for use with adolescents, (ii) it establishes baseline safety and well-being indicators for students in Grades 10–11, and (iii) it presents contextually rich qualitative data aimed at informing future national reforms.

The researcher employed a mixed-methods approach distinguished by its explanatory and sequential nature, specifically devised to address the three research questions (RQ1–RQ3) along with their corresponding hypotheses (H1–H3). The study used a cross-sectional, explanatory-sequential mixed-methods design: a brief questionnaire provided the primary (quantitative) evidence, and qualitative data were collected to explain and contextualise the observed patterns. A succinct student questionnaire constituted the foundation for the quantitative data, while focus groups and interviews conducted with educators facilitated qualitative insights. The quantitative components focused on RQ1 (baseline levels), RQ2 (differences based on grade and gender), and RQ3 (associational) modelling. The qualitative aspects enriched the interpretation of any discerned numerical patterns rather than serving as a parallel test of hypotheses.

All eligible Grades 10 and 11 students in the selected classes were invited to participate; 111 students participated in the study (N = 111; 53 tenth-graders and 58 eleventh-graders). Grade 10 students (n = 53) had a mean age of M = 16.04, SD = 1.07, and 39.9% identified as female; Grade 11 students (n = 58) had a mean age of M = 16.91, SD = 1.38, and 44.8% identified as female. The distribution of locations is presented in Table 1. Six subject teachers, comprising four females and two males, instructed the surveyed classes and participated in individual interviews. Given the pilot scope and sample size, the study is positioned as exploratory and intended to inform a larger planned data collection.

Questionnaires were screened prior to analysis. After screening, 111 student questionnaires were retained for quantitative analyses (0 excluded). For scale scoring and regression, missing item-level data were handled using listwise deletion at the analysis level (i.e., cases with missing values on variables required for a given model were excluded from that model).

The data collection was conducted from April 3 to April 10, 2025. Researchers distributed paper questionnaires during one lesson period in each class, with an average completion time of 5 min. Subsequently, voluntary focus group discussions comprising six to eight students were convened in a quiet room. Focus groups were conducted with students from the same classes that completed the questionnaires (i.e., the qualitative sample was drawn from the surveyed cohort). Focus groups were held, each lasting approximately 25 min, guided by prompts on classroom situations that increased safety, teacher responses to mistakes/questions, and peer interaction norms. Teacher interviews were arranged to take place after school hours on the same day. All sessions were audio recorded and transcribed verbatim.

The Research Ethics Committee at Shakarim University has approved the study. Participation in the study was entirely voluntary. Students aged 16 years and older provided written consent, whereas students under the age of 16 utilised a parental opt-out letter. No identifying information was collected during the study, and all data files were stored using encryption methods to ensure confidentiality and data security.

For Research Question 1 (RQ1) and Hypothesis 1 (H1), descriptive statistics and scale reliability were computed using SPSS version 29.0. Research Question 2 (RQ2) and Hypothesis 2 were examined using independent-samples t-tests to compare psychological safety and classroom well-being across grade and gender groups. Research Question 3 (RQ3) and Hypothesis 3 were examined using hierarchical linear regression, entering grade and gender in Step 1 and psychological safety in Step 2. Regression results are reported with regression coefficients, standard errors, and overall model fit indices including R² and the F-statistic. Two researchers independently coded qualitative transcripts using Braun and Clarke’s six-phase thematic analysis, and inter-coder agreement exceeded 0.75. Transcripts were segmented into meaning units and labelled by participant type (student/teacher) and grade (G10/G11) to support pattern comparisons. Coding discrepancies were resolved through discussion, and an audit trail (codebook + decision notes) was maintained.

The Cronbach’s alpha coefficients demonstrated strong internal consistency for both scales, with values of 0.84 for Psychological Safety and 0.81 for Classroom Well-being. As shown in Table 2, the overall mean for psychological safety (M = 3.02, SD = 0.56) exceeded the neutral midpoint, thus providing partial support for Hypothesis 1. The mean score for classroom well-being was 3.18 (SD = 0.60), indicating a moderately more positive emotional environment.

The grade comparison directly evaluated Hypothesis 2, which predicted a decline in psychological safety from Grade 10 to Grade 11. Tenth-grade students reported a higher mean safety score (M = 3.15, SD = 0.55) compared to their eleventh-grade counterparts (M = 2.91, SD = 0.53); this difference was statistically significant, t(109) = 2.20, p = 0.030, exhibiting a small-to-moderate practical effect size (Cohen’s d = 0.42). Classroom well-being scores were also slightly lower in Grade 11 (Grade 10: M = 3.22, SD = 0.61; Grade 11: M = 3.14, SD = 0.59), and this grade difference reached statistical significance (p = 0.035), although the magnitude of the difference was small. Gender did not exert a reliable influence on either construct. Consequently, the results support Hypothesis 2 for psychological safety and indicate a small grade-linked difference in well-being, suggesting that the transition to the high-stakes examination year is associated with reduced interpersonal safety and a modest decline in reported classroom well-being.

To examine whether psychological safety predicts classroom well-being beyond demographic variables, a hierarchical regression analysis was conducted. As shown in Table 3, the initial model (Step 1) including only grade and gender accounted for 5% of the variance in well-being scores [F(2, 108) = 2.84, p = 0.063]. Grade showed a modest negative effect (β = −0.21, p = 0.035), suggesting lower well-being in Grade 11; gender had no significant impact. When psychological safety was added in Step 2 (Table 4), it significantly improved the model, contributing an additional 26% of explained variance and raising the total adjusted R² to 0.29 [F(3, 107) = 16.03, p < 0.001; ΔF(1, 107) = 40.33, p < 0.001]. Psychological safety emerged as a strong positive associate (β = 0.55, p < 0.001), while the effects of grade and gender became statistically non-significant. This provides exploratory support for Hypothesis 3, indicating that students who feel psychologically safe are more likely to report higher levels of classroom well-being—regardless of their grade or gender. At the same time, the association should be interpreted cautiously given the pilot sample and the dispersion of individual scores, suggesting that additional unmeasured classroom factors also contribute to well-being, which is illustrated in Figure 1, where individual student scores show a positive association with considerable variability between psychological safety and well-being across the sample.

Four student focus groups, comprising 27 participants, and six teacher interviews resulted in the identification of 142 meaning units that coalesced around four themes, elucidating the quantitative patterns, as presented in Table 5. Teacher approachability and norms of peer respect were consistently recognised as facilitators of safety. Conversely, fear of public evaluation and time constraints emerged as significant detractors, particularly in Grade 11 classes preparing for high-stakes examinations. For example, Grade 11 students frequently described hesitation to answer in front of peers (e.g., “If I answer incorrectly, others will laugh”), while time pressure was linked to reduced opportunities to ask clarifying questions (e.g., “We have to rush, so there’s no time to explain mistakes”). These narratives enhance the understanding of the decline associated with grade levels and further substantiate the associational significance of safety within the regression model.

Theme 1 – Teacher approachability: Students described feeling safer when teachers normalised mistakes (e.g., using supportive humour and “mistakes are normal” statements), which reduced hesitation to ask questions in whole-class settings. This theme appeared in both grades, but was described as harder to sustain in Grade 11 under faster pacing.

Theme 2 – Peer-respect norms: Students highlighted explicit “no mocking” norms as a protective factor, often framed as a shared class agreement that made speaking up less risky. Where these norms were absent or inconsistently enforced, participants linked this to avoidance of participation (especially during public answering).

Theme 3 – Fear of public evaluation: Participants connected anxiety to highly visible performance moments (e.g., answering at the board or being singled out), describing freezing and fear of ridicule; this theme was emphasised in Grade 11 narratives and aligns with the observed grade-linked drop in safety.

Theme 4 – Physical and time pressure: Time scarcity (short transitions, rushed explanations, rapid questioning) was reported as limiting clarification opportunities and increasing tension; teachers also noted that faster pacing can unintentionally silence quieter students.

Together, these theme-specific accounts clarify how classroom routines (error response, peer norms, public evaluation, pacing) plausibly produce the observed grade-linked safety differences and the safety–well-being association.

Collectively, the data provide coherent responses to the three guiding questions. The baseline safety level is moderate, yet fragile, deteriorating during the examination-intensive final year. This decline coincides with an increase in qualitatively identified fears concerning public mistakes. Psychological safety emerges as the strongest statistical correlate of classroom well-being, constituting nearly one-third of its variance. Qualitative evidence supports this finding, indicating that when educators normalise errors and peers alleviate ridicule, students experience emotional security, even amidst the pressures of the curriculum.

The first research question inquired whether upper-secondary students experienced a psychologically safe and emotionally supportive classroom environment. Mean scores that slightly exceeded the neutral midpoint indicated only moderate levels of safety and well-being, thereby providing partial support for hypothesis H1. These scores were consistent with the “middling” patterns observed in recent Asian samples (Chin et al., 2024). They were lower than those reported in earlier studies based on Edmondson’s framework, which often concentrated on lower-stakes contexts. The qualitative theme of teacher approachability played a crucial role in elucidating these findings. Students expressed a sense of security when errors were normalised and humour was permitted. However, such conducive conditions were inconsistently manifested across various lessons. Overall, a foundational level of safety was present but proved to be fragile. It was adequate for routine interactions but insufficient for sustained intellectual risk-taking in whole-class settings.

In alignment with H2, Grade 10 students reported a significantly higher level of psychological safety compared to their Grade 11 counterparts; however, their overall well-being did not exhibit any significant difference. This observed grade effect is consistent with prior data from England, which indicates a decline in perceived safety following the transition to high-stakes examinations (UCL, 2025). Insights from focus group discussions identified two immediate factors contributing to this decline: increased public assessments and shorter transitions between lessons. For eleventh-grade students, the Blackboard test emerged as a symbolic stressor, described as “the moment everyone is watching,” thereby magnifying their fear of embarrassment. Educators acknowledged that rapid questioning tends to “shut quieter students down,” thereby supporting the connection between evaluative practices and perceived risk.

The prominence of grade differences—relative to gender (and the limited age variation within grades)—suggests that institutional conditions may matter more than stable individual characteristics in this context. In Kazakhstan, Grade 11 typically concentrates examination preparation and increases evaluative intensity; thus, the grade effect can be interpreted as reflecting a shift in classroom routines (assessment visibility, pacing, and error tolerance) rather than a developmental change alone. At the same time, developmental explanations cannot be ruled out, as older adolescents may be more sensitive to peer evaluation; however, the narrow age range and the alignment of age with grade make institutional factors the more plausible proximal explanation in this pilot sample. The lack of observable gender differences implies that examination-related pressure diminished safety broadly rather than selectively, although the modest sample size may have limited power to detect small gender effects. Together, these findings point to grade-linked instructional and assessment practices—rather than demographic characteristics—as practical targets for improving classroom psychological safety.

Hierarchical regression provided exploratory support for H3 after controlling for grade and gender. Psychological safety accounted for an additional 24% of the variance in classroom well-being, rendering demographic effects non-significant. This pattern aligns with previous findings indicating that supportive environments mitigate burnout and enhance engagement (Edmondson, 1999). However, the association should be interpreted cautiously: this was a cross-sectional pilot, and the scatter in individual scores indicates meaningful unexplained variability, suggesting that other classroom factors also contribute to well-being. Qualitative data elucidated plausible classroom mechanisms that may underlie this association. When peers enforced norms of courtesy and teachers encouraged the expression of divergent views, students reported feeling “relaxed” and “heard,” experiences that correspond directly with the well-being scale items. In contrast, public shaming and time pressure diminished both safety and comfort. Thus, the convergence of mixed-methods evidence emphasises safety not merely as a peripheral nicety but rather as a salient correlate of adolescents’ affective states in upper-secondary classrooms, while underscoring the value of qualitative accounts for interpreting the quantitative pattern.

The findings derived from all three questions converge on a salient message: instructional interaction style and assessment format serve as critical levers for fostering psychological safety within Kazakh upper-secondary classrooms. Professional development workshops should prioritise discourse routines that accommodate errors, promote student questioning, and institutionalise peer-respect charters. School leaders may undertake an audit of high-stakes oral testing practices, replacing private or small-group demonstrations wherever feasible. At the policy level, the national directive for “modern learning environments” could integrate explicit safety indicators, including types of teacher prompts, student participation ratios, and incidences of ridicule, to enhance existing physical environment benchmarks.

The cross-sectional design precluded causal inference and obscured within-student trajectories. Although the sample encompassed four schools, it represented a single urban district. Rural or highly selective schools may differ in their approaches. Self-report measures might inflate associations due to common-method variance. However, the large effect size and qualitative corroboration alleviate this concern. Finally, teacher interviews involved volunteers and may underrepresent less reflective practitioners.