Gendered Pathways to Career Exploration and Academic Persistence Among STEM Undergraduates in South Korea

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

General Comments The manuscript provides a timely and highly relevant investigation into the gendered structural mechanisms underlying STEM undergraduates' career exploration and academic persistence intentions. By applying the Social Cognitive Career Theory (SCCT) to a large, representative sample (n=2,393) within the South Korean educational context, the authors make a valuable contribution to a literature largely dominated by Western perspectives. The finding that career exploration negatively predicts persistence intentions among female students is particularly intriguing and adds significant nuance to traditional SCCT assumptions. Overall, the paper is well-written, theoretically sound, and methodologically transparent.

Major Comments (Methodology & Results)

• Justification of Model Fit Indices: The structural models exhibit marginal absolute fit indices (e.g., RMSEA = 0.112 for female students, 0.121 for male students). The manuscript's explanation—attributing these values to the inherent complexity of the model with numerous latent indicators (k=71)—is well-received, and the structural paths remain theoretically coherent. To further bulletproof this acceptable justification for the broader readership, the authors might consider adding a brief citation to methodological literature that explicitly discusses how RMSEA can be inflated in highly complex models. No further re-specification of the model is necessary.

• Interpretation of Suppression Effects: We positively value the authors' caution in interpreting the suppression effects observed within the model (specifically, the direct path from career barriers to self-efficacy becoming positive, and the path from career exploration to persistence intentions becoming negative in the multivariate SEM). Exercising this level of methodological prudence when dealing with sign reversals in complex structural models is highly commendable and strengthens the overall trustworthiness of the analysis.

Minor Comments (Discussion & Implications)

• Contextualizing Practical Implications: The practical implications derived from the study are sound, particularly the suggestion that institutions should offer "structured advising" and "moderated career-information workshops" to prevent career exploration from deterring female students. To increase the impact of this section, the authors could provide 1-2 brief examples of what these interventions might look like specifically within the South Korean higher education context or leveraging existing national agencies like WISET.

• Future Research Directions: In the limitations section, the authors correctly point out that the cross-sectional design limits causal inference. They could briefly expand on this by suggesting that future longitudinal research should specifically track the timeline of female students' career exploration to pinpoint exactly when and how the feelings of "uncertainty" or "constraint" emerge during their undergraduate years.

• Table Formatting: Please review the formatting of Table 2. Given the large number of variables, ensuring clear spacing and alignment of the sub-variables will improve readability for the final published version.

Author Response

Responses to reviewer 1

We sincerely appreciate the reviewers' valuable comments and have thoroughly revised the manuscript to address their feedback to the best of our ability. Responses to the reviewers' comments are provided, and the revised sections of the manuscript are highlighted in yellow.

1. Justification of Model Fit Indices: The structural models exhibit marginal absolute fit indices (e.g., RMSEA = 0.112 for female students, 0.121 for male students). The manuscript's explanation—attributing these values to the inherent complexity of the model with numerous latent indicators (k=71)—is well-received, and the structural paths remain theoretically coherent. To further bulletproof this acceptable justification for the broader readership, the authors might consider adding a brief citation to methodological literature that explicitly discusses how RMSEA can be inflated in highly complex models. No further re-specification of the model is necessary.

Response 1: We appreciate the reviewer’s positive evaluation of our explanation regarding the marginal RMSEA values. To further strengthen the methodological justification, we have added an explicit citation to methodological literature discussing the inflation of RMSEA in complex models with large numbers of indicators and degrees of freedom. This addition reinforces the appropriateness of our model evaluation and enhances transparency for a broader readership. The relevant discussion has been revised in the Results (Section 4.3). 

2. Interpretation of Suppression Effects: We positively value the authors' caution in interpreting the suppression effects observed within the model (specifically, the direct path from career barriers to self-efficacy becoming positive, and the path from career exploration to persistence intentions becoming negative in the multivariate SEM). Exercising this level of methodological prudence when dealing with sign reversals in complex structural models is highly commendable and strengthens the overall trustworthiness of the analysis.

Response 2: We are grateful for the reviewer’s recognition of our cautious interpretation of suppression effects. To further improve clarity, we have refined the relevant discussion by explicitly emphasizing that these sign reversals arise from shared variance among predictors and should not be overinterpreted substantively. This revision strengthens the methodological transparency and interpretive rigor of the analysis.

3. Contextualizing Practical Implications: The practical implications derived from the study are sound, particularly the suggestion that institutions should offer "structured advising" and "moderated career-information workshops" to prevent career exploration from deterring female students. To increase the impact of this section, the authors could provide 1-2 brief examples of what these interventions might look like specifically within the South Korean higher education context or leveraging existing national agencies like WISET.

Response 3: Following the reviewer’s suggestion, we have expanded the Discussion section to include concrete examples of practical interventions within the South Korean higher education context.

In particular, we now illustrate: how structured advising can be implemented through university career support systems to help students interpret career exploration as an adaptive developmental process, and how moderated career-information workshops, potentially supported by national agencies such as WISET, can assist students in critically evaluating career information and reducing uncertainty. These additions enhance the practical relevance and contextual grounding of the study.

4. Future Research Directions: In the limitations section, the authors correctly point out that the cross-sectional design limits causal inference. They could briefly expand on this by suggesting that future longitudinal research should specifically track the timeline of female students' career exploration to pinpoint exactly when and how the feelings of "uncertainty" or "constraint" emerge during their undergraduate years.

Response 4: We appreciate the reviewer’s suggestion to elaborate on future research directions. In response to the reviewer’s comment, we have expanded the limitations section by adding a brief but more specific recommendation for future research. In addition to noting the limitations of the cross-sectional design, we now emphasize the need for longitudinal studies that examine when and how perceptions of uncertainty or constraint emerge during undergraduate study, thereby clarifying the temporal dynamics linking career exploration and persistence.

5. Table Formatting: Please review the formatting of Table 2. Given the large number of variables, ensuring clear spacing and alignment of the sub-variables will improve readability for the final published version.

Response 5: We appreciate the reviewer’s attention to presentation quality. Tables have been revised to improve readability, with clearer spacing and alignment of sub-variables, ensuring that the structure of the variables is more easily interpretable in the final version.

We believe that these revisions have strengthened the manuscript in terms of methodological clarity, contextual relevance, and overall readability. We sincerely appreciate the reviewer’s valuable feedback. Thank you for your useful comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for submitting your manuscript titled "Gendered Pathways to Career Exploration and Academic Persistence Among STEM Undergraduates in South Korea" for review. I have read it with great interest. The originality of your research question, the coherent application of the SCCT framework, and the large nationally representative sample are genuine strengths of this work. Regrettably, however, after careful consideration I have concluded that the manuscript cannot be recommended for publication in its current form. The methodological concerns outlined below are, in my judgment, substantial enough to warrant rejection rather than revision, as they affect the foundational validity of the measurement model and, consequently, the trustworthiness of all structural conclusions drawn from it. I sincerely hope that this feedback will nonetheless prove useful as you develop the work further, whether for resubmission to this journal or elsewhere.

1. Discriminant Validity Has Not Been Established
While Table 3 reports AVE and CR values supporting convergent validity, no evidence of discriminant validity is provided. Neither the Fornell-Larcker criterion — which requires that the square root of each construct's AVE exceed its correlations with all other constructs — nor HTMT (Heterotrait-Monotrait Ratio of Correlations) ratios are reported. Without these tests, it is not possible to confirm that the constructs are sufficiently distinct from one another, and the interpretability of the structural results is substantially compromised.

2. Standardized Path Coefficients Reach Implausible Values
Several standardized path coefficients are methodologically concerning. The path from contextual supports to engineering self-efficacy reaches β = 1.00 among male students, and the path from engineering self-efficacy to major motivation yields β = 0.98–0.99 across both groups. Standardized coefficients at or near the theoretical ceiling strongly suggest either that the constructs in question are not sufficiently differentiated or that multicollinearity is present in the model. This concern is directly consistent with the absence of discriminant validity evidence noted above, and together these two issues cast doubt on the structural conclusions of the study.

3. Common Method Bias Has Not Been Assessed
All study variables were collected simultaneously from the same participants via self-report, yet no procedure for detecting or controlling Common Method Bias is reported. Established approaches such as Harman's single-factor test, the Common Latent Factor method, or a marker variable technique should be applied and their results reported. Without such controls, it remains unclear whether the observed structural relationships reflect genuine latent associations or are partly attributable to method-related inflation — a concern that is particularly salient given the implausibly high path coefficients described above.

4. Model Fit Indices Require More Adequate Justification
RMSEA values exceed conventional acceptability thresholds in both the female (.112) and male (.121) subgroup models. While the authors acknowledge this and attribute it to model complexity, this explanation alone is insufficient. A more thorough discussion is needed, including examination of modification indices, consideration of alternative model specifications, or at a minimum a more explicit acknowledgment of the limitations these values impose on the interpretation of the findings.

Author Response

Responses to reviewer 2

We sincerely appreciate the reviewers' valuable comments and have thoroughly revised the manuscript to address their feedback to the best of our ability. Responses to the reviewers' comments are provided, and the revised sections of the manuscript are highlighted in yellow. 

1. Discriminant Validity Has Not Been Established

While Table 3 reports AVE and CR values supporting convergent validity, no evidence of discriminant validity is provided. Neither the Fornell-Larcker criterion — which requires that the square root of each construct's AVE exceed its correlations with all other constructs — nor HTMT (Heterotrait-Monotrait Ratio of Correlations) ratios are reported. Without these tests, it is not possible to confirm that the constructs are sufficiently distinct from one another, and the interpretability of the structural results is substantially compromised.

Response 1: We thank the reviewer for this important comment regarding discriminant validity. We agree that establishing the distinctiveness of the constructs is essential for the interpretability of the structural model. In response, we revised the manuscript to provide additional evidence of discriminant validity using both the Fornell–Larcker criterion and the HTMT (Heterotrait–Monotrait) ratio.

First, discriminant validity was assessed using the Fornell–Larcker criterion. As presented in Table 4, the square root of the AVE for each construct exceeds the corresponding inter-construct correlations, supporting adequate discriminant validity.

Second, discriminant validity was further evaluated using HTMT ratios. As reported in Table 5, all HTMT values in the full construct matrix were below the commonly accepted threshold of 0.90. The highest HTMT value was 0.866 between engineering self-efficacy and major motivation. Although this value slightly exceeds the more conservative threshold of 0.85, it remains below 0.90 and is theoretically consistent with the close conceptual relationship between these constructs within the SCCT framework.

Taken together, these results provide converging evidence supporting the empirical distinctiveness of the constructs and address the reviewer’s concern. Relevant explanations and tables have been added to Section 4.2 (Measurement model) and Section 3.3 (Data analysis) of the revised manuscript. 

2. Standardized Path Coefficients Reach Implausible Values

Several standardized path coefficients are methodologically concerning. The path from contextual supports to engineering self-efficacy reaches β = 1.00 among male students, and the path from engineering self-efficacy to major motivation yields β = 0.98–0.99 across both groups. Standardized coefficients at or near the theoretical ceiling strongly suggest either that the constructs in question are not sufficiently differentiated or that multicollinearity is present in the model. This concern is directly consistent with the absence of discriminant validity evidence noted above, and together these two issues cast doubt on the structural conclusions of the study.

Response 2: We thank the reviewer for this insightful comment regarding the unusually high standardized path coefficients observed in the structural model. We agree that coefficients approaching the upper bound (e.g., β = 1.00 and β ≈ 0.98–0.99) warrant careful examination, as they may raise concerns regarding multicollinearity or insufficient discriminant validity. In response, we undertook additional analyses and revisions to address this concern.

First, discriminant validity was reassessed using both the Fornell–Larcker criterion and the HTMT ratio. As reported in Tables 4 and 5, all constructs met the Fornell–Larcker criterion, and all HTMT values were below the commonly accepted threshold of 0.90. Although the highest HTMT value (0.866) slightly exceeded the more conservative threshold of 0.85, it remained within commonly accepted limits and is theoretically consistent with the close relationship between engineering self-efficacy and major motivation within the SCCT framework. These results support the empirical distinctiveness of the constructs.

Second, potential multicollinearity was examined at the latent construct level through inter-construct correlations and HTMT results, which do not suggest problematic multicollinearity.

Third, we revised the manuscript to explicitly acknowledge and interpret these high coefficients. In the revised Results section, we note that although some path coefficients are high, they may reflect strong conceptual linkages expected within the SCCT framework rather than necessarily indicating statistical redundancy. At the same time, we clarify that these estimates should be interpreted with appropriate caution.

These revisions have been incorporated into Section 4.3 (Path analysis results) for both female and male groups. We believe these clarifications adequately address the reviewer’s concern and strengthen the interpretability of the structural findings.

3. Common Method Bias Has Not Been Assessed

All study variables were collected simultaneously from the same participants via self-report, yet no procedure for detecting or controlling Common Method Bias is reported. Established approaches such as Harman's single-factor test, the Common Latent Factor method, or a marker variable technique should be applied and their results reported. Without such controls, it remains unclear whether the observed structural relationships reflect genuine latent associations or are partly attributable to method-related inflation — a concern that is particularly salient given the implausibly high path coefficients described above.

Response 3: We thank the reviewer for this important comment regarding the potential influence of common method bias. We agree that, because all variables were collected simultaneously using self-report measures, it is important to assess whether observed relationships may be affected by method-related inflation. In response, we revised the manuscript to evaluate common method bias using Harman’s single-factor test. An unrotated exploratory factor analysis including all measurement items was conducted, and the first factor accounted for 24.13% of the total variance, well below the commonly used benchmark of 50%. This suggests that common method bias is unlikely to pose a serious threat to the validity of the findings.

We also acknowledge the limitations of Harman’s single-factor test and have clarified this point in the revised manuscript. These revisions have been incorporated into Section 3.3 (Data analysis) and the relevant Results section. We believe these additions adequately address the reviewer’s concern and strengthen the methodological rigor of the study.

4. Model Fit Indices Require More Adequate Justification

RMSEA values exceed conventional acceptability thresholds in both the female (.112) and male (.121) subgroup models. While the authors acknowledge this and attribute it to model complexity, this explanation alone is insufficient. A more thorough discussion is needed, including examination of modification indices, consideration of alternative model specifications, or at a minimum a more explicit acknowledgment of the limitations these values impose on the interpretation of the findings.

Response 4: We thank the reviewer for this valuable comment regarding the interpretation of model fit indices, particularly the elevated RMSEA values observed in the subgroup analyses. We agree that a more comprehensive justification is necessary beyond attributing these values solely to model complexity. In response, we revised the manuscript to provide a more transparent discussion of model fit.

First, we explicitly report and interpret the RMSEA values for both female (.112) and male (.121) groups, acknowledging that these exceed conventional thresholds and warrant cautious interpretation of the structural relationships.

Second, we strengthened the methodological justification by incorporating relevant literature indicating that RMSEA may overestimate model misfit in complex models with large numbers of indicators and degrees of freedom (Kenny et al., 2015), and that model fit should be evaluated holistically across multiple fit indices (Marsh et al., 2004). In this context, acceptable incremental fit indices (e.g., CFI, TLI, IFI, NFI), together with theoretical coherence, support retaining the proposed model.

Third, we examined alternative model specifications and considered possible modifications. However, no theoretically meaningful modifications resulted in substantively improved fit without compromising interpretability; therefore, the current model was retained based on theoretical coherence.

These revisions have been incorporated into Section 4.3 (Path analysis results). We believe the expanded discussion provides a more balanced justification of model fit and adequately addresses the reviewer’s concern.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The article “Gendered Pathways to Career Exploration and Academic Persistence Among STEM Undergraduates in South Korea“ examines gendered pathways among contextual supports, career barriers, engineering self-efficacy, major motivation, career exploration behaviors, and academic persistence intentions among 2,393 STEM undergraduates in South Korea, using Social Cognitive Career Theory (SCCT) and a multi-group SEM comparison by gender. A major strength of the article is its clearly specified model, large sample, precisely defined constructs, and practically useful findings for STEM policy and student support programs. The article is useful because it goes beyond mean-level gender differences and analyzes the structural relationships among the constructs themselves, showing that academic persistence intentions develop through different pathways for male and female students.

The Abstract should first be revised. It is informative, but it contains too many results and recommendations in a single paragraph. It would be better to separate more clearly the study purpose, sample and method, main statistical findings, and then include only one brief practical implication at the end.

The Introduction is strong because it clearly presents the scope of the problem, the South Korean context, the SCCT framework, and two explicit research questions. However, this section should be shortened and made more precise. Some statements about women’s lower self-efficacy, the stronger impact of career barriers, and gender differences in SCCT processes are repeated again later in the Literature Review. As a result, the introduction loses focus. This section would be stronger if it distinguished more sharply among three elements: the problem context, the research gap, and the specific contribution of this study, how this model differs from previous SCCT studies and why career exploration behaviors and academic persistence intentions are included together in the same model.

The Literature Review is broad and draws on many sources, but it should be reorganized more clearly. Across Sections 2.1, 2.2, and 2.3, the same themes recur several times: women’s lower engineering self-efficacy, stronger career barriers, the greater importance of support, and gender differences in exploration and persistence. This creates the impression that the review accumulates similar claims rather than building step by step toward the proposed model. I would recommend structuring each subsection more directly around the role of a specific construct in the model: for example, how the literature supports the link between contextual supports and engineering self-efficacy, between career barriers and career exploration behaviors, and between major motivation and academic persistence intentions. It would also help to reduce interpretive summaries that are repeated in similar wording across multiple sections.

The Methods section is generally clear: the sample is described, the measures are presented, and the subscales and reliability coefficients are reported. However, several points need clarification. First, the sampling logic should be explained more explicitly: it is not fully clear whether the balance by gender and region reflects the actual population or the study design. Second, major motivation should be justified more clearly as a combined construct consisting of expected outcomes in the major and interest in the major, because this is a key theoretical point in the model. Third, in the multi-group SEM section, it would strengthen the paper to report not only that configural, metric, and scalar invariance were tested, but also the specific criteria used to determine whether invariance was acceptable. These revisions would improve methodological transparency.

In the Results section, the most important issue is interpretive precision. In the description of the correlations, the article states that career barriers had a negative correlation with career exploration behaviors, but the reported range r = -0.044~0.208 includes positive values, so this sentence needs correction. In addition, the statement that “when needed, bootstrapping procedures can be applied” does not belong in the results section unless bootstrapping was actually used; the text should state clearly whether it was applied or not. The model fit indices are also an important weakness: in the overall model RMSEA = 0.09, and in the group-specific models RMSEA = 0.112 and 0.121, so it is not sufficient simply to describe the fit as “acceptable” or “modest.” The paper needs a more direct discussion of what these values imply for the strength of the interpretations.

Recommended revision: Figures 2 and 3 should present the path coefficients more clearly and consistently. At present, several coefficients are placed too close to the arrows or surrounding constructs, which makes the diagrams harder to read.  A cleaner layout would improve readability and make the structural relationships easier to follow.

Lines 224, 239, 258, 694–695 and etc.: the dashes are too long and should be formatted normally. Please revise the punctuation in this sentence so that it is consistent with the rest of the manuscript.

Comments on the Quality of English Language

The English could be improved to more clearly express the research.

Author Response

Responses to reviewer 3

We sincerely thank the reviewer for the detailed and constructive feedback. We greatly appreciate the recognition of the study’s strengths, including its clearly specified model, large sample, and contribution to understanding gendered structural mechanisms in STEM persistence. We have carefully revised the manuscript in response to all comments. Responses to the reviewers' comments are provided, and the revised sections of the manuscript are highlighted in yellow.

1. The Abstract should first be revised. It is informative, but it contains too many results and recommendations in a single paragraph. It would be better to separate more clearly the study purpose, sample and method, main statistical findings, and then include only one brief practical implication at the end.

Response 1: We appreciate the reviewer’s suggestion to improve the clarity and structure of the Abstract. In response, we have revised the Abstract to more clearly distinguish: the study purpose, sample and methodological approach, key findings, and a single concise practical implication. We have reduced redundancy and limited the number of results and recommendations presented, thereby improving readability and focus.

2. The Introduction is strong because it clearly presents the scope of the problem, the South Korean context, the SCCT framework, and two explicit research questions. However, this section should be shortened and made more precise. Some statements about women’s lower self-efficacy, the stronger impact of career barriers, and gender differences in SCCT processes are repeated again later in the Literature Review. As a result, the introduction loses focus. This section would be stronger if it distinguished more sharply among three elements: the problem context, the research gap, and the specific contribution of this study, how this model differs from previous SCCT studies and why career exploration behaviors and academic persistence intentions are included together in the same model.

Response 2: We thank the reviewer for this insightful comment. The Introduction has been revised to improve precision and reduce repetition with the Literature Review. Specifically, repetitive statements regarding gender differences (e.g., self-efficacy, career barriers) have been reduced. The section has been reorganized to clearly distinguish: the problem context, the research gap, and the specific contribution of this study. We have also clarified how the present model extends prior SCCT research by jointly examining career exploration behaviors and academic persistence intentions within a single structural framework.

3. The Literature Review is broad and draws on many sources, but it should be reorganized more clearly. Across Sections 2.1, 2.2, and 2.3, the same themes recur several times: women’s lower engineering self-efficacy, stronger career barriers, the greater importance of support, and gender differences in exploration and persistence. This creates the impression that the review accumulates similar claims rather than building step by step toward the proposed model. I would recommend structuring each subsection more directly around the role of a specific construct in the model: for example, how the literature supports the link between contextual supports and engineering self-efficacy, between career barriers and career exploration behaviors, and between major motivation and academic persistence intentions. It would also help to reduce interpretive summaries that are repeated in similar wording across multiple sections.

Response 3: We appreciate the reviewer’s suggestion to improve the coherence of the Literature Review. In response, Sections 2.1–2.3 have been revised to reduce repetition and improve logical progression. Specifically, each subsection has been reorganized around the role of key constructs in the model (e.g., contextual supports, career barriers, self-efficacy, motivation, exploration, persistence). Repetitive interpretive summaries have been reduced. Greater emphasis has been placed on linking prior research directly to the hypothesized structural relationships. These revisions enhance conceptual clarity and alignment between theory and the proposed model.

4. The Methods section is generally clear: the sample is described, the measures are presented, and the subscales and reliability coefficients are reported. However, several points need clarification. First, the sampling logic should be explained more explicitly: it is not fully clear whether the balance by gender and region reflects the actual population or the study design. Second, major motivation should be justified more clearly as a combined construct consisting of expected outcomes in the major and interest in the major, because this is a key theoretical point in the model. Third, in the multi-group SEM section, it would strengthen the paper to report not only that configural, metric, and scalar invariance were tested, but also the specific criteria used to determine whether invariance was acceptable. These revisions would improve methodological transparency.

Response 4: We thank the reviewer for highlighting the need for greater methodological transparency.

The following revisions have been made:

• First, sampling logic clarified: We now explicitly state that the gender and regional distributions reflect the original WISET dataset and were not artificially balanced, and should be interpreted accordingly.
• Second, major motivation construct: We have strengthened the theoretical justification for treating major motivation as an integrated construct combining expected outcomes and intrinsic interest, grounded in SCCT.
• Third, measurement invariance criteria: We have added explicit criteria for evaluating configural, metric, and scalar invariance (e.g., ΔCFI ≤ .01), thereby improving methodological rigor.

5. In the Results section, the most important issue is interpretive precision. In the description of the correlations, the article states that career barriers had a negative correlation with career exploration behaviors, but the reported range r = -0.044~0.208 includes positive values, so this sentence needs correction. In addition, the statement that “when needed, bootstrapping procedures can be applied” does not belong in the results section unless bootstrapping was actually used; the text should state clearly whether it was applied or not. The model fit indices are also an important weakness: in the overall model RMSEA = 0.09, and in the group-specific models RMSEA = 0.112 and 0.121, so it is not sufficient simply to describe the fit as “acceptable” or “modest.” The paper needs a more direct discussion of what these values imply for the strength of the interpretations.

Response 5: We appreciate the reviewer’s careful reading of the Results section.

The following revisions have been made:

• The description of correlations has been corrected to accurately reflect the reported range, acknowledging that associations between career barriers and exploration are weak and mixed rather than uniformly negative.
• The sentence suggesting that “bootstrapping procedures can be applied” has been removed to reflect actual analytical procedures.
• The discussion of model fit has been strengthened: RMSEA values are now interpreted more cautiously and explicitly. In addition, we clarify that the model fit is acceptable primarily in light of incremental indices and theoretical coherence, rather than relying on general descriptors. These revisions improve interpretive accuracy and transparency.

6. Recommended revision: Figures 2 and 3 should present the path coefficients more clearly and consistently. At present, several coefficients are placed too close to the arrows or surrounding constructs, which makes the diagrams harder to read. A cleaner layout would improve readability and make the structural relationships easier to follow.

Response 6: We thank the reviewer for this helpful suggestion. Figures 2 and 3 have been revised to improve clarity and readability. Specifically, Path coefficients have been repositioned to avoid overlapping with arrows and constructs. The layout has been adjusted to ensure consistent spacing and alignment. These changes enhance the visual interpretability of the structural models.

7. Lines 224, 239, 258, 694–695 and etc.: the dashes are too long and should be formatted normally. Please revise the punctuation in this sentence so that it is consistent with the rest of the manuscript.

Response 7: We appreciate the reviewer’s attention to detail. All instances of overly long dashes (e.g., Lines 224, 239, 258, 694–695) have been corrected to ensure consistency with standard punctuation throughout the manuscript.

We believe that these revisions have substantially improved the manuscript in terms of clarity, structure, methodological transparency, and readability. We sincerely thank the reviewer for the detailed and constructive feedback, which has greatly strengthened the quality of this work.

8. Response to Comments on the Quality of English Language Point 1: The English is fine and does not require any improvement.

Response 8: Thank you for your positive evaluation regarding the quality of English. Although the language was deemed acceptable, we had the final version of the manuscript professionally proofread by a native English speaker to ensure clarity, fluency, and consistency throughout. Thank you for your useful comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for your detailed responses and for incorporating the suggested revisions. In light of your clarifications and the additional information provided, I find the manuscript suitable for publication in its current form.

Author Response

Thank you very much for your careful review and for your positive evaluation of our manuscript. We sincerely appreciate your constructive comments throughout the review process, which have greatly helped us improve the quality and clarity of the paper. We are pleased that the revised version meets the requirements for publication. Thank you again for your valuable time and insightful feedback.

Reviewer 3 Report

Comments and Suggestions for Authors

I thank the authors for the revisions. Many of the previous concerns have been adequately addressed, and the manuscript has improved in clarity, structure, and methodological transparency. However, a few minor issues remain, particularly regarding the consistency and readability of Figures 2 and 3.

Figures 2 and 3 still require correction because several displayed coefficients appear inconsistent with the values reported in the tables and narrative text. Since both figures are labelled as standardized path coefficient models, the coefficients shown in the diagrams should match the standardized estimates reported in Tables 6–8. For female students, Table 6 and the text report the path from career exploration behaviors to academic persistence intentions as β = –0.11, whereas Figure 2 appears to display –.019. Similarly, for male students, Table 7, Table 8, and the text report this path as β = –0.02, whereas Figure 3 appears to display –.002. In addition, some theoretically important paths, such as major motivation → academic persistence intentions and engineering self-efficacy → academic persistence intentions, are either missing, unclear, or not labelled with the values reported in the tables. The authors should therefore check all coefficients in Figures 2 and 3 against Tables 6–8 and ensure that the figures consistently report the same standardized estimates as the text and tables.

Minor proofreading is still needed. For example, the citation “West al., 1995” should be corrected to “West et al., 1995,”.

I recommend that the authors carefully review the entire manuscript to ensure consistency among the text, tables, figures, citations, and formatting before publication.

Author Response

Thank you for your careful review and for your constructive comments.

We have thoroughly revised the manuscript to address all of the issues you raised. Specifically, we carefully checked Figures 2 and 3 and corrected all path coefficients to ensure full consistency with the standardized estimates reported in Tables 6–8 and the corresponding narrative text. Missing or unclear paths, including major motivation → academic persistence intentions and engineering self-efficacy → academic persistence intentions, have been clearly included and properly labeled in the revised figures.

In addition, we have conducted a comprehensive proofreading of the entire manuscript. All identified errors, including the citation “West al., 1995,” have been corrected (now “West et al., 1995”, L 411), and we have carefully reviewed the text, tables, figures, citations, and formatting to ensure overall consistency and clarity.

For the reviewer’s convenience, all revised portions in the manuscript have been highlighted in green.

We appreciate your valuable feedback, which has helped us further improve the quality of our manuscript.