Institutional Conditions Associated with Youth Motivation and Retention in Technical and Vocational Education and Training: A Critical Narrative Review and Conceptual Framework for Dropout Diagnosis and Prevention

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a well written article well worth publishing. The topic is of great interest and supported by recent references. My only comment is that while the article focuses on Brazil, are there any implications for other countries? This might strengthen the article making it more relevant for international readers.

Thank you for the opportunity to review your work and all the best with publishing your work.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

Thank you for submitting this thought-provoking manuscript on the institutional determinants of youth motivation and retention in Brazilian Technical and Vocational Education and Training (TVET). The piece tackles a socially relevant question, namely the persistent dropout paradox in a country where the demand for qualified workers grows alongside underutilised vocational tracks. The decision to mobilise Self-Determination Theory (SDT) as the main analytical lens, while triangulating it with Tinto's Student Integration Theory, Expectancy-Value Theory, and Social Cognitive Theory, is a sound and ambitious choice that gives the review conceptual breadth. I particularly appreciated the attempt to shift the discourse from a deficit-centred view of dropout to a systemic, student–institution co-production model, and the effort to translate this theoretical move into a concrete three-level diagnostic model and four strategic prevention axes. Below I share my observations, in the hope that they will be useful for refining the manuscript.

Regarding the introduction, the framing of the problem is generally clear and the paradox between labour-market demand and low enrolment in vocational tracks is well established. However, line 46 contains a sentence that needs punctuation revision: "while demand for qualified workers grows youth enrollment in technical programs remains substantially below..." A comma after "grows" would significantly improve readability. The reliance on only two sources (FIESC, 2023; OECD, 2022) to justify such a strong cross-national comparison feels somewhat thin; the argument would benefit from complementary references on the structural reasons behind these disparities. I would also encourage the authors to state the research questions or guiding propositions more explicitly at the end of the introduction, rather than letting them emerge only implicitly through the two stated objectives.

A more fundamental concern relates to how the abstract represents the scope of the study. A careful inventory of the reference list reveals that a substantial portion of the cited works are Brazilian in origin or directly concern the Brazilian context: Bzuneck and Guimarães (2007), Clement, Custódio and Alves Filho (2015), Cunha (2005), Dayrell (2003), Dore and Lüscher (2011), FIRJAN and PNUD (2023), FIESC (2023), Guimarães and Bzuneck (2008), IBGE (2024), INEP (2024), Moreira and Cardoso (2024), Novaes (2006), Nunes and Silvano (2024), Observatório de Educação/Instituto Unibanco (2026), Raitz and Vanzuita (2021), Scacchetti et al. (2014), SENAI/IPRI (2024), Sobral (2003), and Souza et al. (2024 and 2025), among others. Arguably more than half of the bibliographic corpus is Brazilian, which strongly signals that the paper is not a generic international review but a contextually situated analysis grounded in Brazilian scholarship and policy documents. The abstract, however, mentions Brazil only briefly in the Background section, while the Methods, Results, and Conclusions speak in generic terms about "the literature on TVET dropout," "institutional dimensions," and "the proposed framework," with no further reference to the Brazilian context or the validated instruments adapted for Brazilian technical education, such as the Scacchetti et al. (2014) adaptation of the Academic Motivation Scale, which plays a central role in the diagnostic model. A reader scanning only the abstract might therefore reasonably conclude that the paper offers a universal SDT-based framework, whereas the body is in fact deeply rooted in Brazilian empirical and policy evidence, with cross-national applicability explicitly flagged as a limitation (lines 523–524). I would strongly recommend revising the abstract to make this regional anchoring explicit, perhaps by stating that the synthesis draws predominantly on Brazilian empirical and policy-oriented literature, and that the diagnostic model is calibrated for the Brazilian TVET system while inviting cross-national adaptation. The keywords could likewise include a term such as "Brazilian technical education" or "educação profissional" to improve indexing and signal the paper's geographical focus.

The methodology section adopts the SANRA framework, which is appropriate for a narrative review, yet the operationalisation is somewhat superficial. The authors acknowledge that "no formal SANRA score was calculated" (line 157), which raises the question of how exactly the six domains "informed review organization." A short table summarising how each SANRA domain was addressed would add transparency. Similarly, the inclusion and exclusion criteria (lines 138–147) are listed in an extremely terse, bullet-point form, with items such as "unrelated studies" and "duplicate evidence" that border on tautology. These criteria deserve elaboration, ideally with examples of what was excluded and why. The search period (January to April 2026) is mentioned as a future date, which is consistent with the journal's prospective publication year but may confuse readers; it might be worth clarifying this is the actual search window. I also noticed that the methodology relies heavily on bulleted lists throughout sections 2.2 to 2.5; converting some into running prose would improve the academic register and reduce visual fragmentation.

The theoretical framework (Section 3) is the strongest part of the manuscript. The exposition of the motivational continuum, the basic psychological needs, and the complementary perspectives is competent and accessible. Still, sub-section 3.2 reads as overly compact: each complementary theory is dispatched in a single sentence (lines 191–195), which weakens the claim that these perspectives genuinely "delimit the explanatory scope of SDT." A short paragraph per theory, explicitly indicating which mechanisms SDT under-specifies and how the other frameworks compensate, would strengthen the conceptual integration. I also detected an awkward construction in line 243: "tend to be affected, finding with direct implications for the diagnosis..." The word "finding" appears misplaced and should likely be removed or rephrased as "with direct implications." A further conceptual gap is the absence of any sociological engagement with classical accounts of educational inequality, which is striking given the Brazilian focus. Bourdieu, in particular, is conspicuously missing, as are critical perspectives on the universality of basic psychological needs across cultural contexts, a debate that is highly pertinent when SDT is applied to populations shaped by deep structural inequalities.

Section 4 on the Brazilian youth profile productively grounds the theoretical discussion in a concrete population, but the connection between socio-demographic descriptions and motivational dynamics could be tightened. At present, some passages read as parallel narratives rather than as a synthesis. The introduction of the PROGRESS-Plus framework on line 279 is interesting but appears almost in passing; the authors should either develop its relevance more substantially or relegate it to a footnote.

Section 5, presenting the institution as motivational agent, is conceptually rich and Table 1 offers a helpful visual synthesis. I would, however, suggest expanding the "Practices" column with slightly more operational examples, since the current entries (for instance, "Tutoring; mentoring") are quite general. The argument in 5.4 about the interdependence of curriculum, teaching, and student support is one of the manuscript's most valuable insights and could be highlighted more strongly, perhaps as a sub-conclusion.

The proposed diagnostic model and prevention axes in Section 6 constitute the manuscript's original contribution. Table 2, however, appears truncated in the version I reviewed: only three rows are visible and the header row is absent in the extracted text, which suggests formatting issues that should be checked before resubmission. Table 3 similarly presents examples that are extremely brief ("Flexible pathways," "Feedback," "Tutoring," "Stratified indicators," "Repeated measurements"); these single-word entries are unlikely to satisfy readers seeking actionable guidance. This minimalism in the tables contrasts sharply with the conceptual density of the surrounding text, producing an asymmetry between structural elements and running prose that should be addressed by substantially enriching both tables.

The Discussion (Section 7) is well-structured and appropriately cautious about the limits of SDT. The point that even instrumentally motivated students can internalise the value of training in supportive environments (lines 481–484) is well-articulated and theoretically important. Nevertheless, the discussion sometimes restates conclusions already drawn in earlier sections rather than engaging dialogically with potential counterarguments. A brief engagement with critical perspectives on SDT itself, such as cross-cultural debates on the universality of basic psychological needs, or sociological readings of dropout that emphasise structural rather than motivational determinants, would strengthen the analytical depth and offer a more balanced positioning of the proposed framework.

The Limitations section (Section 8) is commendably honest and the five points raised are pertinent. The Conclusions (Section 9) close the manuscript effectively, though the final sentence (lines 545–547) contains a grammatical slip: "Dropout prevention should not be reduced to persuade students to remain enrolled" should read "should not be reduced to persuading students."

I now turn to a broader observation that I believe deserves frank discussion, as it intersects with both the editorial integrity of the paper and the suggestions I have made above. Reading the manuscript closely, several stylistic features suggest the possible involvement of generative artificial intelligence in its preparation, or at the very least a writing style that has converged toward patterns commonly associated with such tools. The prose exhibits a recurring rhetorical structure in which abstract claims are introduced, immediately reformulated, and closed with a synthetic statement, frequently using triadic constructions: "curriculum, teaching practices, and student support"; "autonomy, competence, and relatedness"; "individual, relational, and institutional"; "cognitive, behavioral, and emotional." While triads are a legitimate rhetorical device, their density throughout the manuscript is unusually high and produces a somewhat mechanical cadence. Several sections rely heavily on bulleted lists with extremely terse, almost telegraphic entries that read less like the product of careful methodological reflection and more like list-generation outputs that have not been fleshed out with academic prose. The asymmetry between the polished running text and the minimal table entries reinforces this impression. The manuscript also employs a number of formulations that are characteristic of large language model outputs in academic registers, such as "operates on two complementary levels," "articulated premises," "interdependent strategic axes," "this transforms X from Y into Z," and the systematic use of "While X, Y" constructions to introduce nuance. The argumentative voice is consistently smooth and never falters or digresses, which, paradoxically, is itself suspicious: genuine human academic writing typically displays more variation in rhythm, occasional infelicities, and idiosyncratic emphases. Finally, the Author Contributions section lists a single author (J.D.M.R.) responsible for all roles, yet the title page contains template placeholders for three authors, an inconsistency that, combined with the uniform style, might suggest a rapidly drafted manuscript that has not yet been thoroughly humanised through multiple authorial revisions.

I want to be cautious here. None of these features, in isolation or even taken together, constitutes definitive evidence of AI authorship. Many human academics, particularly those writing in English as a second language or those who have internalised the conventions of international journal publication, produce prose with similar characteristics. The use of AI as a writing aid is increasingly common and, when disclosed and used responsibly, is not in itself problematic. What would be problematic is undisclosed substantive use, or use that compromises the intellectual ownership and accountability of the work. My recommendation, therefore, would be to invite the authors, in the spirit of transparency expected by most journals, including those in the MDPI family, to disclose any use of generative AI tools in the preparation of the manuscript, in accordance with current editorial guidelines. Beyond that, the manuscript would benefit from a careful authorial pass aimed at humanising the prose, introducing more variation in sentence structure, expanding the underdeveloped passages, engaging more critically with the limits of SDT, and incorporating a more distinctive scholarly voice that reflects the authors' own positioning within Brazilian educational research.

Concerning minor and editorial issues, several deserve attention. The forms of in-text citation are not entirely consistent: most references use the author–date style (Deci & Ryan, 2000), while line 600 in the reference list switches to "O'Neill J, Tabish H, Welch V, et al. ... 2014;67(1):56–64," which follows a Vancouver-style format incompatible with the rest of the bibliography. The reference list also includes "Vallerand, R.J.; et al." (line 627), which is unusual for a full bibliographic entry; all co-authors should normally be listed. The reference to "Souza, B.; Cardoso, A.; Sousa, R." (2024) and "Souza, H.C.; Cardoso, F.M.C.B.; Sousa, M.M." (2025) raises the question of whether both correspond to the citations of "Souza et al., 2025" used in the text, which should be verified to avoid ambiguity.

Italicised text appears to be used appropriately for journal names in the reference list, although I noticed some inconsistencies (for example, the title of Bandura's 1997 book is italicised but other book titles vary in formatting). The abbreviation TVET is correctly introduced, but SDT is defined in the abstract and re-introduced again in the introduction, which is acceptable; however, BPN (basic psychological needs) appears in Tables 2 and 3 without earlier formal definition in the running text, the abbreviation being spelled out only in the table notes. I would suggest introducing it at first use in Section 3.4.

Some sentences are excessively long and would benefit from being broken up. For instance, lines 70–80 contain a single sentence that combines the study's premise, two objectives, and a claim of originality; splitting this passage would improve clarity. Line 332 ("The literature consistently distinguishes two poles of teaching practice...") similarly bundles too many ideas in one period.

Looking forward, I would encourage the authors to consider, in future work, piloting the diagnostic model in at least one TVET institution, even as a proof-of-concept, since this would substantially increase the manuscript's impact and address the most prominent limitation: the absence of empirical validation. A subsequent paper could also explore methodological triangulation between the Academic Motivation Scale and qualitative student listening protocols, perhaps using mixed-methods longitudinal designs aligned with the third future research priority listed in the conclusions. An exploration of how artificial intelligence and learning analytics could support the longitudinal monitoring axis would also resonate with current debates in educational research, and would have the added merit of being an intellectually consistent use of AI within a paper whose own production methods might benefit from greater transparency.

Overall, this is a conceptually sound manuscript with a clear contribution to the literature on TVET retention in Brazil. The two interconnected issues I have emphasised, namely the abstract's failure to acknowledge the Brazilian focus and the somewhat generic, formulaic quality of the prose, both point to a manuscript that, regardless of how it was produced, has not yet fully claimed its own specificity and voice. With careful attention to these matters, alongside the editorial corrections, the enrichment of the proposed tables, the refinement of the methodology section, and the polishing of the references, the paper has strong potential to make a meaningful and lasting contribution. I commend the authors for the intellectual coherence of the project and for advancing a perspective that meaningfully repositions dropout as an institutional, rather than purely individual, phenomenon.

Author Response

Please see the attachment.

 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The central argument, that dropout in TVET reflects institutional conditions rather than individual failure alone, is well-reasoned and the framing through Self-Determination Theory is theoretically appropriate. The three-level diagnostic model and the four SDT-based prevention axes are among the more useful conceptual contributions the field has seen in this area, and the authors' choice to address the Brazilian TVET context specifically gives the paper a useful specificity that broader reviews often lack. The paper is clearly organized, and the English is strong throughout.

The methods section is the weakest part of the manuscript. The authors say the review was "informed by" SANRA but that no formal score was calculated. That is acceptable for a narrative review, but it leaves the reader uncertain about how the inclusion and exclusion process actually worked in practice. The criteria listed in Section 2.3 appear as very short fragments rather than explained principles, and there is no description of how ambiguous cases were handled or how the final corpus of sources was arrived at. The authors should add a paragraph explaining how the search process moved from retrieved records to the sources actually cited in the paper. Even a rough account of how many sources were identified, reviewed, and ultimately included would strengthen the transparency claims considerably. As it stands, a reader cannot distinguish a disciplined selection process from a more ad hoc one. The search strategy section lists databases and keywords but does not describe the volume of records identified or screened. While the authors are explicit that this is not a systematic review and not subject to PRISMA, some accounting of scope is still expected for a published narrative review. Even the SANRA criteria the paper invokes ask for meaningful transparency around search decisions.

In several places the manuscript moves from the synthesized literature into the proposed framework without clearly marking the transition. The three-level diagnostic model and the four strategic prevention axes are useful conceptual proposals, but the authors should be more explicit about signaling when they are summarizing existing evidence and when they are making original propositions that rest on their interpretation of that evidence. The SANRA criteria, which the paper cites, specifically identify this distinction as a quality indicator for narrative reviews. Adding hedging language at key transitions, or a brief orienting statement at the opening of Section 6, would help readers calibrate confidence appropriately.

Section 4.2 introduces the PROGRESS-Plus framework as a tool for identifying socially stratifying factors, but it receives only three short paragraphs and is not integrated into the diagnostic model or the strategic axes in any developed way. This feels like a missed opportunity. The authors make a persuasive argument that motivational processes cannot be interpreted independently from social determinants, but then the equity dimension largely disappears from the diagnostic and prevention framework that follows. The table in Section 6.2 includes "equity-sensitive monitoring" as a fifth axis, but this axis is described in only a single sentence in the main text. Equity deserves more sustained attention, particularly given how clearly the authors establish its importance in Section 4.2. The authors should either develop the equity axis more fully in Section 6.2 or honestly acknowledge in the limitations section that equity integration remains underdeveloped in the proposed framework.

There are also several smaller issues that should be corrected. The conceptual model in Figure 1 appears twice on page 10, which appears to be a layout error. There is a grammatical error in Section 3.4, paragraph 4, where "finding with direct implications" should read "findings with direct implications." The lists in Sections 2.3, 2.4, and 2.5 are formatted as very short bullet fragments that read more like placeholder notes than completed scholarly prose, and these sections would benefit from being written out as explanatory paragraphs consistent with the rest of the manuscript's register. Finally, the reference to Observatório de Educação/Instituto Unibanco (2026) in Section 5.3 and in the reference list cites a source dated 2026, and the authors should verify this date and confirm the source is publicly available and accessible to reviewers, since institutional web sources can be difficult to verify.

The authors acknowledge in Section 8 that findings derive predominantly from the Brazilian TVET context, which is an appropriate limitation to name, but it would be useful for the discussion section to say something more specific about which elements of the framework are likely to be context-sensitive versus which are more broadly applicable. The SDT theoretical core clearly travels well across contexts, but the specific institutional mechanisms described, particularly those tied to Brazilian federal guidelines and specific programs like SENAI, are more locally grounded. A paragraph in the discussion addressing this would help readers working outside Brazil understand how to engage with the framework productively.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors.

Thank you for having responded carefully to the first review and the substantial improvements you have made. In particular, the Brazilian anchoring is now explicit in the abstract; the search strategy is considerably more transparent; SANRA is operationalised in a dedicated table; the theoretical framework has been expanded; PROGRESS-Plus is better contextualised; the diagnostic and prevention tables are more informative; and the limitations now acknowledge the single-author selection process and lack of empirical validation. These changes directly address most of the concerns raised in my first review. The revised manuscript is therefore much stronger. Nevertheless, I see a few remaining issues that should be addressed before publication.

1. The central diagnostic framework remains more conceptual than the manuscript sometimes suggests

This is the most important remaining issue. The manuscript appropriately states that the three-level diagnostic model is a conceptual proposal rather than an empirically validated instrument. It also acknowledges that the proposed timing of measurement (entry, mid-course, and transition/exit) is an authorial recommendation requiring empirical testing.

However, the presentation in Section 6 and Table 3 sometimes moves quite quickly from evidence that particular variables are associated with motivation or dropout-related outcomes to a proposed institutional diagnostic architecture. For example, the table combines the Academic Motivation Scale, need satisfaction/frustration indicators, attendance and performance records, teacher-style measures, classroom observation, focus groups, dashboards, and equity stratification into one integrated diagnostic system.

The problem is not proposing this framework; that is arguably the manuscript’s main contribution. The problem is that the reader needs an even sharper distinction between:

1. validated instruments,
2. variables with empirical associations with dropout-related outcomes,
3. qualitative/administrative indicators that are plausible diagnostic inputs, and
4. components newly synthesised or proposed by the authors.

I recommend adding a short column to Table 3 such as “Evidence status”, using categories such as validated measure, empirically supported indicator, contextual indicator, and authorial proposal. This would make the originality of the framework clearer while preventing readers from interpreting the table as an already validated diagnostic instrument.

2. The manuscript should distinguish more explicitly between “diagnosis” and “prediction”

Related to the previous point, the terminology of diagnostic model, risk identification, and dropout prevention occasionally comes close to implying predictive validity.

The manuscript itself acknowledges that the framework has not yet been empirically validated and that some existing studies use dropout intention rather than observed dropout. This caution is appropriate.

Nevertheless, if the proposed framework is intended ultimately to identify students “before dropout becomes irreversible,” the manuscript should state more explicitly what future validation would have to demonstrate: for example, predictive validity against actual subsequent withdrawal, sensitivity/specificity or other appropriate prediction metrics, incremental validity over administrative risk indicators, and whether the institutional-level variables add predictive value beyond student-level motivation.

At present, the framework is better characterised as a diagnostic conceptual architecture than as a predictive model. I recommend making that distinction explicit throughout Section 6 and, preferably, in the abstract.

3. The evidence supporting the five prevention axes is uneven

The prevention table is now substantially improved and considerably more actionable. However, the empirical basis of the five axes is not equally strong.

The authors themselves acknowledge, for example, that much of the evidence for autonomy-supportive teacher training comes from educational settings outside Brazilian TVET, and that local adaptation and evaluation are necessary. Similarly, the Danish intervention provides suggestive evidence but is non-randomised and cannot exclude residual confounding.

I therefore recommend explicitly labelling the five axes as evidence-informed propositions rather than evidence-based interventions. The distinction is important because the manuscript's conceptual synthesis is strong, but there is not yet sufficient Brazilian intervention evidence to claim that implementation of these five axes will reduce dropout.

A brief evidence-strength indicator in Table 4 would solve this efficiently.

4. The institutional emphasis is convincing, but causal language should remain consistently cautious

The revised Discussion does a good job of stating that institutions do not “fully determine” dropout and that the argument concerns motivational conditions through which students interpret their educational experiences.

Still, some formulations elsewhere remain stronger than the underlying evidence warrants. For example, the statement that curriculum is “the institution's most structuring influence on student motivation” is a strong comparative claim, yet the review does not appear to establish that curriculum is more influential than teaching, belonging, material support, or other institutional factors. Similarly, statements that certain institutional dimensions “shape” persistence should distinguish theoretical mechanisms from empirically demonstrated longitudinal effects.

I suggest a final pass specifically searching for verbs such as determine, shape, produce, cause, lead to, influence, and replacing them where necessary with may contribute to, are associated with, provide conditions for, or are theoretically expected to affect, unless the cited evidence genuinely supports the stronger causal formulation.

This would be particularly important because the manuscript's strongest methodological virtue in the revised version is its increasingly careful treatment of associative versus causal evidence.

5. The search is much more transparent, but the two search phases still have different evidentiary status

The authors have made a commendable effort to document the August update search, including retrieval and deduplication counts. They also appropriately acknowledge that the original January–April search did not have a contemporaneous screening log.

I would nevertheless recommend making this distinction even more prominent. The reader could otherwise interpret the 471 references from the update search as if they constituted the complete evidence base from which the final 56 sources were systematically selected.

A concise sentence in the Methods could clarify that the 56-source corpus is the result of a narrative synthesis combining the earlier search, the prospectively documented update, citation chaining, and targeted verification, rather than the endpoint of a single reproducible systematic screening process.

This would align perfectly with the authors' stated narrative-review design and avoid any unintended appearance of quasi-systematic methodology.

6. The “single-author” limitation has been acknowledged, but its implications for the framework deserve one more sentence

The revised Limitations section appropriately states that source identification, selection, extraction, and appraisal were conducted by one author and that no independent duplicate screening was undertaken.

Because the article's main contribution is a conceptual synthesis, however, this limitation affects not only the literature-selection process but potentially the construction of the framework itself. The coding of evidence into autonomy, competence, relatedness, curriculum, teaching, support, and equity dimensions necessarily involves interpretive decisions.

I recommend adding one sentence acknowledging that the three-level architecture and five prevention axes should be considered interpretive syntheses whose structure may be sensitive to authorial coding decisions, and therefore should be subjected to independent replication or expert review in future work.

7. PROGRESS-Plus is now better integrated, but its adaptation remains under-specified

The revision correctly describes PROGRESS-Plus as a heuristic lens rather than a validated educational scale. The diagnostic table nevertheless operationalizes it through stratification by socioeconomic status, gender, race/ethnicity, disability, geography, and work-study variables.

The remaining issue is practical: how should institutions avoid turning these equity markers into deficit labels or risk scores attached to particular groups?

The manuscript makes the important argument that the same motivational score may require different institutional responses depending on social position and structural barriers. This is one of the stronger aspects of the paper. But the framework would benefit from one explicit warning that equity stratification should be used to identify institutional disparities and barriers, rather than to assign inherently higher “risk” to particular demographic groups.

That clarification would strengthen the ethical and conceptual consistency of the framework.

8. Minor point: the manuscript could be slightly more concise

The authors have improved the mechanical style and reduced fragmentation, but some repetition remains between Sections 6, 7, 8, and 9. The central qualification that the model is conceptual, requires validation, and should not be interpreted as proving institutional causation is stated several times.

The repetition is not severe, but modest tightening would make the contribution more forceful. The strongest version of the paper would spend slightly less space reiterating its limitations and slightly more space explaining what empirical validation of the proposed model would look like.

Author Response

Please see the attachment.

Author Response File: Author Response.docx