Universal Design for Learning

1. From individual accommodation to anticipatory design

Individual accommodation modifies time, format or task once someone has already failed to enter the design that was planned. It can be fair. Its limit is curricular: the environment was conceived for a supposed student, and variation appears afterwards, as an exception. Gómez Martín et al. (2026) describe UDL 3.0 as a framework for anticipating barriers, widening participation and promoting progress. It is a conceptual framework, in the opening of a psychometric study, not a result of the questionnaire. Daley (2025), in a theoretical analysis, proposes designing for human variation, accepting that design for specific students may benefit everyone, and sustaining focused goals with flexible means. Self-determination and socio-emotional learning illustrate that application as examples of the theoretical framework, not as trials.

The framework weighs more as policy than as a measured intervention. Boysen (2024), in a critical analysis, calls it a highly influential source of education policy: CAST presents the guidelines as supported by learning, cognitive psychology and the brain. Hollingshead et al. (2026) interviewed 19 experts on policy, measurement and future research; a keyword analysis produced five themes that the abstract does not name. The qualitative finding is that intersection. Dolan (2026) argues, with Ireland as a country that has adopted UDL, that the framework can depoliticise inclusion even if it is useful: an emancipatory education would also ask about ableist, gender, racial and class hierarchies. The question of this review is what effects UDL has on learning, participation and access, under what implementation and with what quality of evidence.

2. What UDL is and is not: principles, guidelines and version 3.0

Guo and Wang (2025) name the three principles of full implementation: multiple means of engagement, of representation and of action and expression. Anastasiou et al. (2025) warn that multiplicity when information is presented can collide with cognitive load. Gómez Martín et al. (2026) place version 3.0 in policies and practices, in addition to anticipating barriers, widening participation and promoting progress. Changes and checkpoints absent from the corpus are not dated. Saenen et al. (2026), in 16 studies and reviews from 2013 to 2023, document the ambiguity between universal design and UDL: multimodal materials, flexible assessments, accessible technologies and nine components that the abstract does not list, together with a holistic framework without clear causal links and with uneven principles.

It is not the same as differentiated instruction. Griful-Freixenet et al. (2021) address the beliefs and practices of pre-service teachers with respect to the two frameworks; without an abstract, there is no effect, sample or conclusion to cite. Nor is UDL an intervention reserved for disability. De la Fuente-González et al. (2025), in 88 articles on teacher education, find three conceptualisations and a dominant and problematic association with disability or with learning difficulties, even though the approach seeks to improve the learning of all students without substantive modifications of the curriculum. Daley (2025) argues in a compatible direction. And “more is better” does not hold. Roski et al. (2021), a four-group experiment, with invariance analysis, MIMIC models and a multigroup panel, do not support the claim that the more extensive environment outperforms the less extensive one. The questionnaire adapted with Universal Design for Assessment detected the gains better. The abstract of Roski et al. (2021) does not indicate the educational level.

3. Method: critical narrative review with a screening flow

This critical narrative review covers articles from 2021 to 2026 and does not estimate an effect of its own. The reason is a methodological inference: self-competence, correlation with a test, interview and self-report do not share a metric. The base was Crossref, closed on 23 October 2026, with 14 seed queries of up to 60 rows on meta-analyses, reviews, disability, teacher education, implementation, school, higher education, online settings, accessibility, differentiated instruction and inclusive attitudes. The seed yielded 412 unique verified records. Excluded were 2 reviews or corrections, 7 duplicate or mirror DOIs, 64 journals of low editorial assurance, 91 works that do not address UDL and 13 of another scope, among them architecture, libraries, rural extension, aviation and theology. The exclusions sum to 177 and leave 235 eligible out of 412. Manual selection retained 21, without double coding. The lexical screening is in screen.py.

Thirteen targeted queries, of 8 to 10 rows, added 3 records verified one by one because the seed did not retrieve them: King-Sears et al. (2023), Boysen (2024) and Pérez-Conde (2026). The corpus contains 24 works, 21 from the seed and 3 complementary; 22 with an abstract from Crossref, OpenAlex or the publisher, and 2 with the title only. There are 0 studies from Mexico and 0 Latin American empirical studies. Alteridad, an Ecuadorian journal, publishes De la Fuente-González et al. (2025), and Revista Fuentes, in Spain, publishes Pérez-Conde (2026), but the home of the journal is not the site of the fieldwork. The studies conducted in Spain are Barrera Ciurana and Moliner García (2024) and Gómez Martín et al. (2026).

4. What the meta-analyses say: positive effects, greater heterogeneity

Almeqdad et al. (2023) present a systematic review and a meta-analysis of 13 empirical pre-post studies, published between 2015 and 2021 in English and Arabic, in six countries and in school and higher education, with a specific or a generic domain, directed at teachers or at students by means of professional development or school interventions, and in the set the single-group design predominated. The meta-analytic finding is a total effect of 3.56, with considerable heterogeneity; significance was concentrated in the single-group designs, in the student participants, in the specific domain and in the quantitative design. The caution that corresponds is a methodological inference: that magnitude is atypical for single-group designs, which do not separate the programme from maturation or from retest, so that overestimation is probable.

Pérez-Conde (2026) applies the PRISMA protocol to 673 articles and retains 10. The finding is a positive and statistically significant effect on the achievement of university students with disabilities, together with more participation and more self-efficacy and with moderate heterogeneity; the abstract does not report the coefficient. Institutional commitment, training, planning and standardisation are the author’s conclusion, not measured moderators, and they do not extend on their own to the whole student body. King-Sears et al. (2023) is the meta-analysis of the achievement of students who receive instruction with UDL; only the title was available, so neither its magnitude nor its direction is reported. AlRawi and AlKahtani (2022) mix quantitative, qualitative and mixed designs — digital environments, e-books, shared stories, audio, software and activities — and call for more participants and for lived experience, and do not pool an effect. The inference that closes the section is that, where the result can be read, the direction is positive, and that this direction cannot be estimated as a common magnitude.

5. Compulsory schooling: few trials and outcomes that are not always learning

Curtiss et al. (2025) are the only cluster-randomised trial that the abstracts describe. In community-based participatory research, 30 schools were randomly assigned to a four-day puberty workshop in a UDL version or a traditional one. In 2054 students, knowledge differed significantly between pretest and posttest. The hierarchical linear models found no difference between conditions in knowledge, and did find one in the sense of self-competence, greater with UDL. In both outcomes there was no association with Title 1 status or with the percentage of students with disabilities. Teacher fidelity was high, the teaching staff perceived that participation was facilitated and, even so, had difficulties with the multiple means of representation, with group management and with technology.

What the workshop already produces in both versions is not attributed to the framework, and teacher perception is not a student scale. That lack of association is not an individual interaction. Craig et al. (2024), a correlational study, in grades 3 to 8 of a district in the United States with UDL as a system-wide initiative, find better state tests — among students with disabilities as well — when the mathematics or language teacher has a higher implementation score. The abstract of Craig et al. (2024) does not indicate the coefficient, and the design does not separate the teacher who implements the framework from the teacher who would already obtain good scores.

Zhang et al. (2024) prevent those results from being added together. In 32 studies from 1999 to 2023, from preschool to grade 12, explicit alignment with the checkpoints, even coverage and theoretical guidance are lacking; the checkpoints overlap. They recommend documenting the least applied ones and their relationships. Ewe and Galvin (2023), in Libsearch, Ebsco and Scopus, see in the European school — set against research concentrated in the university — qualitative perceptions, few learning outcomes and a scarcity of preschool work; the abstract of Ewe and Galvin (2023) does not indicate how many articles were included. The abstract of Roski et al. (2021) does not indicate the educational level; the experiment shows that adding principles does not improve in a monotonic way and that an inaccessible test can conceal a gain.

Guo and Wang (2025) are not an effect in compulsory schooling: longitudinal mixed methods, 2473 students, 87 institutions, diverse contexts. With the three principles under full implementation they report +37.4 % in overall achievement and +42.8 % among disengaged students, and a significant relationship (p < 0.001) more marked in engaged students than in disengaged ones. The qualitative factors are administrative support, continuous professional development, technology that is available, collaborative pedagogy and flexible use of data. The authors limit generalisation. Without a randomised contrast described, the two comparisons do not allow one to decide, taken together, who benefits more.

6. Higher education and online settings: access, participation and student voice

Saenen et al. (2026) start from the distance between inclusive ambition and university practice. Universal design is promoted, but its implementation and its evaluation remain fragmented: the actions are often interpreted as UDL, and without a causal link it is not known which principle produces what. Pérez-Conde (2026) is the most direct figure on university students with disabilities — a positive and significant effect on achievement, more participation and more self-efficacy, moderate heterogeneity, 10 studies out of 673 — and it likewise does not identify the principle that is responsible. “UDL worked” remains a package. The training and the institutional commitment that this review calls for are its conclusion, not a moderator that can be verified within the 10 studies.

Barrera Ciurana and Moliner García (2024) interviewed seven students with autism spectrum disorder at a Spanish university. Several named key concepts or progressive information; the majority became more involved if their opinions and their effort were taken into account. UDL could benefit them and others, but a discourse that would change practice is lacking. It is a qualitative finding: there is neither prevalence nor a comparison of courses. Beck Wells (2022), with mixed methods and a Likert scale given to 57 undergraduate students, explores perceived prevalence in online courses, independent studies and virtual groups. The abstract of Beck Wells (2022) does not indicate which modality predominated and does not give means; the framing of achievement, retention and international students is not a result. Dolan (2026) argues, without evaluating Ireland, that a normative UDL can depoliticise inclusion. As a reading inference, in this stretch access, participation and voice weigh more than controlled contrasts of learning.

7. STEM and technology: the promise of access, mixed evidence on achievement

Kefalis et al. (2026) review UDL in STEM and in settings that use technology, from 2002 to 2025, in Scopus, Web of Science, ERIC and IEEE Xplore. Of 13 studies, 11 are empirical and 2 are theoretical. The empirical evidence is promising and heterogeneous: those practices can support access, engagement, self-regulation, persistence and inclusive participation, while standardised achievement is more mixed. Multimodal representation, flexible participation, scaffolding and socio-emotional support recur. The framework guides practice, they conclude, when the environment sustains cognitive access, emotional engagement, collaboration and diverse needs. It is a review finding. There is no STEM effect size, and many of the primary studies predate 2021.

Anastasiou et al. (2025) do not add an empirical study. Their theoretical analysis identifies seven problems and makes explicit the doubtful universality, the possible clash between multiplicity and cognitive load, the difficulty of proving efficacy and the distinction between effective and ineffective technology. The problems on which the abstract is silent are not filled in here. An inference, not a mediation: mixed achievement fits integrations that sometimes add access and sometimes add demand. Curtiss et al. (2025) show that, with high fidelity, educators stumbled over technology, representation and group management in a puberty workshop, which is not a STEM intervention. The design inference is that those components can fail while the fidelity checklist remains high.

8. The problem of measuring implementation fidelity

If it is not said which principle was implemented, a positive outcome does not credit the framework and a null outcome does not refute it. Zhang et al. (2024) place that break at the start of the chain. Rusconi and Gäng-Pacifico (2024) present the UDL-OMT for observing UDL in teaching sequences, translated from English into French and into Italian, and still under validation: it is a measurement framework, not a validity finding. Gómez Martín et al. (2026) offer preliminary evidence for the CID-DUA3. From 20 items reviewed by 10 experts the set was reduced to 12, administered to 91 teachers. α = 0.876 (standardised, 0.879). The ordinal exploratory analysis, with polychoric correlations, retained one factor; the confirmatory analysis, with lavaan and mean- and variance-adjusted weighted least squares, fitted (CFI = 0.975; TLI = 0.969; RMSEA = 0.068). The more knowledge was declared, the higher the score (p < 0.001). What is measured is self-reported frequency, not observation, and the link is cross-sectional.

The abstract of Craig et al. (2024) does not indicate whether the implementation score is observation or self-report. Curtiss et al. (2025) show high fidelity with knowledge equal across conditions: the global judgement can conceal the principle that fails, and there is no rate given as a percentage. Guo and Wang (2025) call fidelity the application of the three principles, associated with 37.4 % and 42.8 % (p < 0.001). It is neither the checklist of a workshop nor the 12-item questionnaire. Aligning those operationalisations is an inference, not psychometric equivalence.

9. Teacher education and beliefs

Rusconi and Squillaci (2023) review 12 studies of UDL courses and of teacher competences in inclusive classrooms, across four dimensions of the European Profile of the Inclusive Teacher that the abstract does not name one by one. There are significant effects in valuing the diversity of the student body; effectiveness appears independent of duration, modality and type of teacher; and accessible planning and implementation are fostered. Collaboration and reflective practice remain uncertain. It is a finding about teacher competences, not about student achievement. “Independent of duration” does not mean that any course will suffice, and the durations are not invented. The neuroscience that the introduction invokes is the authors’ framework, not a result of the 12 studies.

De la Fuente-González et al. (2025) do not measure whether the course changes practice: in 88 articles they see three conceptualisations and a problematic predominance that associates UDL with disability and, on their account, restricts application. Calling that a mechanism is an inference. The study by Griful-Freixenet et al. (2021) of the beliefs and practices of pre-service teachers faced with UDL and with differentiated instruction marks a different object and contributes no result. Gómez Martín et al. (2026) associate, in 91 teachers, more declared knowledge with a higher CID-DUA3 score (p < 0.001): a cross-sectional signal, not the efficacy of a course. Curtiss et al. (2025) show difficulties with representation, group management and technology despite high fidelity. A design inference: raising a checklist does not, by itself, prepare the gestures that fail in the classroom. Guo and Wang (2025) include professional development among five qualitative factors, without isolating its weight and under the same caution about generalisation.

10. Equity and disability: design for all or a label for some?

De la Fuente-González et al. (2025) document that the teacher-education literature sticks UDL to disability and narrows practice, even though the framework is presented as an improvement in everyone’s learning without reforming the curriculum at its root. That says how the concept is taught, not who learns more. Daley (2025) argues that including students with disabilities makes educational psychology pertinent: designing for variation, designing for specific students because it may benefit everyone, and fixing goals with flexible means. The argument does not demonstrate that UDL raises self-determination, and it is not the same kind of proof.

Denying the subgroup would be the inverse error. Barrera Ciurana and Moliner García (2024) gather, from seven students with autism spectrum disorder in Spain, key concepts, progressive information, being listened to and seeing effort valued. Treating that as a diagnosis reproduces the reduction; as general teaching it resembles Daley (2025). Pérez-Conde (2026) justifies reporting the university subgroup and not defining the framework as a disability intervention. AlRawi and AlKahtani (2022) read effectiveness in seven studies and ask for greater clarity. Without a pooled effect across the two, the subgroup evidence can be real and it remains thin. Curtiss et al. (2025) did not associate the condition with Title 1 or with the disability percentage, and that does not confirm equality at the individual level. Dolan (2026) warns that an uncritical UDL can reduce equity to accessible materials.

Mexico and Latin America are a limit of the corpus. There are 0 studies from Mexico and 0 Latin American empirical studies. Alteridad and Revista Fuentes are the homes of journals, not samples: the debate already has an Ibero-American home. Without national standards, programmes or figures, no magnitude in this corpus — neither 3.56, nor the coefficient that the abstract of Pérez-Conde (2026) does not report, nor 37.4 % — travels to a classroom in the region as if the system and teacher education were the same. The European Profile of the Inclusive Teacher does not describe local initial teacher education. Seven voices from a single Spanish university do not stand in for a situated voice. An adoption in Mexico would be a design decision under uncertainty: reasonable if it requires an explicit principle, observation of fidelity, an accessible measure and learning separated from perception; unjustified if it asserts that the literature has already estimated the local effect.

11. Critiques of the construct: vagueness, cognitive load and cited evidence

Anastasiou et al. (2025) criticise the construct as a theoretical analysis, not as a classroom finding. Of seven problems, the abstract makes explicit the doubtful universality, the possible clash between multiplicity and cognitive load, the difficulty of proving efficacy and the distinction between effective and ineffective technology. A framework that does not say what would count as a failed application is difficult to put to the test. Boysen (2024) examines the studies cited by CAST for three guidelines, in relation to choice, learning and the brain. They contribute little evidence: most offered neither choice nor a measure of learning; none addressed brain function; taken together they supported single and effective techniques rather than multiple means. The cited evidence is weak, basic research and implementation research are needed, and the analysis does not cover every citation used by CAST.

Zhang et al. (2024) turn the objection — an unclear definition, implementation that is difficult, efficacy that is little evidenced — into a review finding: no alignment, uneven coverage, overlap and no theoretical guidance. Recording the same gesture twice, if two guidelines describe it, is an inference drawn from that overlap. Roski et al. (2021) show that more is not better and that the instrument decides whether the gain is seen; they do not measure load, so the convergence with Anastasiou et al. (2025) is not a proof of that theory. Saenen et al. (2026) describe uneven principles, with no causal link and with ambiguity between universal design and UDL, and they maintain that universal design can be an effective holistic framework.

Daley (2025) upholds the three principles as theoretical discipline, with disability as a lens and not as a compartment: the argument does not need the 3.56, and it does not answer Boysen (2024). Kefalis et al. (2026) offer a conditional empirical counterweight, promising and heterogeneous for access and participation, with mixed standardised achievement; the two theoretical texts among those 13 are not trials of efficacy. The construct is vulnerable where it is vague, where multiplicity ignores load, where the cited evidence on the brain and on choice is weak, and where implementation is not aligned with checkpoints. It remains usable if the decision is stated, is observed and is tied to an outcome that is not perception alone.

12. Criteria for turning principles into verifiable decisions

The nine rules are a design inference. They are not guidelines of version 3.0, nor the results of a trial.

1. State the principle and, if one exists, the checkpoint. “UDL was applied” is not enough (Zhang et al., 2024; Guo & Wang, 2025).
2. Measure fidelity by observation as well as by self-report. The UDL-OMT moves in that direction and remains under validation (Rusconi & Gäng-Pacifico, 2024). The CID-DUA3 is a preliminary self-report (Gómez Martín et al., 2026).
3. Do not read high fidelity as learning. In Curtiss et al. (2025) fidelity was high and knowledge did not differ between conditions; representation, group management and technology failed.
4. Assess with accessible instruments. In Roski et al. (2021) the version that used Universal Design for Assessment detected the gains better.
5. Separate perception, participation and learning (Ewe & Galvin, 2023; Beck Wells, 2022; Curtiss et al., 2025).
6. Treat multiplicity as a hypothesis set against cognitive load, not as a sum of channels (Anastasiou et al., 2025; Roski et al., 2021).
7. Do not reduce the framework to disability and, even so, analyse subgroups (De la Fuente-González et al., 2025; Daley, 2025; Pérez-Conde, 2026; AlRawi & AlKahtani, 2022; Barrera Ciurana & Moliner García, 2024).
8. Distinguish effective technology from ineffective technology and record the difficulties that appear alongside high fidelity (Anastasiou et al., 2025; Kefalis et al., 2026; Curtiss et al., 2025).
9. Report an association as an association. Craig et al. (2024) and Guo and Wang (2025) do not substitute for a comparison that isolates the design. It is worth stating which principle was not covered and which outcome did not move.

13. Limitations

The base is Crossref plus 13 queries that recovered three texts absent from the seed. What is not indexed is not here. screen.py and the manual selection, without double coding, leave open the possibility that another reader would have retained other works among the 235 eligible, and they leave out many accounts of practice. King-Sears et al. (2023) and Griful-Freixenet et al. (2021) are cited by title only; any figure about them would be an invention. The five themes of Hollingshead et al. (2026), the nine components of Saenen et al. (2026), the coefficient of Pérez-Conde (2026), the number of studies in Ewe and Galvin (2023) and the coefficients of Rusconi and Gäng-Pacifico (2024) were not filled in by conjecture.

Pre-post designs, correlation and perception predominate. The only cluster-randomised trial is Curtiss et al. (2025): it does not suffice for causes at the level of a system. Guo and Wang (2025) report 37.4 % and 42.8 %, extraordinary figures, with a caution about generalisation and without a randomised contrast described. There is no meta-analysis of our own. Publication bias is an unmeasured risk. The 64 journals set aside by a heuristic may have excluded a serious paper; the exclusions of scope, 91 and 13, are more transparent. There are no studies from Mexico and no Latin American empirical studies. Several primary studies predate 2021: 2008–2018 (AlRawi & AlKahtani, 2022), 1999–2023 (Zhang et al., 2024), 2015–2021 (Almeqdad et al., 2023), 2002–2025 (Kefalis et al., 2026) and 2013–2023 (Saenen et al., 2026). The window is the date of the synthesis.

14. Conclusions

Moving from individual accommodation to a design that anticipates barriers is coherent. Gómez Martín et al. (2026) formulate that move for version 3.0. Daley (2025) formulates it as expected variation and as fixed goals with flexible means. Boysen (2024), Hollingshead et al. (2026) and Dolan (2026) show that this language organises policy. Influence is not validity.

The quantitative syntheses are positive and unstable. Almeqdad et al. (2023) report a total effect of 3.56, with considerable heterogeneity, in 13 pre-post studies that are above all single-group designs. Pérez-Conde (2026) finds a positive and significant effect among university students with disabilities, together with more participation and self-efficacy and with moderate heterogeneity, in 10 studies out of 673. Of King-Sears et al. (2023) only the title was available and, for that reason, the result is not summarised.

In Curtiss et al. (2025), with 30 schools and 2054 students, knowledge rose from pretest to posttest and did not differ between versions; self-competence was greater with UDL, and fidelity was high. Craig et al. (2024) contribute a correlation, not that comparison. Guo and Wang (2025) report 37.4 % and 42.8 % and limit generalisation. Kefalis et al. (2026) see access and participation in STEM that are promising and heterogeneous, and standardised achievement that is mixed. Roski et al. (2021) reject the claim that more UDL is better.

Zhang et al. (2024) document alignment that is absent, coverage that is uneven and overlap. The CID-DUA3 is a self-report (Gómez Martín et al., 2026); the UDL-OMT observes and remains under validation (Rusconi & Gäng-Pacifico, 2024). Rusconi and Squillaci (2023) find effects for diversity and for accessible planning, not for collaboration or for reflection. De la Fuente-González et al. (2025) show that teacher education sticks the framework to disability. The value of UDL lies in observable decisions that separate perception from learning. Without an empirical study in Mexico or in Latin America, that is design under uncertainty, not local evidence.

NEXTECH.IA Editorial Lab / Ingeniero Mitre

References

1. Almeqdad, Q. I., Alodat, A. M., Alquraan, M. F., Mohaidat, M. A. y Al-Makhzoomy, A. K. (2023). The effectiveness of universal design for learning: A systematic review of the literature and meta-analysis. Cogent Education, 10(1), Article 2218191. https://doi.org/10.1080/2331186x.2023.2218191
2. AlRawi, J. M. y AlKahtani, M. A. (2022). Universal design for learning for educating students with intellectual disabilities: a systematic review. International Journal of Developmental Disabilities, 68(6), 800–808. https://doi.org/10.1080/20473869.2021.1900505
3. Anastasiou, D., Wiley, A. L. y Kauffman, J. M. (2025). A Critical Analysis of Theoretical Underpinnings of Universal Design for Learning. Exceptionality, 33(3), 145–162. https://doi.org/10.1080/09362835.2024.2426801
4. Barrera Ciurana, M. y Moliner García, O. (2024). ‘How does universal design for learning help me to learn?’: students with autism spectrum disorder voices in higher education. Studies in Higher Education, 49(6), 899–912. https://doi.org/10.1080/03075079.2023.2259932
5. Beck Wells, M. (2022). Student perspectives on the use of universal design for learning in virtual formats in higher education. Smart Learning Environments, 9(1), Article 37. https://doi.org/10.1186/s40561-022-00218-6
6. Boysen, G. A. (2024). A critical analysis of the research evidence behind CAST’s universal design for learning guidelines. Policy Futures in Education, 22(7), 1219–1238. https://doi.org/10.1177/14782103241255428
7. Craig, S. L., Smith, S. J. y Frey, B. B. (2024). Universal design for learning: connecting teacher implementation to student outcomes. International Journal of Inclusive Education, 28(11), 2458–2474. https://doi.org/10.1080/13603116.2022.2113459
8. Curtiss, S. L., Stoffers, M., Brown, A. y Mindler, P. (2025). The Effects of Universal Design for Learning on Fifth-Grade Puberty Instruction to Enable Accessibility for Students with Disabilities. Exceptional Children, 92(1), 54–72. https://doi.org/10.1177/00144029251350061
9. Daley, S. G. (2025). Universal design of educational psychology? Improving theory and application by focusing on students with disabilities. Educational Psychologist, 60(3), 141–153. https://doi.org/10.1080/00461520.2024.2441661
10. De la Fuente-González, S., Menéndez Álvarez-Hevia, D. y Rodríguez-Martín, A. (2025). Universal Design for Learning. A systematic review of its role in Teacher Education. Alteridad, 20(1), 113–128. https://doi.org/10.17163/alt.v20n1.2025.09
11. Dolan, E. (2026). Universal design for learning as emancipatory education?: politics, disagreement, discomfort. Teaching in Higher Education, 31(5), 883–899. https://doi.org/10.1080/13562517.2026.2612981
12. Ewe, L. y Galvin, T. (2023). Universal Design for Learning across Formal School Structures in Europe—A Systematic Review. Education Sciences, 13(9), 867. https://doi.org/10.3390/educsci13090867
13. Gómez Martín, M., Cruz Cruz, P., Martínez Cano, A. y Martín Cruz, I. (2026). Psychometric evidence for the universal design for learning 3.0 teacher implementation questionnaire (CID-DUA3): development and preliminary validation. Frontiers in Education, 11, Article 1921354. https://doi.org/10.3389/feduc.2026.1921354
14. Griful-Freixenet, J., Struyven, K. y Vantieghem, W. (2021). Exploring pre-service teachers’ beliefs and practices about two inclusive frameworks: Universal Design for Learning and differentiated instruction. Teaching and Teacher Education, 107, 103503. https://doi.org/10.1016/j.tate.2021.103503
15. Guo, P. y Wang, Z. (2025). Implementation fidelity of universal design for learning and effects on student achievement engagement and belonging. Scientific Reports, 15(1), Article 45430. https://doi.org/10.1038/s41598-025-30079-1
16. Hollingshead, A., Lowrey, K. A. y Howery, K. (2026). Universal Design for Learning Research: Where Policy, Implementation, and Evidence Collide. Journal of Disability Policy Studies, 37(1), 3–13. https://doi.org/10.1177/10442073261435095
17. Kefalis, C., Skordoulis, C., Papoutsi, C. y Drigas, A. (2026). Universal Design for Learning in STEM: Evidence from a Systematic Review and Directions for Inclusive Practices. Computers, 15(7), 433. https://doi.org/10.3390/computers15070433
18. King-Sears, M. E., Stefanidis, A., Evmenova, A. S., Rao, K., Mergen, R. L., Owen, L. S. y Strimel, M. M. (2023). Achievement of learners receiving UDL instruction: A meta-analysis. Teaching and Teacher Education, 122, 103956. https://doi.org/10.1016/j.tate.2022.103956
19. Pérez-Conde, M. L. (2026). Impacto del Diseño Universal para el Aprendizaje en estudiantes universitarios con discapacidad: Revisión y metaanálisis. Revista Fuentes, 1(28), 1–13. https://doi.org/10.12795/revistafuentes.2026.28573
20. Roski, M., Walkowiak, M. y Nehring, A. (2021). Universal Design for Learning: The More, the Better? Education Sciences, 11(4), 164. https://doi.org/10.3390/educsci11040164
21. Rusconi, L. y Gäng-Pacifico, D. (2024). Universal Design for Learning Observation – Measurement Tool. Revue suisse de pédagogie spécialisée, 14(03), 17–25. https://doi.org/10.57161/r2024-03-03
22. Rusconi, L. y Squillaci, M. (2023). Effects of a Universal Design for Learning (UDL) Training Course on the Development Teachers’ Competences: A Systematic Review. Education Sciences, 13(5), 466. https://doi.org/10.3390/educsci13050466
23. Saenen, L., Stevens, R., Hermans, K., Struyven, K. y Emmers, E. (2026). From principles to practice: Universal design in higher education—A systematic review. Review of Education, 14(1), Article e70129. https://doi.org/10.1002/rev3.70129
24. Zhang, L., Carter, R. A., Greene, J. A. y Bernacki, M. L. (2024). Unraveling Challenges with the Implementation of Universal Design for Learning: A Systematic Literature Review. Educational Psychology Review, 36(1), Article 35. https://doi.org/10.1007/s10648-024-09860-7