Toward a sensory-gradient framework for neuroinclusive secondary-school design: an integrative review and contextualization for a tropical Nigerian setting

1 Introduction

Neurodivergent adolescents—including autistic students, students with ADHD, dyslexia, dyspraxia, and mixed sensory-processing differences—attend secondary schools whose acoustic, visual, and spatial conditions were overwhelmingly designed around a narrow, unstated sensory norm. Autistic adolescents themselves describe mainstream secondary schools as sites of sensory overwhelm that shape their schooling experience (McKinlay et al., 2025; Price and Romualdez, 2025), and qualitative work following students through the transition into secondary school documents them actively developing tactics—muting, filtering, transforming space—to navigate sensorially demanding buildings not designed with them in mind (Birkett et al., 2022).

Architecture is plausibly a regulatory rather than a purely accommodating variable in this experience, though the evidentiary strength of that claim varies by mechanism and population. Classroom acoustic conditions have been associated with attention, behavior, and speech comprehension, with autistic and ADHD populations identified as disproportionately affected in several reviews (Black et al., 2022; Masiero et al., 2025; Mealings and Buchholz, 2026), though much of the underlying evidence is observational or self-report rather than experimental. Lighting, spatial sequencing, circulation predictability, and the availability of retreat space have each been linked, across a fragmented but growing literature, to sensory load, self-regulation, attention, and participation (Mostafa, 2014; Tola et al., 2021; Al-Harasis et al., 2025).

Several established frameworks already address components of this problem. Mostafa’s Autism ASPECTSS™ Design Index (Mostafa, 2014) identifies seven recurring design qualities for autism-friendly environments: acoustics, spatial sequencing, escape space, compartmentalization, transition spaces, sensory zoning, and safety, which were subsequently tested and refined through post-occupancy evaluation (Mostafa et al., 2024). Three organizing factors, sensory quality, intelligibility, and predictability, were identified independently by Tola et al., 2021 in a scoping review of 21 included studies. Black et al. (2022) synthesized 28 studies to develop design recommendations for construction, lighting, sound, aesthetics, temperature, and air quality. Most recently, Wang et al. (2026) proposed a sensory-inclusive built-environment (SIBE) framework spanning four sensory modalities, eleven factors, and six design qualities. Universal Design for Learning (UDL) addresses curricular and instructional variability rather than spatial organization directly (CAST, n.d.), and Attention Restoration Theory (ART) explains why low-demand, naturalistic environments can restore depleted attentional capacity (Kaplan, 1995), operationalized architecturally through biophilic design (Barbiero et al., 2021).

Given this existing cluster of built-environment reviews, several of which were published within the last 5 years and one (Wang et al., 2026) explicitly proposing a multi-factor sensory framework, this review’s task is narrower and more precisely defined than it might first appear. It is not to identify, for the first time, that sensory design qualities matter for neurodivergent learners; that has now been established by several independent reviews using different methods and samples. Its task is to test whether a specific architectural proposition developed independently of this literature, a graduated whole-campus sequence of stimulus zones connected by transition infrastructure, originating in earlier unpublished postgraduate architectural design work, is supported, requires modification or should be rejected when placed against that evidence, and to establish what, if anything, it adds to Tola et al. (2021), Black et al. (2022) and Wang et al. (2026) specifically. The earlier design work is treated only as an a priori proposition under review, not as empirical evidence.

A complementary architectural literature provides a broader basis for evaluating the framework’s claims about adjacency, circulation and legibility. Foundational wayfinding work treats spatial organisation and environmental information as integral to navigation rather than as signage problems alone (Arthur and Passini, 1992). School-specific spatial-configuration studies likewise show that layout, connectivity and circulation structure can be examined as variables related to movement, social interaction, user evaluation and participatory planning (Kishimoto and Taguchi, 2014; Saghafi and Mirzaei, 2021; Psathiti and Charalambous, 2025). These sources are not neurodivergence-specific and are therefore used here as architectural background for interpreting spatial mechanisms, not as direct evidence that particular neuroinclusive interventions are effective.

A further gap concerns age and geography. Much of the built-environment literature, including Tola et al., 2021 and Black et al. (2022), draws predominantly on evidence spanning early childhood through adulthood rather than specifically on secondary-school adolescents, and almost none of it is set in West Africa. A tropical, lower-resource context such as Ota, Ogun State, additionally requires design strategies to be reconciled with passive cooling, daylighting, and rain-noise management, since mechanical climate control cannot be assumed (Mba et al., 2022; Xi, Sa’ad, Liu, Sun, Wang & Guo, 2025), and with a policy environment in which inclusive-education commitments have been reported to outpace physical-infrastructure delivery (Federal Ministry of Education, 2015).

This review therefore has three aims: first, to test a pre-existing Sensory Gradient proposition against interdisciplinary evidence, reporting explicitly which of its original elements are supported, modified, rejected or newly added; second, to establish what a refined Sensory-Gradient Framework adds to existing built-environment reviews and frameworks (ASPECTSS, UDL, ART, Tola et al., 2021; Black et al., 2022; Wang et al., 2026) rather than assuming novelty; and third, to develop an initial contextual application of the resulting framework for Ota without claiming local empirical validation.

1.1 Aim, objectives, and research questions

This review aims to test, refine, and contextualize a Sensory-Gradient Framework for neuroinclusive secondary-school design against interdisciplinary evidence on sensory processing, environmental psychology, neuroarchitecture, and inclusive education, for application in Ota, Nigeria. Its objectives are: (i) to identify the sensory, cognitive, spatial and social needs of neurodivergent adolescents in school environments, distinguished by neurodivergent group where the evidence allows; (ii) to examine architectural and environmental features associated with sensory regulation, attention, participation and wellbeing; (iii) to establish precisely what the Sensory-Gradient Framework adds beyond ASPECTSS, UDL, ART and the Tola et al., 2021, Black et al., 2022 and Wang et al. (2026) reviews; (iv) to test the framework’s original single-axis stimulus-gradient logic against the evidence and revise it where the evidence does not support it; (v) to specify testable propositions; (vi) to contextualise the framework for Ota’s climatic, economic, institutional and policy conditions; and (vii) to set out a research agenda for empirical testing.

The primary research question is: Does the evidence support, and how should it modify, a graduated whole-campus spatial-sequencing framework for neuroinclusive secondary-school design? Supporting questions are (RQ1) what sensory, cognitive, spatial and social requirements do neurodivergent secondary-school learners have (RQ2) which built-environment features are associated with sensory regulation, attention, participation and well-being, and through what mechanisms (RQ3) is stimulus intensity better represented as a single axis or as a multidimensional demand profile (RQ4) what do ASPECTSS, UDL, ART, Tola et al. (2021), Black et al. (2022) and Wang et al. (2026) already establish, and what, if anything, remains unaddressed (RQ5) how should transition magnitude and learner-perceived control be incorporated as design variables; and (RQ6) which climatic, material, economic, regulatory and institutional conditions moderate the framework’s application in Ota.

1.2 Relationship to prior work: an abductive review, not an inductive origin story

The Sensory Gradient originated in earlier, unpublished postgraduate architectural design work completed before this review. The pre-review proposition organized a secondary-school campus as a broadly graduated sequence from more public and activity-intensive settings toward more restorative learning and support settings, with transition zones functioning as sensory buffers. The earlier design work is used here only as the a priori conceptual proposition subjected to evidential testing; it is not treated as empirical data or as evidence supporting the framework.

For transparency, Box 1 summarises the four components of the proposition entering the review before any literature-based modification. This makes it possible to distinguish the original architectural assumptions from the revised framework generated through the abductive audit.

BOX 1. Pre-review Sensory Gradient proposition entering the evidence audit.

This review does not claim that the Sensory-Gradient Framework emerged inductively from the literature reviewed in Sections 3, 4; that would misrepresent the review’s logic. The review is abductive: the pre-existing architectural proposition is treated as a hypothesis to be tested against published evidence, and the framework presented in Section 5 is derived by revising that proposition rather than independently deriving it. The audit reported in Section 5.4 explicitly identifies which components of the original single-axis Sensory Gradient were supported, modified, rejected, or newly added. The most consequential change is reported in Sections 4.6 and 5: the evidence does not support treating stimulus intensity as a single interchangeable axis, so the original high-to-low formulation is replaced by a multidimensional demand-profile model.

2 Methods

2.1 Review design and eligibility criteria

This study employed an integrative review design following Whittemore and Knafl (2005), with a systematic literature search embedded within it. The methodological components served distinct purposes. Whittemore and Knafl’s approach structured the integrative review and synthesis of heterogeneous evidence; systematic database searching was used to identify eligible literature; PRISMA 2020 was used specifically as a transparency and reporting framework for identification, screening, eligibility, and inclusion; critical appraisal was used to judge methodological quality; and an abductive proposition audit was used to test the pre-existing Sensory Gradient against the synthesized evidence. The review was therefore not presented as a conventional intervention-focused systematic review, and PRISMA reporting did not replace the integrative-review methodology. The earlier architectural proposition was not treated as a data source or as evidence in support of itself.

Eligible evidence concerned neurodivergent learners, particularly autistic learners and learners with ADHD, dyslexia, dyspraxia, or sensory-processing differences, principally within the secondary-school age range of approximately 11–19 years. Studies involving younger children, adults, or general populations were retained only where they provided transferable evidence on relevant environmental mechanisms and were classified as indirect evidence. Eligible environmental variables included acoustics, lighting, visual complexity, spatial configuration, wayfinding, circulation, sensory zoning, transition and retreat spaces, learner control, social density, biophilic design, thermal comfort, and ventilation. Outcomes included sensory regulation, distress, attention, participation, navigation, autonomy, safety, belonging, and wellbeing. Peer-reviewed empirical studies, evidence reviews and relevant framework-development literature published between January 2000 and July 2026 were eligible, with foundational pre-2000 theoretical sources retained where necessary. Government policy and clearly attributable practice sources were used only for contextualisation and were not treated as equivalent to peer-reviewed empirical evidence.

2.2 Search strategy and study selection

Evidence identification proceeded through two complementary routes. Stream A comprised systematic searches of Scopus, Web of Science Core Collection, ERIC, PsycINFO, MEDLINE/PubMed, and the Avery Index to Architectural Periodicals, completed on 15 July 2026. Searches covered January 2000 to 15 July 2026, targeted English-language peer-reviewed journal evidence, and combined three concept groups: neurodivergence/sensory processing, secondary-school/adolescent learning environments, and architecture/built-environment design. The common Boolean expression was (autism* OR ADHD OR dyslexia* OR dyspraxia* OR “sensory processing” OR neurodivergen) AND (“secondary school” OR adolescen OR classroom) AND (architecture OR “built environment” OR acoustic* OR lighting OR “spatial sequencing” OR biophilic), with field wrappers and platform syntax adapted to each database. Supplementary Table S1 reports the database-specific implementation, limits, and yields.

The six database searches yielded 356 records: Scopus (n = 112), Web of Science Core Collection (n = 84), ERIC (n = 63), PsycINFO (n = 48), MEDLINE/PubMed (n = 32), and Avery Index (n = 17). After 72 duplicates were removed, 284 titles/abstracts were screened, and 222 were excluded. Sixty-two reports were sought and assessed at full text; 47 were excluded because of wrong population (n = 18), wrong setting (n = 14), absence of a relevant built-environment/environmental variable (n = 9), or ineligible publication type (n = 6). Fifteen empirical/review sources were retained from Stream A. Backward and forward citation tracking of key reviews was used as a coverage check; the principal anchor sources included Tola et al., 2021, Black et al. (2022), Masiero et al. (2025), and Wang et al. (2026).

Stream B captured evidence identified through other methods that was intentionally outside the main 2000–2026 peer-reviewed database stream. Two foundational theoretical/framework sources were retained through citation/framework searching because they were necessary for construct interpretation despite falling outside the main database eligibility route. Seven Nigeria/Ota-relevant policy, climatic, and infrastructural sources were identified through a purposive supplementary search in Google Scholar and regional repositories, including the Covenant University repository, completed in July 2026. Search terms included (‘Nigeria’ OR ‘Ota’ OR ‘Lagos’) AND (‘inclusive education policy’ OR ‘passive school design’ OR ‘tropical rainfall’ OR ‘school architecture’), supplemented by forward citation tracking from Nigeria’s National Policy on Special Needs Education and relevant local studies. Sources were selected for geographical relevance and applicability to educational infrastructure, policy, or local climate. Two additional practice-oriented sources concerning dyslexia/dyspraxia and the visual environment were retained as low-confidence illustrative context rather than peer-reviewed empirical evidence. Stream B therefore contributed 11 sources in total: two theoretical/framework, seven Nigeria/Ota contextual, and two practice-oriented sources. The two streams are shown separately and merged only at the final synthesis stage in the complete PRISMA 2020 flow diagram (Figure 1).

PRISMA 2020 was used specifically to report identification, screening, eligibility, and inclusion transparently within the broader integrative-review design. Figure 1 presents the complete parallel-stream PRISMA 2020 flow diagram, including database-level yields, deduplication, title/abstract exclusions, full-text exclusion reasons, other-methods identification, and the final 26-source synthesis.

Contextual and practice evidence from Stream B was kept analytically separate from the peer-reviewed evidence used to test neurodivergence-specific propositions. These sources were used to identify climatic, infrastructural, policy, implementation, or illustrative neurotype-specific boundary conditions and were not treated as evidence of equivalent efficacy for the framework.

The resulting evidence base, therefore, comprised two transparently reported identification streams rather than a single undifferentiated pool. This separation is carried through the synthesis and evidence-to-framework audit so that local contextualisation is not mistaken for empirical validation of the neuroinclusive propositions.

2.3 Data extraction and critical appraisal

Data were extracted using a structured study-characteristics form capturing author, year, country, study design, sample, participant age, neurodivergent group, setting, environmental variable, principal findings and study limitations. Evidence was additionally classified by population directness and context directness so that secondary-school evidence could be distinguished from evidence transferred from younger, adult or general populations and from non-school settings.

Methodological quality was assessed separately from evidential directness because these dimensions answer different questions: methodological quality concerns the internal credibility of a source, whereas directness concerns how closely its population and setting match neurodivergent secondary-school learners and educational built environments. Primary qualitative, quantitative and mixed-methods studies were appraised using the Mixed Methods Appraisal Tool (MMAT) (Hong et al., 2018). Systematic, scoping, and structured reviews were appraised using AMSTAR 2 item-level domains (Shea et al., 2017) as a structured methodological appraisal framework. Because AMSTAR 2 was developed principally for systematic reviews of intervention studies and the included evidence syntheses were methodologically broader, item-level judgments were used to inform the descriptive quality categories reported in this integrative review rather than being represented as formal AMSTAR 2 overall confidence ratings or numerical scores.

Theoretical and conceptual sources were not assigned empirical quality ratings but were used to define constructs and compare frameworks. Appraisal findings informed the weight given to each source during synthesis. Descriptive categories (High, Moderate-high, and Moderate) were therefore synthesis-level judgments informed by the relevant appraisal items; methodological quality was not combined mathematically with population or context directness into a single confidence score.

To make the descriptive methodological-quality categories reproducible, the same synthesis-level rules were applied across sources. High indicated that all or nearly all applicable appraisal criteria were satisfied and that no important methodological weakness was identified; Moderate-high indicated that the core appraisal criteria were satisfied but one or more non-critical limitations remained; and Moderate indicated that minimum credibility criteria were met but multiple limitations and/or one important methodological concern reduced confidence. For primary studies appraised with MMAT, these judgments were based on the five design-specific MMAT criteria after the screening questions. For review evidence, judgments were based on applicable AMSTAR 2 domains, including protocol and search transparency, study selection and data extraction, appraisal of included studies, appropriateness of synthesis, and interpretation of limitations. These labels were descriptive synthesis judgments rather than numerical scores or official AMSTAR 2 overall confidence ratings. Both authors appraised sources independently, compared judgments, and agreed the final category by consensus where initial ratings differed.

2.4 Synthesis approach

Because of heterogeneity in study designs, populations, and outcomes, findings were synthesized narratively using a hybrid deductive-inductive approach. Evidence was organized across environmental-demand dimensions, including acoustic, visual, movement/activity, social, thermal, and other sensory demands; predictability/legibility was coded separately as a cross-cutting characteristic affecting anticipation of those demands rather than as another demand dimension. Evidence was then mapped through the analytical chain: environmental condition -> learner experience -> mechanism -> architectural response -> outcome, shown schematically in Figure 2. This chain was also used to distinguish findings directly demonstrated in the reviewed studies from architectural translations proposed by the framework.

Negative cases, conflicting findings, and boundary conditions were actively sought. The original Sensory Gradient proposition was then subjected to an abductive proposition audit using explicit decision rules. A component was retained when convergent evidence supported both its underlying mechanism and its original scope; modified when the underlying principle remained plausible but the evidence required a change in scope, mechanism, or architectural expression; rejected when a defining assumption was contradicted by the evidence or by recurrent heterogeneity that made the original formulation untenable; and added when recurrent evidence identified a construct absent from the pre-review model. Audit judgments considered the direction and convergence of findings, methodological quality, population directness, and context directness separately. Conflicting findings were not averaged into a single score; they were retained as negative cases or boundary conditions, thereby reducing the strength assigned to the corresponding proposition. This procedure led to rejection of the original single high-to-low stimulus axis and its replacement with a multidimensional environmental-demand model incorporating transition magnitude, predictability/legibility, and learner-perceived control.

The proposition-audit decisions were conducted through independent coding by both authors (OOO and OAT). Following an independent assessment, the authors held consensus meetings to compare judgments on whether each component should be retained, modified, rejected, or added. Where initial interpretations differed, the relevant sources were re-examined with particular attention to methodological quality, population directness, context directness, evidential convergence, and contradictory findings until a final consensus was reached.

To assess incremental contribution rather than assume novelty, the refined framework was compared with ASPECTSS, Universal Design for Learning, Attention Restoration Theory, and the recent built-environment syntheses of Tola et al. (2021), Black et al. (2022), and Wang et al. (2026). Nigerian and Ota-relevant evidence on school infrastructure, passive design, climate, and policy was synthesized separately as contextual evidence and used to identify implementation trade-offs rather than to validate neurodivergence-specific propositions. Figure 3 summarises this overall review-to-framework workflow without duplicating the PRISMA record counts.

3 Characteristics of the evidence base and study selection

The systematic database stream identified 356 records, of which 284 remained after deduplication. After title/abstract screening, 62 reports were assessed at full text, and 15 empirical/review sources were retained from Stream A. Stream B contributed 11 sources identified through other methods: two theoretical/framework sources used for construct interpretation, seven Nigeria/Ota contextual sources, and two practice-oriented sources. The two streams, therefore, yielded 26 evidence sources in the final integrative synthesis. Five methodological standards (Whittemore and Knafl, 2005; Torraco, 2005; Page et al., 2021; Hong et al., 2018; Shea et al., 2017) guided the review but were not counted among the 26 evidence sources; additional architectural and regional contextual sources cited to situate the discussion were used only as supplementary background and were not counted among the 26 audited evidence sources or treated as neurodivergence-specific evidence. Table 1 shows that the 15 core sources were methodologically heterogeneous: four were rated High, five Moderate-high, and five Moderate, with one framework-development source (Mostafa, 2014) not formally rated on an empirical quality scale. Population directness was also uneven: six sources were direct to school-age or secondary-school autistic populations, one was mixed/partly direct, and eight relied on mixed-age, younger, or otherwise indirect populations. This distribution explains why later framework claims are weighted most strongly toward autism and why transfer to ADHD, dyslexia, and dyspraxia is treated as provisional rather than assumed. Methodological quality and directness are interpreted together in the narrative below without collapsing them into a single score.

TABLE 1. Study-level characteristics and appraisal of the 15 core empirical and review sources included in the synthesis. Descriptive methodological-quality categories are defined in Section 2.3.

4 Synthesis of evidence

4.1 Learner needs, kept distinguished by the neurodivergent group

Autistic secondary-school students report school environments as sources of sensory overwhelm, with mainstream secondary school specifically identified as more difficult than primary school (McKinlay et al., 2025). In participatory photo-voice work, participants described the sensory environment as ‘unnatural’ and identified unpredictable multisensory conditions as sources of distress that affect behavior and learning (Price and Romualdez, 2025). A related study following the transition into secondary school found autistic children actively developing sensory ‘tactics’ — muting, filtering, transforming space—to navigate demanding buildings (Birkett et al., 2022), indicating that the transition itself is a distinct point of vulnerability.

Sensory needs are not uniform across neurodivergent groups, and the evidence does not justify treating ‘neurodivergence’ as a homogeneous design category. Classroom acoustic difficulties are reported across autism, ADHD, and language-impairment populations, but the strength and directness of evidence differ. Mealings and Buchholz’s (2026) scoping review of classroom environments for children with ADHD found that most studies addressed learning outcomes rather than listening or wellbeing directly, identifying a gap that the present review inherits rather than resolves. Practice-oriented sources on dyslexia and dyspraxia emphasize visual rather than auditory sensitivity - including glare and high-contrast lighting for some dyspraxic students and daylight considerations for dyslexic students (Architecture, 2024; Dowsing and Reynolds, 2024) - but this evidence is practice-based and is therefore treated as low-confidence illustrative context rather than as a tested design rule. Across groups, the synthesis more strongly supports a learner-environment fit perspective than a diagnosis-specific prescription for universally lower stimulation: the relevant problem is a mismatch between an individual’s regulatory profile and the changing demands of a setting. This distinction becomes central to the rejection of a single-axis gradient in Section 4.6.

4.2 Environmental features, mechanisms, and architectural responses

Across the included literature, evidence is strongest for the links between environmental conditions, learner experience, and plausible regulatory mechanisms; it is weaker for some of the specific architectural responses proposed later in the framework. The synthesis, therefore, follows the analytical chain in Figure 2 and distinguishes between studies that directly demonstrate and design translations that remain propositions for testing.

Acoustic and visual demand. Classroom acoustic conditions are associated with attention, speech comprehension, and behavior, with autistic and ADHD populations reported to be disproportionately affected (Black et al., 2022; Masiero et al., 2025; Mealings and Buchholz, 2026). Masiero et al. (2025) further argue that conventional acoustic standards are largely based on neurotypical hearing assumptions and may not adequately account for autistic sensory-processing differences. Across these sources, the recurring mechanism is not simply ‘noise is bad’, but that background sound, variability, and poor signal-to-noise conditions can increase listening effort, sensory load, and task interference. Visual evidence is less experimentally developed but converges around glare, lighting adjustability, and visual complexity: diffuse or controllable lighting and reduced clutter recur across autism-focused reviews (Tola et al., 2021; Black et al., 2022; Wang et al., 2026). These findings support treating acoustic and visual demand as separable dimensions of a space rather than merging them into a single stimulus-intensity score. Architectural responses such as absorptive finishes, lighting controls or screening are therefore evidence-informed translations of these mechanisms; their effectiveness depends on context and should not be read as uniformly tested interventions across all learner groups.

The same distinction between condition and response is important when interpreting design recommendations: evidence that glare or clutter contributes to discomfort does not by itself establish one universally optimal lighting level, color scheme, or display density. The framework therefore prioritizes adjustability, legibility, and learner-environment fit over a fixed sensory prescription.

Spatial sequencing, circulation and wayfinding. ASPECTSS identifies spatial sequencing, compartmentalisation, transition spaces, sensory zoning and safety as core autism-oriented design qualities (Mostafa, 2014), while its later post-occupancy application includes colour-coded navigation and transition alcoves as concrete design interventions (Mostafa et al., 2024; Al-Harasis et al., 2025) taxonomy likewise recommends simplified circulation and threshold marking, while noting that one-way circulation can be difficult to implement in large multi-age schools. Broader architectural research adds a complementary, non-neurodivergence-specific perspective: wayfinding theory treats spatial organization, decision points, and environmental information as interdependent (Arthur and Passini, 1992), while school studies have operationalised spatial configuration through connectivity, integration, intelligibility, participatory spatial analysis and user perceptions (Kishimoto and Taguchi, 2014; Saghafi and Mirzaei, 2021; Psathiti and Charalambous, 2025). Together these literature support treating adjacency, route structure and legibility as legitimate architectural variables. They do not, however, directly validate the specific campus-scale demand-profile network proposed in Section 5; that network is an architectural synthesis derived from the evidence and requires empirical testing.

Restorative opportunity and biophilic design. Kaplan’s (1995) Attention Restoration Theory proposes that exposure to softly fascinating natural stimuli can restore directed-attention capacity. In a 3-year longitudinal case study, biophilic classroom retrofits were associated with higher perceived restorativeness and improved sustained-attention performance relative to conventional classrooms, although the sample was small, primary-school-aged and not neurodivergent-specific (Barbiero et al., 2021). The evidence therefore supports restoration as a plausible mechanism and justifies retaining restorative opportunity in the analytical model, but it does not directly establish a neurodivergence-specific requirement for a particular type or quantity of biophilic intervention in secondary schools.

Nigerian evidence on restorative and biophilic design remains limited but emerging. A user-centred qualitative study of a proposed therapeutic student centre at Covenant University found that students prioritized natural lighting, private low-stimulus rooms, and access to counselling and outdoor space (Babalola et al., 2024), while separate Nigerian reviews of therapeutic and biophilic design principles reinforce daylight, nature contact, and spatial privacy as recurring restorative design levers (Ekhaese and Salami, 2026; Sholanke et al., 2026). An empirical study of Lagos office buildings similarly found biophilic design features associated with occupant productivity, although the biophilic elements evaluated occurred as isolated rather than integrated interventions (Aduwo et al., 2021). None of these studies involves neurodivergent learners or is set in a secondary school, so they are treated as convergent contextual support for the restorative mechanism rather than direct evidence for the framework’s propositions.

Learner control and environmental agency. The strongest direct mechanistic evidence for the learner-agency proposition comes from studies in which aspects of sensory exposure or multi-sensory-environment use were manipulable. Unwin et al. (2022) found experimentally that autistic children aged 4–12 showed increased attention and fewer repetitive/sensory behaviors when they had control over multi-sensory-environment equipment compared with no-control conditions; this is indirect with respect to age but directly relevant to the mechanism of control. Dyer et al. (2026) studied autistic pupils aged 11–15 and found that self-directed use of multi-sensory environments supported wellbeing and the management of sensory overwhelm, with participants valuing the flexibility to adapt the space to context-dependent needs. Jones et al. (2026) experimentally manipulated classroom noise and visual displays and found effects on task engagement, strengthening the case that specific environmental variables can influence attention. These studies directly support the importance of agency over manipulable environmental conditions. Extending that mechanism to learner-selectable campus routes, alternative seating locations or other spatial choices is an architectural proposition of this review rather than an intervention directly tested in those studies.

4.3 Existing frameworks and reviews: what they already establish

Six bodies of work recur across the literature synthesized above, and each is compared directly against the proposed Sensory-Gradient Framework in Table 2. The Autism ASPECTSS™ Design Index (Mostafa, 2014) offers architecturally specific, post-occupancy-tested design qualities including spatial sequencing and sensory zoning, but does not specify a whole-campus sequencing rule connecting zones in a particular order with specified transition magnitude. Tola et al. (2021), Black et al. (2022) and Wang et al. (2026) each independently synthesise built-environment evidence into organising factors or design-quality catalogues, but none of the three specifies campus-scale adjacency and sequencing as a primary organising variable, nor do any of them centre transition magnitude or learner-perceived control as design variables in their own right, though Wang et al. (2026) comes closest by explicitly incorporating sensory modality, factor and design-quality dimensions. UDL (CAST, n.d.) addresses instructional rather than spatial variability. ART (Kaplan, 1995), operationalised through biophilic design (Barbiero et al., 2021), supplies a restoration mechanism but was not developed for, nor extensively tested with, neurodivergent adolescents specifically.

TABLE 2. Structural comparison of the Sensory-Gradient Framework against ASPECTSS, UDL, ART/biophilic design, and three recent built-environment reviews.

Table 2 makes this comparison explicit rather than asserting it. The Sensory-Gradient Framework’s distinctive contribution, on this comparison, is not any single design quality: acoustics, zoning and transitions all already appear in ASPECTSS and in Tola et al., 2021, but the combination of campus-scale adjacency and sequencing, transition magnitude as an explicit, measurable design variable, and learner-perceived control as a central moderating mechanism, applied across all three spatial scales (campus, building, room) simultaneously.

4.4 Direct evidence from secondary-school-age neurodivergent adolescents

A cluster of recent, direct, participatory studies of autistic adolescents specifically in mainstream secondary schools strengthens this review’s confidence in several propositions beyond what the four anchor reviews (which draw on mixed-age samples) can offer alone. McKinlay et al. (2025) synthesised 36 qualitative studies (of 117 identified) on autistic adolescents’ and adults’ accounts of positive secondary-school experience, generating themes of understanding and acceptance, environments that suit me, and agency to meet my needs—the latter two mapping directly onto this review’s demand-profile and learner-control constructs. Price and Romualdez (2025) and Dyer et al. (2026) provide small-sample but direct self-report evidence from students aged 12–16 and 11–15 respectively. Jones et al. (2026) and Unwin et al. (2022) provide experimental (rather than purely self-report) evidence, strengthening the mechanistic case for the acoustic/visual and learner-control propositions specifically, though Unwin et al.'s sample was younger than this review’s target population.

4.5 Nigerian and Ota-relevant evidence

A survey of 47 staff across three Lagos special-needs schools found gaps in spatial flexibility (mean 2.6/5), accessibility (mean 2.8/5) and sensory design, with sensory and physical impairments rated the most severe student challenge (mean 4.2/5); the study recommends universal-design standards and post-occupancy evaluation (Alabi et al., 2026). As geographically proximate evidence from southwestern Nigeria, the study is relevant to implementation conditions but is not Ota-specific and does not validate the framework. A parallel study of Abuja public primary schools reported persistent basic-accessibility barriers, including the absence of ramps in the surveyed schools (Ahmed et al., 2024). Nigeria’s National Policy on Special Needs Education (Federal Ministry of Education, 2015) provides the policy context, while secondary commentary on implementation gaps (Kareem, 2025) is used only as contextual supplementation. These sources establish constraints on implementation and inclusion; they are not treated as direct evidence for neurodivergence-specific sensory mechanisms.

More recent Nigerian evidence extends this pattern beyond special-needs and primary schools to other public building typologies. A study of universal-design compliance in south-west and south-south Nigerian cultural centres, conducted by Covenant University-based researchers, found approachability criteria more often met than accessibility and usability requirements (Aduwo and Olu-Oseh, 2026). Although the study concerns neither neurodivergent learners nor secondary schools specifically, it corroborates, from an independent building typology, the implementation gap between policy commitments and built reality that Proposition 6 treats as a boundary condition on the framework’s application in Ota.

On climate, Nigerian empirical studies confirm the centrality of passive design to classroom comfort: orientation, shading and cross-ventilation can improve thermal conditions and reduce mechanical-cooling dependence (Mba et al., 2022; Xi et al., 2025). Ota-specific rainfall measurements further document pronounced tropical rainfall conditions and high rain rates at Covenant University (Omotosho et al., 2017). These findings are not neurodivergence-specific, but they establish important contextual boundary conditions because natural ventilation, daylight and rain protection can interact directly with acoustic exposure, glare and visual distraction.

4.6 Negative-case and boundary evidence, including the case against a single-axis gradient

Consistent with an integrative review obligation to seek disconfirming evidence, this section reports the finding that most directly required a revision to the framework. The original Sensory Gradient proposition treated stimulus intensity as a single axis, from high to low. Three lines of evidence argue against this. First, the environmental-demand dimensions identified across the reviewed literature acoustic, visual, movement, social, olfactory/tactile and thermal demand (Wang et al., 2026; Al-Harasis et al., 2025), vary independently: a school cafeteria is plausibly high in social, movement and acoustic demand simultaneously while being only moderate in visual complexity, whereas a science laboratory may be high in olfactory/tactile demand and task predictability but only moderate in social demand. Collapsing these into one score would misrepresent both spaces. Second, the desirable stimulus level is not uniform across or even within neurodivergent groups: hyposensitive individuals may actively seek stronger sensory input, while hypersensitive individuals require attenuation, a distinction noted across the reviewed sensory-processing literature and directly recognized in Wang et al. (2026) multi-modality framework. Third, a strict high-to-low sequence risks recreating the homogeneous conception of neurodivergence that this review otherwise critiques (Section 4.1). For these reasons, Section 5 replaces the single-axis gradient with a multidimensional demand-profile model.

Other design tensions are carried forward as explicit boundary conditions rather than resolved: daylight admission versus glare; natural ventilation versus external acoustic intrusion (Proposition 7); legible wayfinding cues versus visual clutter; quiet retreat provision versus the risk of stigma or use as informal disciplinary space; predictability versus flexibility of use; sensory attenuation versus sensory-seeking behaviour; compartmentalisation versus social inclusion; individual learner control versus institutional supervision and safety requirements; and one-way, simplified circulation (preferred by many autistic individuals) versus the practical difficulty of enforcing single-direction movement across a large, multi-age, shared-timetable secondary school (Al-Harasis et al., 2025). ASPECTSS’s own founding 2008 classroom study has been noted, including by researchers otherwise supportive of the framework, to have used a small, non-blinded sample (Mostafa et al., 2024), a limitation this review treats as a caution against over-reliance on any single source, including the sources this review itself leans on most heavily.

5 The sensory-gradient framework

The framework below results from testing the pre-existing single-axis Sensory Gradient proposition against the evidence in Section 4 and modifying it accordingly. It organizes decisions across campus, building, and room scales but represents environmental demand in multiple dimensions rather than as a single stimulus score. Importantly, not every architectural response shown in the framework has been directly tested in the cited studies. Where evidence establishes a learner experience or mechanism but does not test its spatial translation, the corresponding architectural response is presented as an evidence-informed proposition requiring empirical evaluation.

5.1 Core model

The refined framework conceptualizes regulation and participation as functions of the interaction between a learner profile and an environmental-demand profile, as well as the magnitude of transitions between adjoining settings. This interaction shapes regulatory experience, including sensory load, cognitive demand, restorative opportunity, and transition difficulty. Perceived control and predictability/legibility are treated as buffering or moderating characteristics, while climatic, cultural, policy, and resource conditions operate as contextual moderators. Architecture can therefore influence the demand profile of each space, the magnitude and legibility of transitions, and the degree of environmental choice available to learners. The conceptual relationships are illustrated in Figure 4.

For this framework, an environmental-demand profile is the configuration of demands associated with a setting across separable dimensions, including acoustic, visual, movement/activity, social, thermal, and olfactory/tactile conditions. Predictability/legibility is deliberately not included as another demand dimension: it describes how well a learner can anticipate, interpret, and prepare for those demands. This distinction allows, for example, a setting to be acoustically intense yet predictable, or relatively quiet yet difficult to anticipate because conditions change abruptly.

Transition magnitude is defined as the degree of change in environmental demand encountered when moving between two adjoining settings across one or more dimensions. The framework does not assume that this must be collapsed into a single universal numerical score. In future empirical testing, dimension-specific change could be operationalised using measures such as sound level and variability for acoustic demand; luminance, glare or visual-complexity ratings for visual demand; pedestrian flow or occupancy for movement/social demand; and operative temperature or thermal variability for thermal demand. Learner-reported anticipation difficulty, perceived control, and regulatory response could then be analyzed alongside those environmental measures. These are candidate operational indicators for future validation, not measurement instruments validated in the present review.

5.2 Campus-scale demand-profile network

At the campus scale, the framework no longer assumes a universal linear progression from objectively high-to low-stimulus space. Instead, it proposes a network of multidimensional environmental-demand profiles connected by buffered transitions and, where feasible, learner-selectable options. Public/arrival areas, learning and social settings, core learning clusters, restorative or low-demand nodes, and specialist support spaces are planning heuristics rather than fixed categories or a mandatory sequence. The design objective is to reduce abrupt, involuntary changes in demand, make upcoming conditions legible, and provide opportunities for regulation without assuming that all learners seek lower levels of stimulation. Figure 5 is therefore an illustrative architectural proposition rather than a validated campus configuration.

5.3 Framework propositions

Seven propositions translate the model into testable claims relevant for design.

P1 Greater transition magnitude - defined as a larger change between adjoining multidimensional demand profiles - is hypothesized to increase transition difficulty; buffering and gradual recalibration may attenuate this effect (Mostafa, 2014; Al-Harasis et al., 2025).

P2 Greater predictability and legibility of upcoming environmental demands before entry may reduce navigational and cognitive demand. Predictability/legibility is treated as a cross-cutting buffering characteristic rather than a component of the demand profile itself (Al-Harasis et al., 2025; Tola et al., 2021).

P3 Access to supportive, non-stigmatizing retreat space may enhance opportunities for self-regulation and continued participation, provided that such space is not treated as a form of discipline (Price and Romualdez, 2025; Dyer et al., 2026).

P4 Learner-perceived control over manipulable environmental conditions - including lighting, seating, proximity, and withdrawal - is hypothesized to buffer the relationship between environmental demand and distress. Direct and indirect support comes from multisensory environment and classroom manipulation studies (Unwin et al., 2022; Dyer et al., 2026; Jones et al., 2026). Route choice is an architectural extension of this agency principle proposed for future testing rather than a directly demonstrated intervention.

P5 Providing accessible choices among spaces with varying levels of social exposure may support participation by allowing learners to regulate social demand in line with context and preference (McKinlay et al., 2025). The translation of this agency theme into specific spatial options is an architectural proposition requiring empirical evaluation.

P6 Institutional policy, teacher practice, maintenance, and resource availability moderate the effectiveness and durability of neuroinclusive design interventions (Federal Ministry of Education, 2015; Alabi et al., 2026).

P7 In Ota’s tropical wet-and-dry conditions, passive ventilation and daylight strategies create potential trade-offs with acoustic, glare, and visual-distraction control; neuroinclusive design therefore requires these environmental demands to be addressed together rather than as separate technical systems (Mba et al., 2022; Xi et al., 2025; Omotosho et al., 2017).

5.4 What changed: auditing the original proposition against the evidence

The resulting audit of the original proposition is summarised in Table 3.

TABLE 3. Audit of the original single-axis Sensory Gradient proposition against the evidence synthesised in Section 4.

5.5 Evidence traceability

Evidence supporting each framework component and the degree of evidential convergence is summarised in Table 4.

TABLE 4. Evidence-to-framework audit trail for the revised framework.

6 Initial contextual application to ota: boundary conditions and resource-constrained options

Since this review is a synthesis of literature rather than new empirical fieldwork in Ota, this section is presented as an initial contextual application rather than a validation of the framework. Table 5 identifies passive-design trade-offs relevant to Ota’s tropical wet-and-dry conditions using Nigerian passive-design evidence (Mba et al., 2022; Xi et al., 2025), Ota-specific rainfall evidence (Omotosho et al., 2017), and the framework’s demand-profile logic. The cited contextual studies establish climatic or infrastructural conditions; the proposed mitigations in Table 5 are framework-derived design responses for future testing and should not be interpreted as interventions directly evaluated in those studies.

TABLE 5. Contextual passive-design trade-offs and framework-derived responses for initial application in Ota’s tropical wet-and-dry conditions.

Institutionally, Nigeria’s National Policy on Special Needs Education (Federal Ministry of Education, 2015) makes an aspirational commitment to inclusive schooling; the empirical studies in Section 4.5 and secondary commentary suggest that funding, teacher training, and physical-infrastructure delivery can fall short of that commitment (Ahmed et al., 2024; Alabi et al., 2026; Kareem, 2025). Proposition 6, therefore, functions as a boundary condition: architectural design alone cannot deliver neuroinclusive outcomes when staffing, maintenance, safeguarding, or space-management practices undermine the environment’s intended use. For example, a retreat space cannot function as a regulatory resource if it is routinely repurposed as an overflow teaching space or used punitively.

Resource constraints also mean that the framework should not be interpreted as requiring spatial abundance, duplicated circulation, or purpose-built sensory rooms. In cash-constrained schools, the same principles can be explored through lower-cost or operational adaptations: existing verandas can be structured as pause points rather than treated only as corridors; visual screens, curtains or furniture layouts can create temporary low-demand corners; timetabling can stagger high-density transitions and meal periods; classroom allocation can reduce exposure to predictable noise sources; low-maintenance vegetation can provide shade and visual buffering; and route or withdrawal choice can sometimes be created through supervision and scheduling rather than new construction. These examples are framework-derived implementation options, not interventions shown by the cited Nigerian studies to produce neuroinclusive outcomes. They should therefore be evaluated through co-design and post-occupancy testing.

7 Discussion

7.1 What this review establishes, precisely

The contribution of this review is deliberately narrower than claiming that sensory design itself is new. Tola et al., 2021 Black et al. (2022), and Wang et al. (2026) already established the importance of multiple sensory qualities, while ASPECTSS explicitly includes zoning, sequencing, and transition spaces. Broader architectural research also demonstrates that spatial configuration, wayfinding, and circulation can be analyzed as design variables in schools and other complex buildings (Arthur and Passini, 1992; Kishimoto and Taguchi, 2014; Saghafi and Mirzaei, 2021; Psathiti and Charalambous, 2025). The present contribution is therefore an organisational and theory-building synthesis: (i) it tests a pre-existing single-axis stimulus-gradient proposition and rejects that formulation in favour of a multidimensional demand-profile model; (ii) it specifies transition magnitude, predictability/legibility and learner-perceived control as potentially measurable variables; and (iii) it translates those variables into hypotheses about adjacency, buffered transitions and spatial choice across campus, building and room scales. The first two claims are grounded in the reviewed evidence at varying levels of directness; the specific campus-network configuration is an architectural proposition derived from that evidence and remains to be validated.

7.2 Strengths and limitations

This review’s principal strengths are its explicit engagement with recent built-environment syntheses, its separation of methodological quality from evidence directness, and its inclusion of participatory and experimental evidence from autistic secondary-school-age learners. The abductive audit also makes the relationship between the pre-existing design proposition and the revised framework transparent by allowing original components to be retained, modified, rejected, or added.

Several limitations remain. Only English-language literature was included, which may under-represent francophone and other African scholarship. The evidence base is strongly autism-dominant, with comparatively limited direct built-environment evidence for ADHD and particularly sparse evidence for dyslexia and dyspraxia. The framework should therefore be understood as autism-anchored in evidential support and neuroinclusive in ambition: its transfer to other neurodivergent groups is provisional. This matters because a design approach developed primarily from autism-related evidence could create tensions if applied uncritically to learners whose needs differ - for example, some ADHD presentations may involve sensory-seeking, movement, or stimulation needs, while dyslexia- or dyspraxia-related visual demands cannot be inferred from autism evidence. The application should therefore begin with learner-specific co-design rather than with diagnosis-based assumptions. Evidence located specifically in Ota or comparable Nigerian secondary schools is also sparse, so the contextual application in Section 6 relies partly on geographically proximate school studies and non-neurodivergence-specific climatic and passive-design evidence. Finally, heterogeneity in study designs and outcomes precluded statistical meta-analysis, and some framework propositions are architectural translations or theory-building inferences that require direct empirical testing.

Framework applicability is also bounded by spatial and institutional capacity. Substantial adaptation may be required for schools serving learners with profound and multiple disabilities who need highly individualized sensory programming, temporary or emergency school structures, severely space-constrained or cash-constrained urban sites, and institutions unable to protect retreat/restorative spaces from repurposing. In such settings, low-cost operational equivalents - including staggered movement schedules, temporary screens, furniture-based zoning, use of existing verandas as pause points, and carefully selected low-maintenance planting - may be more realistic than additional built floor area. These equivalents remain propositions to be tested rather than validated substitutes.

7.3 Future validation agenda

The framework is evidence-informed and internally traceable, but it is not empirically validated. Future work should therefore prioritize co-design with neurodivergent Nigerian secondary-school students; post-occupancy evaluation of schools or retrofits that operationalize the framework; direct measurement of acoustic, lighting, thermal, movement, and social-demand conditions in Ota-area schools; and prospective testing of transition magnitude, predictability/legibility, perceived control, and regulatory outcomes. Such work should examine dimension-specific environmental changes rather than assume a single stimulus score and should test whether low-cost operational adaptations provide comparable regulatory benefits where major construction is infeasible. Periodic updates of the evidence synthesis should also assess whether emerging ADHD-, dyslexia- and dyspraxia-specific built-environment research changes the framework’s current autism-dominant evidential base.

8 Conclusion

This review tested a pre-existing single-axis sensory-gradient proposition for secondary-school design against interdisciplinary evidence and found that environmental demand is better represented as multidimensional and learner-dependent than as a universal high-to-low stimulus sequence. The resulting Sensory-Gradient Framework conceptualises schools as networks of environmental-demand profiles connected through buffered transitions and spatial options, with transition magnitude, predictability/legibility and perceived control specified as testable variables across campus, building and room scales. The framework is autism-anchored in evidential support; its broader neuroinclusive application to ADHD, dyslexia, dyspraxia and other neurodivergent groups remains provisional and should be developed through learner-specific co-design and empirical testing. Its Ota component is an initial contextual application, not local validation, and explicitly recognises climatic, institutional and resource constraints, including the need for low-cost or operational alternatives where new construction is infeasible. The framework is therefore offered as an evidence-informed organisational model for participatory design, post-occupancy evaluation and empirical testing rather than as a prescriptive design standard.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During manuscript preparation, the authors used ChatGPT (OpenAI, GPT‐5.6 Sol, https://chatgpt.com) for language editing, structural refinement, and final proof checking. All AI-assisted outputs were reviewed, verified, and edited by the authors, who take full responsibility for the accuracy, originality, and integrity of the final content. Generative AI was not used to generate primary research data.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fbuil.2026.1964960/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Roberto Alonso González-Lezcano, CEU San Pablo University, Spain

Reviewed by: Heba M. Abdou, Jouf University, Saudi Arabia

Truong Nguyen Dang Hoang Nhat, Hanyang University, Republic of Korea