Vertical Architecture and Mental Health: Assessment of Depressive Symptoms Among Dwellers in Apartments and Multi-Storey Houses

Abstract

Depression represents one of the most prevalent mental health challenges globally, affecting individuals across diverse populations and settings. Based on the neurogenesis-informed hypothesis that stair use may likely elevate brain-derived neurotrophic factor (BDNF) in humans that in turn may have an antidepressant effect, this study takes residential buildings as a controlled environment to test whether there is a difference in depression symptoms based on single- or multi-storey housing. This study examined associations between staying at home and depression symptoms using the Public Health Questionnaire-8 (PHQ-8) data from 128 adults in England who spend most of their time at home. Residents in single-storey flats in apartment buildings had significantly higher overall depression scores than multi-storey house residents. Among the PHQ-8 items, only Item 8, psychomotor agitation/retardation (moving or speaking too slowly, or restlessly moving around more than usual), approached but did not reach statistical significance after Bonferroni correction (p = 0.056). After adjusting for gender, age, number of residents, activity level, and income, apartment living (vs. multi-storey houses) (β = −0.362, p < 0.001) and loneliness (β = 0.221, p = 0.016) were significant independent predictors of psychomotor agitation/retardation. Future research is needed to explore this relationship using a larger sample size and to explore whether the use of stairs explains this potential relationship through a change in BDNF.

1. Introduction

Depression is among the most prevalent mental health conditions globally, affecting an estimated 280 million people across diverse populations worldwide [1]. Despite decades of research into its biological, psychological, and social determinants, the built environment has received comparatively limited attention as a modifiable risk factor. Buildings may shape mental health through affecting different neurobiological pathways, and understanding which architectural features drive this relationship is therefore a question with direct public health consequences.

The home is the primary environment of adult life, and its growing dominance makes the question of how residential architecture shapes mental health increasingly urgent. Adults spend approximately 90% of their time indoors [2], a proportion that has risen steadily over recent decades [3] and accelerated further following the COVID-19 pandemic [4]. Our previous pilot study found that individuals who identify as homebodies are at higher risk of depression, identifying the residential environment as a priority context for further investigation [5]. These findings suggest that the home can be an active determinant of mental health.

Several large cohort studies have since established that housing typology independently predicts depression risk across diverse national contexts. A Danish register-based cohort of 14,387 individuals found higher depression incidence among apartment renters than detached house residents, with approximately 8% of the excess depression risk attributable to loneliness and a further 11% mediated by indoor annoyances [6]. An Italian study of 8177 students during the COVID-19 lockdown found that smaller apartment size significantly increased depression odds, that poor indoor quality independently doubled the odds of moderate-to-severe depression, and that poor outdoor views contributed further [7]. A French CONSTANCES cohort analysis of 22,042 participants identified a dose–response relationship between room count and depression risk, with single-room apartment dwellers showing the highest rates and each additional room associated with incrementally lower risk, an effect that persisted after adjustment for income and socioeconomic status [8]. Collectively, these studies confirm that the physical characteristics of the home can increase depression risk along other dimensions. Together, this body of evidence positions housing as a structurally significant yet modifiable determinant of mental health, one that warrants deeper investigation into whether staircases may be operative through metabolic pathways [9], a hypothesis that is yet to be tested. Yet, it is needed to support the ongoing discourse on the associations between housing and mental health [10,11,12].

The Brain Booster Buildings model formalises the hypothesis that stair use, structurally imposed by multi-storey houses but absent in single-storey apartments, constitutes a predictable daily source of incidental moderate-to-vigorous physical activity sufficient to elevate serum brain-derived neurotrophic factor (BDNF), support hippocampal neurogenesis, and confer resilience against depression [9]. This architectural pathway aligns with a broader environmental affordance framework that quantifies the built environment’s capacity to sustain BDNF release by reaching the metabolic equivalents (METs) [13]. The empirical case for stair use as a health-relevant behaviour is substantial: a prospective cohort of over 442,000 UK Biobank participants found that any level of stair climbing was associated with progressively lower incidence of cardiovascular disease, type 2 diabetes, chronic obstructive pulmonary disease, and all-cause dementia, with dose–response relationships evident across outcomes [14], and a further study of nearly 390,000 UK adults identified significant sex differences in the cardiovascular protective effects of stair climbing, with women deriving more consistent protection across all levels of genetic risk [15]. The pattern of stair use also shapes its physiological consequences: randomised trials have demonstrated that brief, high-intensity stair climbing bouts, exercise snacks, can produce significant cardiorespiratory improvements in inactive adults [16], and that dispersed stair snack protocols elicit more positive affective responses and broader physical activity ripple effects than structured continuous sessions [17]. Beyond its physical effects, stair use carries a psychological dimension: perceived difficulty using stairs can be positively associated with anxiety and depression across self-report studies, while objective stair performance is negatively associated, suggesting that psychological barriers to stair use may decouple objective behaviour from its expected benefits [18]. This bidirectionality positions stair use within a broader framework of allostatic stress, in which architecturally mediated physical activity, if designed to be a stressor rather than a resource, may undermine rather than support the neurobiological resilience it is proposed to confer [19], warranting its examination as a voluntary free-living physical activity.

The theoretical basis for the stair use hypothesis lies in the relationship between physical activity, BDNF, and adult hippocampal neurogenesis. BDNF is the most abundantly expressed neurotrophin in the mammalian brain, acting through the TrkB receptor to promote neuronal survival, synaptic plasticity, dendritic branching, and the generation of new neurons in the adult dentate gyrus [20,21], a process that persists throughout the adult human lifespan at a rate of approximately 1400 neurons per day in both hemispheres [22,23]. The neurotrophin hypothesis of depression proposes that reduced BDNF-mediated neurotrophic support leads to hippocampal neuronal atrophy, impaired neurogenesis, and the emergence of depressive symptoms [24,25], a hypothesis supported by a large meta-analysis showing significantly lower serum BDNF in depressed patients, with levels normalising following effective antidepressant treatment [26,27,28]. Physical activity is among the most robustly evidenced non-pharmacological modulators of BDNF [29,30,31]. The relationship is bidirectional and mutually reinforcing; exercise raises BDNF, elevated BDNF supports neurogenesis, and sustained neurogenesis confers resilience against depression, while physical inactivity suppresses BDNF and incrementally increases depression vulnerability [32,33]. Within this neurobiological framework, the absence of a domestic staircase can remove the primary structural affordance for incidental moderate-to-vigorous physical activity within the home [9], eliminating any possible architectural affordances for physical activity.

What none of the existing studies has examined is whether this housing typology difference, in the presence and absence of verticality, predicts not just overall depression but specific symptom dimensions. The neurogenesis-informed hypothesis raises a precise question [5,9,13]: whether the absence of stair infrastructure in single-storey apartments manifests selectively in particular depressive symptom profiles. This study tests that question by examining Public Health Questionnaire-8 (PHQ-8) symptom profiles across single- and multi-storey housing typologies in 128 adults in England.

2. Materials and Methods

This analysis uses the dataset from our recent pilot study of 142 healthy adults in England [5]. A survey (see Ref. [5] for the complete survey design) was distributed to participants through a third party, Pollfish (Prodege, El Segundo, CA, USA) and attention check questions were embedded in the survey by the researcher to ensure that all received responses are of high quality. Responses that failed to correctly answer the attention check questions were excluded. Participants were asked how they would describe themselves as homebodies or venturers on a Likert scale (1 = Strong homebody, e.g., at home for 19 to 24 h, and outside for 0 to 5 h; 2 = Moderate homebody, e.g., at home for 15 to 18 h, outside for 6 to 9 h; 3 = Neutral, e.g., at home for 10 to 14 h, outside for 10 to 14 h; 4 = Moderate venturer, e.g., at home for 5 to 9 h, outside for 15 to 19 h; 5 = Strong venturer, e.g., at home for 0 to 5 h, outside for 19 to 24 h).

Data were collected after the COVID-19 pandemic lockdown, after May 2024, when the ethical approval was obtained. The anonymous online survey was distributed through a survey platform to participants aged 18+ living in England, UK. Participants provided informed consent before completing the survey and received compensation anonymously through the platform. The researcher designed the survey and set eligibility criteria, but had no direct participant contact. Responses were anonymised with IDs, while the platform provided demographic data (city, age, gender, economic status, marital status) to maintain both anonymity and sample insights.

Depression was assessed via Public Health Questionnaire-8 (PHQ-8) (8 items, scored 1 = not at all to 4 = nearly every day). The eight items capture anhedonia (1), depressed mood (2), sleep issues (3), fatigue (4), appetite changes (5), guilt/worthlessness (6), concentration (7), and psychomotor agitation or retardation (8). We used the PHQ-8 rather than the PHQ-9 because the PHQ-9 includes an item assessing suicidal ideation (Item 9), which is primarily intended for clinical settings where immediate intervention and support can be provided if needed. The PHQ-8 is more appropriate for large-scale population screenings and community-based research where such clinical infrastructure is not available. Omitting the suicidal ideation item reduces potential distress to respondents and avoids ethical complexities when immediate mental health support cannot be guaranteed. The PHQ-8 demonstrated good internal consistency in our sample (Cronbach’s α = 0.892).

House type was self-reported prior to the survey, and we aimed to achieve equal numbers of participants across groups. In our earlier study [5], based on the limitations, we excluded studio apartments (n = 14). In the final sample (mean age 43 ± 15 years; 58% female), n = 128 (apartment n = 33, terraced house n = 28, semi-detached house n = 32, detached house n = 35). To explore the associations among the PHQ items, we first used the Shapiro–Wilk test for normality and Levene’s test for homogeneity. Accordingly, Welch’s ANOVA was used for mean differences. Bonferroni correction was applied to item-level Welch’s ANOVAs to account for multiple testing. A Games–Howell post hoc test was used for pairwise comparisons of the significant PHQ items (Item 8 was the only symptom significantly predicted by house type). Bonferroni correction was applied to the item-level Welch’s ANOVAs to control for Type I error inflation due to conducting eight separate tests (one for each PHQ-8 item; adjusted alpha = 0.05/8 = 0.00625). A Games–Howell post hoc test was selected for pairwise comparisons because it does not assume homogeneity of variance.

Assumptions for linear regression were checked before including covariates: gender, age, number of residents, activity level, and loneliness. Activity level was assessed using a Likert scale question: “Over the past 4 weeks at your house, were you mostly…” The answer options were as follows: 1 = Sedentary (sitting, lying down, or engaged in activities with minimal movement), 2 = Lightly active (standing, light housework, or activities with occasional walking), 3 = Moderately active (regular housework, gardening, or activities with frequent walking), or 4 = Very active (physically demanding housework, exercise, or activities with continuous movement). Loneliness was assessed using a Likert scale question as well: At your house over the past 4 weeks, how often have you felt lonely? The answers options were 1 = Never, 2 = Rarely, 3 = Sometimes, 4 = Often, and 5 = Very Often.

Two participants were excluded from regression analyses due to missing income data (final n = 126). Using G*Power 3.1, post hoc power analysis for PHQ-8 Item 8 (psychomotor agitation/retardation), the only item that approached significance after Bonferroni correction (adjusted p = 0.056), indicated that with n = 126, α = 0.05, and seven predictors in the model, the study achieved power of 0.986 to detect the observed effect size (f² = 0.14), representing a small-to-medium effect. This observed effect size corresponds to housing type (apartments compared to multi-storey houses), uniquely explaining 10.4% of variance in psychomotor symptoms beyond other covariates (ΔR² = 0.104).

The Mann–Whitney U test was conducted to compare loneliness scores between single-storey apartment and multi-storey house residents (Shapiro–Wilk < 0.05). All statistical analyses were performed using SPSS v.30; α = 0.05.

3. Results

Demographic characteristics are presented in

Table 1. Demographic characteristics by housing type.

	Apartment (n = 33)	Terraced House (n = 28)	Semi-Detached House (n = 32)	Detached House (n = 35)
Age (years), M ± SD	40.9 ± 15.4	43.1 ± 11.8	38.7 ± 11.6	48.6 ± 17.5
Female, (%)	66.7%	60.7%	59.4%	45.7%
Number of residents, M ± SD	2.1 ± 1.5	3.0 ± 1.3	3.4 ± 1.2	2.8 ± 1.3
Residency length in years, M ± SD	5.6 ± 1.9	7.2 ± 2.1	6.8 ± 1.8	7.2 ± 1.6
Homebody(1)/Venturer(4), M ± SD	1.9 ± 1.1	2.2 ± 0.9	2.4 ± 0.9	2.3 ± 1.0
Income level	2.1 ± 1.3	2.3 ± 1.3	2.8 ± 1.4	3.2 ± 1.4
Sedentary(1)/Active(4), M ± SD	2.0 ± 0.8	2.0 ± 0.9	2.5 ± 0.7	2.4 ± 0.9
Loneliness	2.8 ± 1.3	2.3 ± 1.1	2.2 ± 1.0	2.2 ± 1.1

Note: M = mean; SD = standard deviation.

and show similar patterns across the different house types. The overall depression mean showed significant differences by house type (Welch’s ANOVA F(3, 68.235) = 3.420, p = 0.022, omega-squared (ω2) = 0.070). These results indicate a significant medium effect size. Games–Howell post hoc (

Table 2. Games–Howell post hoc for depression mean differences.

(I) House Type	(J) House Type	Mean Difference	Std. Error	Sig.	95% Confidence Interval
					Lower Bound	Upper Bound
Apartment	Terraced house	0.47506	0.18099	0.052	−0.0035	0.9536
	Semi- detached house	0.49645 *	0.17319	0.028	0.0387	0.9542
	Detached house	0.49735 *	0.18007	0.037	0.0223	0.9724
Terraced house	Apartment	−0.47506	0.18099	0.052	−0.9536	0.0035
	Semi- detached house	0.02138	0.15902	0.999	−0.3996	0.4424
	Detached house	0.02229	0.16648	0.999	−0.4176	0.4622
Semi- detached house	Apartment	−0.49645 *	0.17319	0.028	−0.9542	−0.0387
	Terraced house	−0.02138	0.15902	0.999	−0.4424	0.3996
	Detached house	0.0009	0.15797	1.000	−0.4157	0.4175
Detached house	Apartment	−0.49735 *	0.18007	0.037	−0.9724	−0.0223
	Terraced house	−0.02229	0.16648	0.999	−0.4622	0.4176
	Detached house	−0.0009	0.15797	1.000	−0.4175	0.4157

* The mean difference is significant at the 0.05 level.

) indicated single-storey flats or apartments had higher scores than semi-detached houses (mean diff = 0.496, p = 0.028) and detached houses (mean diff = 0.497, p = 0.037), but it approached yet did not reach statistical significance with terraced houses (mean diff = 0.475, p = 0.052).

After checking for normality and homogeneity of variance differences, and given the sample size, we used Welch’s ANOVA. Item-level Welch’s ANOVAs revealed that only Item 4 (p = 0.047) and Item 8 (p = 0.007) were significant (

Table 3. Welch’s ANOVA results for differences in Public Health Questionnaire-8 (PHQ-8) item scores by housing type.

	Statistic	df1	df2	Sig.
PHQ-1	1.534	3	67.841	0.214
PHQ-2	2.246	3	67.317	0.091
PHQ-3	1.784	3	67.452	0.159
PHQ-4	2.788	3	68.186	0.047
PHQ-5	2.152	3	67.385	0.102
PHQ-6	1.867	3	67.818	0.143
PHQ-7	2.424	3	67.661	0.073
PHQ-8	4.390	3	66.768	0.007

); others ranged from p = 0.091 (Item 2: depressed mood) to p = 0.214 (Item 1: anhedonia). After Bonferroni correction across items, only Item 8 (psychomotor agitation/retardation) approached significance (adjusted p = 0.056), warranting post hoc examination.

For Item 8, Games–Howell results (

Table 4. Games–Howell post hoc results for PHQ item 8 (psychomotor agitation/retardation).

(I) House Type	(J) House Type	Mean Difference	Std. Error	Sig.	95% Confidence Interval
					Lower Bound	Upper Bound
Apartment	Terraced house	0.626 *	0.173	0.005	0.16	1.09
	Semi- detached house	0.541 *	0.184	0.025	0.05	1.03
	Detached house	0.526 *	0.182	0.03	0.04	1.01
Terraced house	Apartment	−0.626 *	0.173	0.005	−1.09	−0.16
	Semi- detached house	−0.085	0.093	0.801	−0.33	0.16
	Detached house	−0.1	0.091	0.693	−0.34	0.14
Semi- detached house	Apartment	−0.541 *	0.184	0.025	−1.03	−0.05
	Terraced house	0.085	0.093	0.801	−0.16	0.33
	Detached house	−0.015	0.11	0.999	−0.31	0.28
Detached house	Apartment	−0.526 *	0.182	0.03	−1.01	−0.04
	Terraced house	0.1	0.091	0.693	−0.14	0.34
	Semi- detached house	0.015	0.11	0.999	−0.28	0.31

* The mean difference is significant at the 0.05 level.

) showed single-storey apartment residents to be higher in psychomotor agitation/retardation than residents of terraced houses (mean diff = 0.626, p = 0.005), semi-detached houses (mean diff = 0.541, p = 0.025), and detached houses (mean diff = 0.526, p = 0.030). No other pairwise differences were significant between the multi-storey houses. These post hoc findings support the association between apartment living and elevated psychomotor agitation/retardation.

While p = 0.056 technically exceeds the alpha = 0.05 threshold set a priori, the marginal nature of this result (just 0.006 above the cutoff) and the medium effect size (omega-squared = 0.070) suggest this finding warrants cautious interpretation rather than outright dismissal. The post hoc pairwise comparisons and regression analyses should be viewed as exploratory, hypothesis-generating analyses rather than confirmatory tests. Future research with larger samples and preregistered hypotheses is needed to determine whether this signal represents a robust, replicable effect.

Accordingly, this study computed a new binary variable (apartment vs. multi-storey houses combined) before conducting multiple regressions.

Assumptions of multiple linear regression were verified for the models predicting psychomotor agitation/retardation (PHQ-8 Item 8) through single-storey apartment vs. multi-storey houses (binary; 1 = apartment, 2 = houses), gender (binary; 1 = females, 2 = males), age, number of residents, inactivity vs. activity level (1 = sedentary, 4 = very active), loneliness (subjectively reported from 1 = never to 5 = very often), and income (low to high). Multicollinearity was assessed using variance inflation factors (VIF), with all values ranging from 1.077 to 1.718 (below the threshold of 5), indicating no substantial multicollinearity among predictors. Independence of errors was confirmed via the Durbin–Watson statistic (2.236), which was close to 2 and within acceptable limits (1.5–2.5). Normality of residuals was evaluated using histograms, which showed an approximately normal distribution (M ≈ 0, SD = 0.972), with minor right skew deemed acceptable given the sample size (n = 128). Linearity and homoscedasticity were examined via scatterplots of standardised residuals against standardised predicted values, revealing a slight, non-severe downward trend and an uneven spread, suggestive of mild heteroscedasticity. One potential outlier was noted (standardised residual > 3), but Cook’s distance values (<1) indicated no influence on the model. These checks confirmed that the assumptions were largely met, supporting the validity of the regression analyses. The regression model was significant (p < 0.001), and the adjusted R-square indicated that the model explained 20.8% of the variance in Item 8 (psychomotor agitation/retardation). As shown in

Table 5. Regression model for PHQ Item 8 (psychomotor agitation/retardation) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	2.141 *	0.417		5.14	<0.001
Housing type (apartment/multi-storey house)	−0.525 *	0.13	−0.362	−4.048	<0.001
Gender (f/m)	−0.082	0.107	−0.064	−0.772	0.442
Age	−0.004	0.004	−0.1	−0.981	0.329
Number of residents	0.046	0.048	0.101	0.969	0.334
Activity level	−0.043	0.066	−0.057	−0.652	0.516
Loneliness	0.124 *	0.051	0.221	2.442	0.016
Income	0.007	0.039	0.015	0.172	0.863

Note: R² = 0.252, F(7, 118) = 5.687, p < 0.001. Housing type (single-storey apartment vs. multi-storey house) uniquely contributed ΔR² = 0.104 beyond other covariates, corresponding to an effect size of f² = 0.14 (small-to-medium effect). * significant at the 0.05 level.

, the single-storey apartment vs. multi-storey house variable was the strongest significant predictor (p < 0.001), followed by loneliness (p = 0.016). No other variables were significant predictors. After Bonferroni correction, only the house type remains a significant predictor (p < 0.00625).

Consequently, a nonparametric independent-samples Mann–Whitney U test was employed to compare loneliness scores between single-storey apartment residents (n = 33, mean rank = 75.65) and multi-storey house residents (n = 95, mean rank = 60.63). The test revealed a statistically significant difference, U = 1199.500, p = 0.037 (two-tailed), with apartment residents reporting higher loneliness. The finding that apartment residents experience significantly higher loneliness (p = 0.037) provides additional context for understanding the relationship between housing type and psychomotor symptoms. This suggests that apartments may trigger depression symptoms not only through the physical space but also through loneliness, which is supported through the analysis of the remaining PHQ symptoms (

Table A1. Regression model for PHQ Item 1 (anhedonia) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	2.277 *	0.585		3.893	<0.001
Housing type (apartment/multi-storey house)	−0.136	0.182	−0.068	−0.745	0.458
Gender (f/m)	0.073	0.150	0.041	0.487	0.628
Age	−0.007	0.006	−0.113	−1.088	0.279
Number of residents	−0.013	0.067	−0.020	−0.192	0.848
Activity level	−0.067	0.093	−0.064	−0.714	0.476
Loneliness	0.209 *	0.072	0.272	2.927	0.004
Income	−0.166 *	0.055	−0.266	−2.990	0.003

* significant at the 0.05 level.

,

Table A2. Regression model for PHQ Item 2 (depressed mood) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	1.477 *	0.497		2.971	0.004
Housing type (apartment/multi-storey house)	−0.175	0.155	−0.086	−1.130	0.261
Gender (f/m)	−0.223	0.127	−0.124	−1.749	0.083
Age	0.002	0.005	0.030	0.344	0.731
Number of residents	−0.053	0.057	−0.082	−0.923	0.358
Activity level	−0.017	0.079	−0.016	−0.210	0.834
Loneliness	0.459 *	0.061	0.586	7.551	<0.001
Income	−0.068	0.047	−0.107	−1.435	0.154

* significant at the 0.05 level.

,

Table A3. Regression model for PHQ Item 3 (sleep disturbances) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	2.444 *	0.657		3.718	<0.001
Housing type (apartment/multi-storey house)	0.073	0.205	0.030	0.356	0.723
Gender (f/m)	−0.289	0.168	−0.134	−1.715	0.089
Age	0.007	0.007	0.098	1.017	0.311
Number of residents	−0.173	0.075	−0.227	−2.302	0.023
Activity level	−0.234 *	0.105	−0.184	−2.229	0.028
Loneliness	0.334 *	0.080	0.356	4.150	<0.001
Income	−0.076	0.062	−0.100	−1.213	0.228

* significant at the 0.05 level.

,

Table A4. Regression model for PHQ Item 4 (fatigue) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	2.674 *	0.703		3.806	<0.001
Housing type (apartment/multi-storey house)	−0.246	0.219	−0.100	−1.123	0.264
Gender (f/m)	−0.413	0.180	−0.188	−2.294	0.024
Age	0.016	0.007	0.216	2.132	0.035
Number of residents	−0.069	0.080	−0.089	−0.855	0.394
Activity level	−0.183	0.112	−0.142	−1.635	0.105
Loneliness	0.276 *	0.086	0.290	3.216	0.002
Income	−0.057	.067	−0.073	−0.850	0.397

* significant at the 0.05 level.

,

Table A5. Regression model for PHQ Item 5 (appetite changes) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	3.002 *	0.636		4.718	<0.001
Housing type (apartment/multi-storey house)	−0.181	0.198	−0.081	−0.913	0.363
Gender (f/m)	−0.483 *	0.163	−0.243	−2.965	0.004
Age	−0.001	0.007	−0.020	−0.201	0.841
Number of residents	−0.037	0.073	−0.052	−0.502	0.617
Activity level	−0.301 *	0.101	−0.256	−2.962	0.004
Loneliness	0.229 *	0.078	0.264	2.942	0.004
Income	0.031	0.060	0.044	0.517	0.606

* significant at the 0.05 level.

,

Table A6. Regression model for PHQ Item 6 (guilt or self-worthlessness) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	1.309 *	0.575		2.275	0.025
Housing type (apartment/multi-storey house)	−0.184	0.179	−0.082	−1.027	0.307
Gender (f/m)	−0.296 *	0.147	−0.149	−2.007	0.047
Age	0.009	0.006	0.139	1.525	0.130
Number of residents	−0.006	0.066	−0.009	−0.092	0.927
Activity level	−0.099	0.092	−0.084	−1.079	0.283
Loneliness	0.481 *	0.070	0.556	6.839	<0.001
Income	−0.066	0.055	−0.094	−1.208	0.230

* significant at the 0.05 level.

and

Table A7. Regression model for PHQ Item 7 (concentration difficulties) predictors.

	Unstandardised B	Coefficients Std. Error	Standardised Coefficient Beta	t	Sig.
(Constant)	2.581 *	0.552		4.674	<0.001
Housing type (apartment/multi-storey house)	−0.258	0.172	−0.130	−1.503	0.136
Gender (f/m)	−0.361 *	0.142	−0.204	−2.551	0.012
Age	0.003	0.006	0.055	0.555	0.580
Number of residents	−0.036	0.063	−0.058	−0.575	0.566
Activity level	−0.248 *	0.088	−0.238	−2.813	0.006
Loneliness	0.238 *	0.068	0.309	3.523	<0.001
Income	−0.003	0.052	−0.005	−0.062	0.951

* significant at the 0.05 level.

in the Appendix A).

4. Discussion

This exploratory study examined associations between residential housing typology and specific depressive symptom profiles in 128 adults in England who spend the majority of their time at home. Residents of single-storey apartments reported significantly higher overall depression scores than residents of multi-storey houses. At the item level, psychomotor agitation and retardation was the only Public Health Questionnaire-8 (PHQ-8) symptom to approach significance after Bonferroni correction (adjusted p = 0.056), and remained a significant independent predictor of apartment living after full covariate adjustment including gender, age, number of residents, activity level, income, and loneliness (β = −0.362, p < 0.001). The selectivity of this association, psychomotor symptoms specifically, while seven other depressive symptom dimensions were unaffected, is the central finding of this study and warrants careful interpretation. In addition to the difference between single- and multi-storey housing, loneliness was the only other significant predictor of psychomotor agitation and retardation (β = 0.221, p = 0.016), and apartment residents reported higher loneliness than multi-storey house residents. It can be hypothesised that apartment living was associated with higher loneliness, which may reflect reduced incidental social encounters, such as where lift use replaces stair use as the primary vertical circulation mode, reducing the informal social interactions that stair use in shared residential environments affords. The convergence of reduced physical activity infrastructure and increased loneliness in the apartment typology may compound psychomotor symptom risk through partially overlapping but independent pathways.

The present findings extend the existing cohort literature by adding that vertical architecture can be a candidate predictor of depression that operates through a mechanism distinct from those previously identified. Prior studies established that room count, apartment size, indoor quality, outdoor views, and layout organisation each independently predict depressive symptoms [6,7,8]; this study suggests that the presence or absence of a domestic staircase may constitute an additional and neurobiologically distinct predictor, one whose effect appears selectively on psychomotor symptoms rather than on the cognitive–affective symptom dimensions that loneliness and spatial confinement appear to drive. It is important that future research takes into account the possibility of staircase presence and stair use patterns contributing towards depressive symptoms among dwellers.

The specificity of the psychomotor finding is architecturally meaningful. Among the eight PHQ-8 symptom dimensions, psychomotor agitation and retardation is the one most directly linked to physical inactivity and its neurobiological consequences. The absence of stair infrastructure in single-storey apartments may limit the most reliable source of incidental moderate-to-vigorous physical activity within the residential environment, an activity that the Brain Booster Buildings framework has modelled as capable of predictably elevating serum BDNF through architecturally mediated stair use [9]. The association with psychomotor symptoms, rather than with any other symptom, is consistent with this proposed pathway. In other words, psychomotor retardation is a reduction in activity, while agitation may not be supported by the lack of sufficient walkability and the absence of stairs. It is important to note, however, that stair use, BDNF, and objective physical activity were not directly measured in this study. The stair-based physical activity hypothesis therefore remains proposed rather than confirmed, and represents the primary direction for future empirical investigation.

The proposed mechanism linking residential building typology to psychomotor symptoms operates through the neurobiological chain introduced earlier, centred on BDNF, hippocampal neurogenesis, and the dopaminergic cortico-striatal circuits implicated in psychomotor function [20,21,22,23,24,25]. Building on this framework, the clinical evidence is substantial: serum BDNF is significantly lower in depressed patients and normalises following effective antidepressant treatment across multiple meta-analyses [26,27,28], and both acute and chronic exercise significantly increase circulating BDNF levels across training modalities [29,30,31]. The relationship is bidirectional and mutually reinforcing, exercise raises BDNF, elevated BDNF supports neurogenesis, and sustained neurogenesis confers resilience against depression, while physical inactivity suppresses BDNF and incrementally increases depression vulnerability [32,33]. Within this mechanistic framework, the architectural difference between single-storey apartments and multi-storey houses is neurobiologically relevant insofar as it may systematically reduce exposure to the category of physical activity most consistently associated with acute BDNF elevation. Whether chronic residence in a stair-free environment translates into measurable suppression of BDNF or attenuation of hippocampal neurogenesis remains an empirical question that the present study cannot answer directly (Figure 1).

The selectivity of the psychomotor finding, one symptom dimension out of eight reaching significance while the remaining seven did not, is not a statistical artefact to be explained away but rather the most theoretically informative result of this study. Within the PHQ-8 structure, the item of psychomotor agitation and retardation occupies a distinct position from the other items. Factor analytic studies of the PHQ-9 identify a two-factor solution separating somatic symptoms, including sleep disturbance, fatigue, appetite changes, and psychomotor disturbance, from cognitive–affective symptoms, comprising anhedonia, depressed mood, guilt, and concentration difficulties [34]. Examining the regression models in the appendix of this study reveals precisely this pattern: loneliness emerged as a significant predictor of anhedonia, depressed mood, guilt, and concentration difficulties, while housing type did not reach significance for any of these items. Psychomotor agitation and retardation represented the sole symptom for which housing type, rather than loneliness or any other covariate, was the dominant independent predictor. This dissociation implies that the cognitive–affective symptom dimensions of depression in this sample were primarily driven by social and psychological factors, while the psychomotor dimension was driven by something structurally environmental, consistent with the proposed architectural pathway towards BDNF and neurogenesis.

The neurobiological distinctiveness of psychomotor symptoms further supports this interpretation. Unlike depressed mood and rumination, which are associated with default mode network hyperconnectivity [35,36], psychomotor retardation is rooted in dopaminergic dysregulation of cortico-striatal motor circuits, particularly involving the basal ganglia and prefrontal cortex [37,38,39]. Physical activity, and aerobic exercise in particular, represents a well-evidenced non-pharmacological mechanism for boosting dopamine transmission and reducing systemic inflammation, both of which are implicated in psychomotor symptom severity [39,40,41]. A framework proposed by Hird and colleagues articulates the central pathway: aerobic exercise reduces systemic inflammation and boosts dopamine transmission, with consequent improvements in effort-based decision-making and propensity to exert effort, thereby specifically improving the interest–activity cluster of depressive symptoms, encompassing anhedonia, fatigue, and subjective cognitive impairment [41]. Complementary work by Felger and colleagues further demonstrates that inflammatory cytokines preferentially target basal ganglia dopamine function to produce both motivational and psychomotor deficits, providing a direct mechanistic bridge between this exercise framework and psychomotor symptom severity specifically [39,40]. The absence of stair infrastructure in single-storey apartments therefore may remove a daily source of dopaminergic stimulation that multi-storey house residents accumulate incidentally, potentially leaving the cortico-striatal motor circuitry of apartment residents chronically under-stimulated in ways that could specifically predispose to psychomotor, rather than cognitive-affective, depressive symptoms.

The relationship between physical inactivity and depression is not unidirectional, and understanding this bidirectionality is essential for interpreting the architectural hypothesis this study advances. In the context of single-storey apartment living, the absence of staircases does not simply reduce one source of incidental physical activity in isolation. Rather, it removes the primary architectural mechanism by which the inactivity–depression loop might be broken on a daily basis without deliberate effort or external motivation. Depression reduces the motivation and energy available for voluntary, discretionary exercise, meaning that the most unwell individuals are precisely those least likely to compensate for an activity-poor environment through intentional physical activity [42,43]. The architecture of a multi-storey house, in contrast, imposes a degree of mandatory vertical movement that is non-optional and non-discretionary, occurring as a structural consequence of daily domestic life regardless of motivational state. The chronicity of this loop matters further when considered through the lens of allostatic overload. Prolonged exposure to a home environment that both limits physical activity and fails to interrupt the depression–inactivity cycle may constitute a sustained low-grade stressor on neurobiological systems, gradually depleting the BDNF reserves that support hippocampal neurogenesis and stress resilience [19].

Several important design implications follow from these findings. The selective association between apartment living and psychomotor symptoms suggests that interventions targeting incidental physical activity infrastructure within residential buildings may be relevant to mental health outcomes. Encouraging stair use over lift use through design, through stair visibility, aesthetics, and placement, has been identified as an effective strategy for increasing physical activity among building occupants [44,45,46]. Integrating indoor-outdoor connections that enable easier access to green space and walking may also address the outdoor access deficit associated with apartment living [47].

This study has some limitations that must be acknowledged. The current study is also cross-sectional, precluding causal inference. The Bonferroni-corrected p-value for Item 8 (adjusted p = 0.056) approached but did not reach conventional significance, and the finding should be treated as exploratory and hypothesis-generating rather than confirmatory. However, post hoc power analysis indicated that the regression model predicting PHQ-8 Item 8 achieved high statistical power (0.986) to detect the observed effect size (f² = 0.14) with n = 126, seven predictors, and α = 0.05. This suggests that the sample, while modest in absolute terms, was adequately powered to detect the reported association between apartment living and psychomotor symptoms. The primary statistical challenge therefore lies not in insufficient power but in the conservative nature of the Bonferroni correction applied across eight correlated PHQ-8 items. Bonferroni correction assumes independence between tests, an assumption that is not fully met when items belong to the same validated scale measuring a common underlying construct. The adjusted p-value of 0.056 for Item 8 should therefore be interpreted in the context of this conservatism, and the finding treated as a directional signal warranting replication rather than a null result. Additionally, some confounding variables were not assessed, such as including floor level for apartment residents, counting frequency of stair use, and accounting for the number of stories climbed in a single ascent. BDNF levels and objective physical activity were not measured, limiting the ability to test the proposed mechanism directly. Lastly, the sample was recruited from England, limiting generalisability to other housing markets and cultural contexts. Replication in larger, preregistered, longitudinal studies with objective physical activity measurement is essential before drawing policy or design conclusions.

To address these limitations, future studies should adopt several methodological improvements. First, a longitudinal or quasi-experimental design, such as tracking residents before and after relocation between housing typologies, would permit stronger causal inference regarding the direction of the relationship between housing type and psychomotor symptoms. Second, objective measurement of stair use (e.g., via accelerometry or dedicated step counters), serum BDNF assays, and clinician-rated psychomotor assessments should be incorporated alongside self-report PHQ-8 data to directly test the proposed mechanistic pathway rather than inferring it from cross-sectional associations. Third, floor level, stair availability, and actual stair use patterns should be recorded for all participants, to disentangle the effects of building typology from the specific physical access to staircases and use within each dwelling. Taken together, these improvements would allow the hypothesis of the impact of stair use on mental health to be tested with the rigour required before drawing policy or design conclusions.

5. Conclusions

This study examined associations between residential housing typology and depressive symptom profiles in 128 adults in England who spend the majority of their time at home. Residents of single-storey apartments reported significantly higher overall depression scores than residents of multi-storey houses. At the item level, psychomotor agitation and retardation was the only Public Health Questionnaire-8 (PHQ-8) symptom dimension to approach significance as a function of housing type after Bonferroni correction, and remained the strongest significant independent predictor of apartment living after full covariate adjustment. Loneliness was the only other significant predictor of psychomotor symptoms, and apartment residents reported significantly higher loneliness than multi-storey house residents, suggesting that social and neurobiological pathways may compound one another within this typology.

The selectivity of this association, psychomotor symptoms specifically, while seven other depressive symptom dimensions were unaffected, is the central finding of this study. It is consistent with the neurogenesis-informed hypothesis that the absence of stair infrastructure in single-storey apartments removes a daily source of incidental physical activity with downstream consequences for BDNF, hippocampal neurogenesis, and specifically the dopaminergic cortico-striatal circuits implicated in psychomotor function. The dissociation between psychomotor and cognitive–affective symptom dimensions, the former predicted by housing type, the latter primarily by loneliness, suggests that architectural and social pathways may act on distinct neurobiological substrates within depression, rather than contributing to a uniform elevation of symptom burden.

These findings contribute to the growing cohort literature linking residential housing characteristics to depression risk, adding vertical architecture, specifically the presence or absence of a domestic staircase, as a candidate variable that may operate through a neurobiological pathway distinct from those previously identified, including room count, indoor quality, ownership status, and layout organisation. The finding also raises a broader principle for architectural research: that mental health consequences of built environments may be symptom-specific rather than global.

However, these findings are exploratory and carry important limitations. While the sample is powered to detect the observed medium effect size, the design is cross-sectional and precludes causal inference, stair use frequency and serum BDNF were not measured, and the Bonferroni-corrected p-value of 0.056 for Item 8 did not reach conventional significance. Replication in larger, longitudinal, preregistered studies incorporating objective stair use measurement, serum BDNF assays, and clinician-rated psychomotor assessments is essential before design or policy conclusions can be drawn.

Additionally, future research should account for how time spent outside the home is distributed across daily life, as variation in outdoor activity patterns may independently influence any prospective association between stair use and neurobiological outcomes. If replicated, these findings would implicate the domestic staircase as a modifiable architectural feature with neurobiological consequences that accumulate gradually over years of daily residential life, likely to increase the risk of depression, especially in the form of psychomotor agitation or retardation.