Measuring attitudes towards historic gardens: development and validation of the Garden Heritage Scale

Historic gardens, defined by the International Charter of Florence as architectural and horticultural compositions of historic or artistic interest, are both cultural heritage and ecologically valuable green spaces. They can offer biodiversity, mitigate urban heat, and support health, while also presenting potential conflicts between conservation aims. Such spaces are promising venues for interdisciplinary learning linking environmental, historical, and aesthetic education. Yet, research on educational uses and psychological outcomes (e.g., attitudes) in historic gardens is scarce, partly due to a lack of tailored measurement tools. Grounded in the tripartite attitude model (affect, cognition, behavior), the study set out to develop and validate the Garden Heritage Scale (GHS) to assess attitudes toward historic gardens, enabling the evaluation of educational interventions and comparison of program formats.

Attitudes are evaluative, relatively stable, latent constructs that can influence behavior and are commonly modeled with affective, cognitive, and behavioral components. While some debate alternatives (one- or two-factor structures; e.g., preservation vs. utilization), recent work supports the tripartite model and clarifies conditions under which attitudes predict intentions and behavior. Environmental attitude measures exist, but they mainly target ecological concerns and are not tailored to cultural heritage green spaces. Prior work on historic gardens/landscapes often used qualitative methods and did not explicitly apply attitude theory. Consequently, there is a gap for a quantitative, theory-driven instrument capturing affective, cognitive, and behavioral aspects of attitudes toward historic gardens.

Scale development followed psychometric best practices with emphasis on item generation, factor extraction, item reduction, and scale evaluation. Item generation: 18 items (six per component: Affect, Behavior, Cognition), with seven adapted conceptually from Sieg et al. (2018). Response format: five-point Likert scale. Exploratory Factor Analysis (EFA): Sample n=233 (143 general population adults, partially online; 90 high-school students via paper-pencil); 60.09% female, 39.06% male. Analyses in SPSS 29 with principal axis factoring, promax rotation; initial factorability confirmed by KMO=0.904. EFA was first forced to three factors based on theory; items with cross-loading differences <0.20 or with loadings contradicting the theoretical classification were removed; the affective subscale was shortened to balance subscales. A final EFA based on eigenvalues (≥1.0) was also conducted. Confirmatory Factor Analysis (CFA): Independent sample n=183 (primarily university students; 71.04% female). Missing values replaced by series means. CFA in SPSS Amos 29 with maximum likelihood estimation. Three models were compared: Model 1 (three-factor: Affect, Behavior, Cognition), Model 2 (two-factor: Behavior vs. combined Affect-Cognition), Model 3 (one-factor). Fit indices: CFI, TLI, RMSEA, SRMR, χ²/df; models compared via likelihood-ratio tests. Criterion-related validity: Combined dataset n=224 (CFA sample plus test–retest pretest). Convergent measure: adapted Nature Interest Scale (NIS; “nature” replaced by “historic gardens”) to index interest in historic gardens. Discriminant measure: adapted Zoderer & Tasser wilderness beliefs short scale (translated to German; context adjusted to wilderness in Germany). Internal consistency (Cronbach’s alpha) computed for both references; then mean scores were correlated (Pearson r) with GHS subscales. Reliability: Internal consistency (Cronbach’s alpha) computed across combined factor analysis datasets (n=416) for subscales and total scale. Test–retest reliability: panel of biology students (n1=41; n2=31 eight weeks later; mostly ages 20–30) with Pearson correlation of mean GHS scores across the two administrations.

EFA: KMO=0.904. After item reduction, balanced subscales remained; affective items loaded together, and behavior and cognition items formed distinct factors in the forced three-factor solution with high loadings (>0.590; average ≈0.779). The eigenvalue-based EFA suggested two components (Affect and Cognition combined), but the scree plot and theory supported three factors. Final retained items (Table 4) included Affect: A5 (.923), A3 (.848), A2 (.769); Behavior: B3 (.845), B5 (.669), B1 (.590); Cognition: C6 (.918), C3 (.668). CFA: Model 1 (three-factor) fit best: p=0.139, CFI=0.988, TLI=0.980, SRMR=0.0392, RMSEA=0.045, χ²=23.3, df=17, χ²/df=1.37. Model 2 (two-factor) fit worse: p<0.001, CFI=0.879, TLI=0.821, SRMR=0.0981, RMSEA=0.136, χ²/df=4.98. Model 3 (one-factor): p<0.001, CFI=0.714, TLI=0.600, SRMR=0.1200, RMSEA=0.203, χ²/df=8.5. Likelihood-ratio tests showed Model 1 significantly superior to Model 2 (Δχ²=59.5, Δdf=2) and Model 3 (Δχ²=146.7, Δdf=3), both p<0.01. Criterion-related validity: Convergent instrument alpha=0.915; discriminant instrument alpha=0.691. Correlations with convergent instrument: Affect r=0.592, Cognition r=0.430, Behavior r=0.461 (avg ≈0.494). Correlations with discriminant instrument (wilderness attitudes): Affect r=0.078, Cognition r=0.094, Behavior r=0.169 (avg ≈0.114). Reliability: Cronbach’s alpha—Affect=0.871, Cognition=0.735, Behavior=0.762, Total scale=0.846. Test–retest reliability over 8 weeks: r=0.782.

Findings support the validity of a tripartite attitude structure (affect, cognition, behavior) for attitudes toward historic gardens. Although eigenvalues from the EFA suggested two components (merging affect and cognition), the CFA provided strong evidence that the three-factor model fits the data best and is significantly superior to one- and two-factor alternatives. Subscale loadings were moderate to high, and inter-factor correlations were neither trivial nor so high as to render factors indistinct. Criterion validity patterns aligned with theory: GHS subscales correlated moderately with a conceptually related interest measure and only weakly with a distinct wilderness attitudes measure. Reliability was adequate, with internal consistency above recommended thresholds and acceptable test–retest stability for attitudes. Together, results indicate that the GHS captures distinct yet related affective, cognitive, and behavioral intention components of attitudes toward historic gardens, enabling nuanced evaluation of educational programs and contributing to environmental and heritage psychology.

The study introduces the Garden Heritage Scale (GHS), a theory-driven, psychometrically evaluated instrument to measure attitudes toward historic gardens. Through EFA, CFA, criterion validity tests, and reliability analyses, the GHS demonstrated adequate dimensionality, validity, and reliability, consistent with a three-component attitude model. The tool can support systematic assessment and improvement of educational offerings in historic green spaces and foster empirical research at the intersection of environmental education and heritage studies. Future research should examine GHS applicability and invariance across diverse target groups and contexts, compare different educational formats (e.g., guided vs. self-guided), and investigate links between reported behavioral intentions and actual behaviors.

Questionnaires used for the EFA and internal consistency assessment differed slightly in design and context items; instruments employed for criterion-related validity also varied slightly in socio-demographic sections. Later-stage questionnaires contained all 18 initial items, though only items identified as appropriate via the initial EFA were analyzed. Samples were relatively small and skewed toward young female adults, limiting representativeness and generalizability; applicability to different target groups should be interpreted cautiously.