The role of cultural memory and place attachment in designing cultural heritage tourism

Abstract

Cultural heritage tourism increasingly aims to provide not only recreation but also cultural learning, identity formation, and immersive engagement. This study examines how cultural memory, place attachment, and Theory of Planned Behavior (TPB) constructs jointly relate to visitors’ experiences and design-oriented insights at Sanhe Ancient Town, China. We develop and empirically examine a Cultural–Affective–Behavioral Intention model that links three cultural-memory pathways to place attachment, TPB components, and an overall cultural heritage tourism experience index. Cross-sectional survey data from 420 on-site visitors were analysed using structural equation modelling and exploratory Random Forest analysis. SEM results indicate that symbolic and narrative pathways are positively associated with place attachment, which in turn is positively associated with TPB constructs and behavioural intention and shows a positive association with higher levels of self-reported tourism experience, whereas the ritual pathway does not exhibit a significant linear effect on attachment. Random Forest analyses further highlight the central roles of behavioural intention, place attachment, and narrative cues while suggesting additional, nonlinear contributions of ritual elements. Taken together, the findings highlight cultural memory and place attachment as core mechanisms for shaping meaningful, sustainability-oriented heritage experiences and offer theory-informed implications for visual and narrative strategies at heritage destinations.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-44178-0.

Introduction

Cultural heritage tourism has become one of the fastest-growing sectors in global tourism, offering not only recreation but also cultural learning, identity formation, and immersive engagement. Visitors increasingly seek experiences that create meaningful connections between personal encounters and collective cultural narratives¹. Unlike conventional sightseeing, heritage tourism operates as a multidimensional environment in which cultural participation, historical interpretation, and community identity interact to support sustainable destination development^2,3. As a result, experience-centered and visitor-oriented design approaches have gained importance in enhancing immersion, strengthening cultural identity, and improving long-term engagement with heritage sites^4,5.

Interpretation research has increasingly emphasized that the cultural, social, and psychological value of heritage tourism relies on well-designed narrative and visual communication strategies⁶. Co-creation and participatory engagement have been shown to reinforce cultural identity and elevate experiential quality^7,8. Digital media, including augmented reality (AR), virtual reality (VR), and mixed-reality storytelling, further expand these possibilities by enabling multisensory immersion, interactive learning, and personalized interpretation^9–11. Across this literature, cultural narrative consistently emerges as a key mechanism for visitor engagement and meaning-making^12–14.

Despite these advances, many heritage sites continue to rely on repetitive, presentation-oriented communication formats. Such standardized practices often reduce visitor resonance, weaken cultural identity, and fail to convey the unique historical significance of place^14,15. More importantly, the integration of visual expression and narrative interpretation has rarely been grounded in a systematic psychological framework^16,17. Existing models that combine place attachment with the Theory of Planned Behavior (TPB) largely emphasize environmental attitudes, satisfaction, or pro-environmental behaviors rather than the role of cultural memory pathways, while memory-based tourism studies seldom embed their constructs within a full TPB structure. As a result, current design practices lack theoretical clarity regarding how cultural cues stimulate emotional attachment, influence cognitive evaluations, and shape behavioral intentions in heritage contexts. This reveals a research gap: the limited development of validated models that explicitly link cultural memory pathways, place-based emotion, and planned behavior mechanisms.

To address this gap, the present study develops an integrated Cultural–Affective–Behavioral Intention framework that draws on cultural memory theory, place attachment, and TPB. Cultural memory organizes symbolic, ritual, and narrative representations that allow visitors to interpret heritage through shared cultural frameworks¹⁸. Place attachment adds an affective dimension that embeds authenticity and transforms heritage interpretation into personally meaningful encounters^19,20. Together, these perspectives offer a theoretical foundation for understanding how cultural narratives and visual cues influence attitudes, subjective norms, perceived behavioral control, and ultimately behavioral intentions.

Building on these foundations, this study addresses the following research questions:

RQ1: How do different pathways of cultural memory influence cultural heritage tourism experiences?

RQ2: How does place attachment shape visitors’ attitudes, subjective norms, perceived behavioral control, and behavioral intentions?

RQ3: How do cultural memory pathways and place attachment jointly inform visual and narrative design strategies in heritage tourism?

RQ1 and RQ2 are examined empirically through a cross-sectional survey and Structural Equation Modeling (SEM) complemented by Random Forest analysis, whereas RQ3 is explored conceptually by mapping the empirically supported pathways onto a VR-based narrative design prototype.

This study makes three main contributions.

First, an integrated and falsifiable model. We formally specify and empirically test a Cultural–Affective–Behavioral Intention model that links the three cultural-memory pathways (symbolic, ritual, narrative) to place attachment and, subsequently, to the TPB components (attitude, subjective norm, perceived behavioral control), behavioral intention, and cultural-heritage tourist experience (Fig. 1). In contrast to prior PA–TPB integrations that typically treat place attachment as a direct predictor of behavioral intention or one of several additional antecedents, our framework positions cultural memory pathways as upstream representational mechanisms and place attachment as an affective bridge between cultural cues and planned behavior. To ensure transparency and refutability, ten directional hypotheses (H1–H10) are evaluated with standard errors and 95% bias-corrected confidence intervals.

Fig. 1.

Conceptual model.

Second, we examine place attachment as a core affective mechanism that is allowed to shape all three TPB antecedents (PA → AB/SN/PBC) in a cultural-heritage tourism setting, rather than treating PA only as a parallel predictor of behavioural intention. This specification extends existing PA–TPB models by positioning attachment as an integrative bridge between cultural‐memory pathways and TPB components, while, as discussed later, coefficients involving PA should be interpreted with caution given the single-site context and the measurement properties of the PA scale in this study (i.e., limited statistical generalizability of effect magnitudes).

Third, a theory-informed design mapping in VR. To connect the proposed framework to practical experience design, we develop a conceptual VR narrative prototype that illustratively aligns each cultural memory pathway with plausible design tactics and their proximal behavioral mechanisms. In this prototype, symbolic cues (SP) are associated with visual symbol mappings of architecture, colors, and motifs intended to strengthen place identity and shape attitudes (AB); ritual elements (RP) are associated with an embodied incense-lighting interaction at Xianyue Pavilion intended to enhance participatory agency and perceived behavioral control (PBC); and narrative cues (NP) are associated with a story-driven river cruise with spatial audio intended to deepen meaning-making and perceived social norms (SN). This pathway → mechanism → TPB mapping is presented purely as a conceptual design artifact—a theory-informed matrix for generating design hypotheses and guiding future A/B tests or experimental evaluations—rather than as additional empirical evidence about the effectiveness of specific VR interventions in this study. Across these contributions, we position the model and the pathway → mechanism mapping as an analytically transferable template (i.e., a testable scaffold that can be adapted and re-estimated across heritage contexts), while treating the current parameter estimates as context-bound and not statistically generalizable beyond the single-site sample.

To examine these questions, we develop a conceptual model consisting of ten hypotheses that articulate the cultural, emotional, and cognitive mechanisms of heritage experience formation. A cross-sectional survey with 420 visitors at Sanhe Ancient Town, a nationally designated 5 A cultural heritage site in China, tests the proposed pathways for RQ1 and RQ2. SEM evaluates the relationships among cultural memory, place attachment, and TPB constructs, while Random Forest analysis provides complementary, exploratory evidence through variable-importance profiles and out-of-sample predictive performance.

Literature review and research hypothesis

The concept of cultural memory has been widely examined in the humanities and social sciences. Halbwachs introduced the idea of collective memory, emphasizing that individual recollection is embedded within the symbolic and cognitive frameworks of social groups²¹. Building on this foundation, Assmann and Czaplicka distinguished cultural memory as an institutionalized form of long-term memory maintained through texts, images, monuments, and rituals¹⁸. Erll further conceptualized cultural memory as dynamic and constantly reconstructed through diverse media, including journalism, art, museums, and academic discourse²². This perspective shifted the study of memory from a notion of static preservation to one of continuous reinterpretation.

Olick and Robbins described memory as a socio-historical practice, transforming past events into narrative resources for the present through repeated performances²³. Rigney argued that memory is always in motion, traveling across platforms and reshaped in new contexts²⁴. These mechanisms support cultural transmission while reinforcing identity and belonging. From a narrative perspective, cultural memory provides the symbolic framework, legitimizing events and symbols that underpin collective meaning making²². Importantly, it also contributes to the development of place attachment by embedding shared meanings into specific locales, thereby strengthening emotional bonds and cultural identification²⁵.

Within heritage tourism, cultural memory is central to understanding how visitors connect with narratives and develop a sense of place identity. Based on the literature, this study identifies three key pathways through which cultural memory shapes place attachment: symbolic, ritual, and narrative. Each pathway highlights a distinct mechanism that links memory to emotional engagement with heritage sites, leading to the formulation of hypotheses.

The symbolic pathway emphasizes material and visual triggers that evoke cultural associations. Symbols such as monuments, architectural motifs, and flags serve as markers of collective memory, activating both cognitive and emotional recall²⁶. These symbolic cues convey belonging and continuity across generations¹⁸ and play a central role in fostering place attachment²⁵. In cultural heritage tourism, such representations provide visitors with emotionally resonant entry points into the cultural significance of sites.

These insights led us to the following hypothesis:

H1: In cultural heritage tourism, the symbolic pathway of cultural memory has a significant positive influence on visitors’ place attachment.

The ritual pathway highlights embodied practices and commemorative events that enact cultural memory. Connerton described ritual as a socially organized process that reinforces collective memory through temporal, spatial, and behavioral structures²⁷. In cultural heritage tourism, rituals allow memory to be relived, transforming visitors from passive observers into active participants. These practices not only transmit cultural knowledge but also generate affective bonds with place²⁸. Engagement in rituals deepens visitors’ sense of belonging and strengthens emotional ties to cultural sites.

Therefore, we hypothesize that:

H2: In cultural heritage tourism, the ritual pathway of cultural memory has a significant positive influence on visitors’ place attachment.

The narrative pathway focuses on storytelling, myths, and traditions as mechanisms of cultural memory. By condensing complex histories into accessible narratives, cultural memory transforms heritage into resonant and meaningful experiences²⁹. Chronis emphasized that narratives bridge historical events and contemporary visitor experiences, turning sites into emotionally charged spaces of lived meaning¹². When tourists co-create narratives, they integrate personal perspectives with historical content, reinforcing symbolic relevance and identity formation.

Accordingly, the following hypothesis emerges:

H3: In cultural heritage tourism, the narrative pathway of cultural memory has a significant positive influence on visitors’ place attachment.

The impact of place attachment on tourists’ planned behavior

Place attachment has been conceptualized as a strong affective and functional bond between individuals and specific locations. Williams and Vaske identified two fundamental dimensions: place identity, which reflects the integration of a location into one’s self-concept, and place dependence, which captures the extent to which a site supports personal goals and activities¹⁹. Building on this foundation, Scannell and Gifford proposed a tripartite framework encompassing the person, process, and place dimensions of attachment²⁵. Lewicka further expanded the scope of place attachment by linking it to memory, social ties, and cultural symbols, thereby demonstrating its interdisciplinary relevance³⁰.

Research on place attachment has progressed from simple unidimensional scales to multidimensional models that capture contextual complexity. Hidalgo and Hernández showed that attachment operates across spatial hierarchies such as home, community, and city, suggesting its layered and scalable nature³¹. Subsequent research further validated these spatial distinctions and demonstrated how place attachment interrelates with constructs such as sense of place, cultural memory, and social cohesion^25,30. In tourism contexts, place attachment has been associated with perceptions of authenticity, satisfaction, loyalty, and pro-environmental behaviors²⁰.

Drawing on the Theory of Planned Behavior (TPB), scholars have proposed that place attachment influences behavioral intentions through three mechanisms. First, it can enhance attitudes by adding emotional value and meaning to tourist behavior. Second, it may reinforce subjective norms by aligning individual choices with community expectations and significant others. Third, it can improve perceived behavioral control by fostering familiarity and self-efficacy in heritage settings^32–34. These mechanisms suggest that attachment operates as an affective catalyst that reinforces the cognitive and social antecedents of intention. Halpenny demonstrated that place attachment predicts park visitors’ pro-environmental behavioural intentions and identity-related bonds, thereby enhancing normative influence³⁵. Ramkissoon et al. confirmed its dual impact on attitudes and norms, leading to sustainable behavior in park contexts³⁶. Research on perceived control further indicates that emotionally bonded tourists express higher confidence in acting responsibly and making informed decisions³⁷. Vada et al. found that immersive engagement increases both feelings of control and readiness to participate³⁸.

Recent scholarship has further advanced the understanding of place attachment by emphasizing its integrative and dynamic role in linking authenticity, satisfaction, and behavioral intentions in tourism contexts. Ramkissoon developed a conceptual framework for cultural tourism in African island economies, illustrating that the perceived authenticity of cultural attractions enhances place satisfaction, which subsequently strengthens attachment and fosters culturally oriented behavioral intentions³⁹. Building on this foundation, Ramkissoon proposed an integrative model of cultural heritage tourism that connects visitors’ perceived positive impacts, cultural heritage place attachment, and memorable heritage experiences with their revisit intentions and recommendations to heritage attractions⁴⁰. The study conceptualizes place attachment as a multidimensional construct—comprising cultural place dependence, identity, affect, and social bonding—that mediates the relationship between heritage experiences and sustainable consumption behaviors, particularly within the post-COVID-19 context. In these models, place attachment is typically specified as a direct predictor or mediator on the pathway to behavioral intention, particularly in relation to pro-environmental or responsible behaviors.

Extending this line of research, recent studies have begun to emphasize the downstream influence of place attachment on cognitive, social, and behavioral antecedents of intention. While classical models mainly describe PA in terms of identity, dependence, affect, and social bonding, work in heritage and nature-based tourism shows that emotional bonds with a place can also shape visitors’ evaluative tendencies, perceived social expectations, and sense of capability in acting within that setting^35,39,40. In this expanded view, PA functions not only as an affective connection, but also as a psychological frame through which individuals interpret and respond to heritage contexts.

Drawing from this extended conceptualization, the present study operationalizes PA as a multidimensional construct that may inform tourists’ attitudinal, normative, and control-related predispositions within the TPB framework. Conceptually, the PA items used here emphasize affective bonding with Sanhe as a place (e.g., feeling attached to the town and reluctant to lose it), whereas the TPB components are measured with items that focus on evaluative judgments about revisiting (attitude), perceived expectations of important others (subjective norm), and perceived ease and control over planning future visits (perceived behavioral control). Empirically, factor-analytic evidence supports treating PA, AB, SN, and PBC as distinct but correlated constructs, even though some content proximity between attachment-infused identity and social-normative items cannot be entirely ruled out. Accordingly, we specify PA as an upstream affective mechanism that can, in this particular heritage context, influence all three TPB antecedents. We do not assume that this pattern holds universally across all settings; rather, it is treated as a context-sensitive hypothesis to be evaluated empirically in this study.

Building upon these theoretical and empirical insights, the following hypotheses are proposed to examine the attitudinal, normative, and control-related effects of place attachment within the TPB framework:

H4: In cultural heritage tourism, place attachment positively influences tourists’ attitudes.

H5: In cultural heritage tourism, place attachment positively influences tourists’ subjective norm.

H6: In cultural heritage tourism, place attachment positively influences tourists’ perceived behavioral control.

The Influence of Tourists’ Behavioral Intentions on Cultural Heritage Tourism Experience

The Theory of Planned Behavior (TPB), introduced by Ajzen, extends the Theory of Reasoned Action to explain intentions and actual behaviors^33,41. It identifies three predictors of behavioral intention: attitude toward the behavior, subjective norm, and perceived behavioral control. The addition of perceived behavioral control addresses behaviors that may not be fully voluntary.

Over the past decades, TPB has been widely applied in domains such as environmental studies, consumer behavior, and tourism, demonstrating strong explanatory power⁴². In tourism research, TPB has been used to examine decision making, revisit intentions, and responsible behaviors^43,44. Extensions of the model incorporate affective factors⁴⁵, values⁴⁶, and constructs such as place attachment³⁶, broadening its relevance for complex tourist behaviors.

Within cultural heritage tourism, each component of TPB contributes to shaping visitor experiences. Attitude toward the behavior influences willingness to participate in conservation practices, cultural activities, and storytelling⁴⁶. Subjective norm strengthens belonging and identification in cultural contexts⁴³. Perceived behavioral control determines the feasibility of engaging in heritage activities and shapes the quality of participation⁴⁴. Integrating TPB with constructs such as cultural memory and place attachment enhances understanding of emotional engagement, cultural identity, and sustainable behavior⁴⁷.

Most existing TPB extensions in tourism have incorporated place attachment as a direct predictor or mediator of behavioral intention, particularly for pro-environmental or revisit behaviors, while treating memory-related constructs as background variables or experiential outcomes. By contrast, the present study links multidimensional cultural memory pathways to place attachment and, from there, to the three TPB components, behavioral intention, and overall heritage experience. This cross-theoretical bridge allows us to examine how symbolic, ritual, and narrative cues feed into the cognitive and social determinants of intention, rather than simply adding place attachment as another predictor of behavioral intention.

Based on these insights, the following hypotheses are proposed:

H7: In cultural heritage tourism, attitude toward the behavior positively influences behavioral intention.

H8: In cultural heritage tourism, subjective norm positively influences behavioral intention.

H9: In cultural heritage tourism, perceived behavioral control positively influences behavioral intention.

H10: In cultural heritage tourism, behavioral intention positively influences the overall tourist experience.

The conceptual framework and hypothesized model

Building upon the theoretical foundations outlined above, this study integrates Cultural Memory Theory, Place Attachment Theory, and the Theory of Planned Behavior (TPB) into a unified conceptual framework (Fig. 1). This framework conceptualizes cultural heritage tourism experiences as a cognitive–affective–behavioral process in which cultural memory activates symbolic, ritual, and narrative pathways, fostering emotional attachment to place and subsequently shaping visitors’ behavioral intentions and experiential outcomes.

Consistent with the research questions outlined in Sect. 1, the structural relationships in Fig. 1 are organized to reflect the theoretical mechanisms motivating the proposed model.

RQ1, which examines how different pathways of cultural memory shape heritage tourism experiences, is operationalized through H1–H3 and the downstream effects captured in H7–H10. These hypotheses collectively assess how symbolic, ritual, and narrative cues influence place attachment and, through TPB constructs, visitors’ experiential outcomes.

RQ2, which focuses on how place attachment informs attitudes, subjective norms, perceived behavioral control, and behavioral intentions, is represented in H4–H9. These paths articulate the affective-to-cognitive transitions through which attachment shapes planned behavior.

RQ3, which concerns how cultural memory pathways and place attachment can inform visual and narrative strategies for heritage tourism, is addressed at a conceptual level by illustrating a theory-informed VR narrative prototype developed for Sanhe Ancient Town (Sect. 4.3). The prototype is aligned with the conceptual pathway structure proposed in this study but is presented solely as a conceptual design artifact, rather than as an additional empirical experiment.

A detailed mapping between each research question and hypothesis is provided in Supplementary Table S1.

Cultural memory pathways

The left side of the model illustrates three interconnected dimensions of cultural memory—the symbolic pathway (SP), the ritual pathway (RP), and the narrative pathway (NP)—each representing a distinct mechanism through which cultural meanings are transmitted and experienced.

Consistent with Assmann and Erll, these pathways are modeled as parallel mechanisms, meaning that each can independently influence visitors’ interpretation of cultural heritage^18,22. No hierarchical ordering is imposed, as symbolic cues, ritual practices, and narrative structures all function as primary carriers of cultural memory in their own right.

Although treated as parallel antecedents in this study’s model, they are not mutually exclusive. Cultural memory scholarship acknowledges conceptual interdependence among pathways: symbolic representations often inform narratives, and ritual practices frequently reinforce symbolic meaning^23,27. However, the literature does not prescribe a dominant pathway or fixed causal sequence. This makes a parallel structure theoretically appropriate and statistically suitable for structural equation modeling.

Accordingly, the present model assumes no mediating or interactive effects among SP, RP, and NP, and each pathway is tested as an independent antecedent of place attachment.

The symbolic pathway emphasizes material and visual representations—such as architecture, artifacts, and design motifs—that evoke shared cultural associations and identity. The ritual pathway refers to participatory practices and performative traditions that sustain social cohesion and allow visitors to experience collective memory. The narrative pathway includes storytelling and interpretive media that transform historical significance into emotionally resonant and personally meaningful encounters.

Together, these pathways constitute the antecedents of place attachment by shaping visitors’ emotional and cognitive engagement with heritage sites (H1–H3).

Place attachment as an integrative mechanism

At the core of the framework, place attachment (PA) functions as an affective–cognitive integrator that translates the influence of cultural memory into planned behavioral outcomes. PA is conceptualized following the widely adopted person–process–place model²⁵, with extensive empirical evidence linking attachment to satisfaction, loyalty, and pro-environmental intentions^20,36.

In this model, PA is hypothesised to positively influence the three TPB antecedents—attitude (AB), subjective norm (SN), and perceived behavioural control (PBC) (H4–H6), reflecting its role in reinforcing visitors’ emotional valuation, normative alignment, and perceived agency in a heritage tourism context. In this sense, place attachment connects the symbolic and affective meanings of heritage with the rational determinants of behavioural intention. Following work that links attachment both to personal evaluations and to social identity and efficacy processes in tourism and environmental behaviour^35,39,40, we therefore model PA as a potential upstream input to all three TPB components in the present study. This should be understood as a theoretically motivated and empirically testable specification for the current heritage context, rather than as a universal assumption about place attachment in all behavioural domains. As discussed later in the Results and Discussion, the magnitudes of the estimated PA→AB/SN/PBC paths should nonetheless be interpreted with some caution given the measurement properties of the PA scale in this dataset.

Behavioral intention and experience formation

The right side of the model applies the TPB structure to explain how attitude (AB), subjective norm (SN), and perceived behavioral control (PBC) jointly predict behavioral intention (BI) (H7–H9). Behavioral intention is then posited to directly influence cultural heritage tourism experience (CHTTE) (H10), reflecting the extent to which visitors internalize, enact, and sustain engagement with meaningful cultural narratives. The model omits the actual behavior stage of TPB for several theoretically grounded and methodological reasons:

Intention as the most proximal determinant of behavior. Ajzen underscores that intention represents the strongest immediate predictor of actual behavior, particularly when the behavior occurs outside the researcher’s temporal control³³. Temporal delay in heritage tourism behaviors. Visitors may revisit, recommend, or participate in cultural activities months after the on-site experience. Measuring actual behavior is therefore not feasible within a cross-sectional design—a limitation widely acknowledged in TPB-based tourism research^43,44. Intention as a valid proxy in behavioral prediction. Meta-analytic evidence confirms that intention is a robust and empirically acceptable proxy when the final behavior is not immediately observable⁴².

Accordingly, in line with established TPB applications in heritage tourism, behavioral intention serves as the endpoint of the predictive chain, and its influence on perceived tourism experience (H10) is examined to link behavioral motivation with holistic experiential outcomes.

Integration and research contributions

This integrative model advances prior research in two main ways.

First, it extends traditional TPB applications in tourism by incorporating multidimensional cultural memory as cognitive–affective antecedents of place attachment, thereby linking sociocultural meaning-making with behavioural prediction. Whereas earlier PA–TPB models have mainly treated place attachment as an additional predictor or mediator on the path to behavioural intention^36,39,40,47, the present framework positions symbolic, ritual, and narrative cultural memory pathways as upstream representational mechanisms and specifies place attachment as an affective bridge from these pathways to the three TPB components, behavioural intention, and overall heritage experience.

Second, it introduces a cross-framework pathway that connects the symbolic, ritual, and narrative dimensions of cultural memory to behavioural intention through the mediating mechanism of place attachment. This approach provides a more comprehensive understanding of how cultural representation, emotional bonding, and planned behaviour jointly shape cultural-heritage tourism experiences. At the same time, because the model is derived from a single-site, cross-sectional study at Sanhe Ancient Town, it is best regarded as a theoretically portable, analytically transferable template that should be adapted and re-estimated in other heritage contexts, rather than as a statistically generalisable model with fixed parameter values.

Methods

Research questionnaire

Research background

Sanhe Ancient Town is located in Feixi County, Hefei City, Anhui Province, and is officially designated as a National 5 A Tourist Attraction by the Ministry of Culture and Tourism of China. According to the Anhui Provincial Department of Culture and Tourism and the Hefei Municipal Culture and Tourism Bureau, Sanhe receives more than two million visitors annually and is recognized for its well-preserved historical architecture, traditional waterways, and cultural heritage assets. Despite its national status, existing interpretation programs remain limited in fully supporting immersive visitor engagement, highlighting the need for empirical research on heritage experience design.

Questionnaire and measurement

The survey consisted of demographic questions and 34 measurement items across the constructs in the conceptual model, all rated on a five-point Likert scale (1 = strongly disagree; 5 = strongly agree). Measurement items were adapted from established scales in cultural heritage, place attachment, and TPB literature, with contextual refinements for applicability to Sanhe Ancient Town (Table 1). Full item sets, source references, and modification rationales are provided in the Supplementary Table S2.

Table 1.

Question items and references.

Constructs	Coding	Items	References
Symbolic Pathway(SP)	SP1	I perceive the traditional architecture, landmarks, and decorative patterns in Sanhe Ancient Town as representations of its unique cultural symbolism.	25,50,51
	SP2	I feel a strong sense of cultural identification with traditional Chinese heritage when encountering symbolic elements in Sanhe Ancient Town, such as ancient bridges, archways, and couplets.
	SP3	Culturally symbolic elements in Sanhe Ancient Town naturally evoke feelings of familiarity and a sense of belonging in me.
Ritual Pathway(RP)	RP1	Experiencing traditional festivals or folk customs in Sanhe Ancient Town has strengthened my emotional attachment to the site.	27,52,53
	RP2	Participating in traditional crafts or cultural experience programs in Sanhe Ancient Town has enhanced my identification with its cultural atmosphere.
	RP3	I believe that the cultural rituals in Sanhe Ancient Town, such as temple fairs and intangible heritage displays, have helped me form a special emotional bond with this place.
Narrative Pathway(NP)	NP1	I can clearly perceive the local historical narratives and cultural stories in Sanhe Ancient Town.	20,54,55
	NP2	These historical stories evoke a sense of emotional identification with Sanhe Ancient Town.
	NP3	The cultural stories of Sanhe Ancient Town strengthen my emotional connection and sense of attachment to this place.
Place Attachment(PA1)	PA11	I hope that people around me would appreciate Sanhe Ancient Town as much as I do.	33,56,57
	PA12	I believe that those who share similar interests with me also consider visiting Sanhe Ancient Town a meaningful activity.
	PA13	My emotional attachment to Sanhe Ancient Town makes me care more about whether others support my decision to visit the place.
Place Attachment(PA2)	PA21	I find it highly meaningful to visit places like Sanhe Ancient Town that hold historical and cultural value.	33,56,57
	PA22	Because of my emotional connection to Sanhe Ancient Town, I am more willing to engage in behaviors that contribute to its preservation, such as avoiding littering and respecting local customs.
	PA23	I believe Sanhe Ancient Town is worth recommending to my friends and family because I personally feel a strong emotional attachment to it.
Place Attachment(PA3)	PA31	I have the ability to decide whether to revisit Sanhe Ancient Town for cultural tourism.	33,58,59
	PA32	If I want to, I can easily find ways to participate in cultural preservation or volunteer activities in Sanhe Ancient Town.
	PA33	Because of my emotional attachment to Sanhe Ancient Town, I believe I am capable of making a meaningful contribution to it, such as through promotion or supportive actions.
Subjective Norm(SN)	SN1	My family and close friends support my participation in cultural tourism at Sanhe Ancient Town.	33,43,44
	SN2	I believe that most people who are important to me expect me to revisit cultural heritage sites such as Sanhe Ancient Town.
	SN3	People around me consider participation in cultural heritage tourism activities to be a worthwhile and supported behavior.
Attitude toward the Behavior(AB)༉	AB1	I believe that visiting cultural heritage sites such as Sanhe Ancient Town is a meaningful experience.	33,43,44
	AB2	I hold a positive attitude toward participating in cultural heritage tourism activities.
	AB3	In my view, cultural heritage tourism is an activity worth investing time and effort in.
Perceived Behavioral Control(PBC)	PBC1	I believe I can easily arrange the time and itinerary to visit cultural heritage sites such as Sanhe Ancient Town.	33,43,44
	PBC2	As long as I am willing, I can easily revisit Sanhe Ancient Town or similar heritage destinations.
	PBC3	Visiting cultural heritage sites is something that I find easy to accomplish.
Behavioral Intention(BI)	BI1	I am willing to revisit Sanhe Ancient Town to further experience its cultural heritage.	60–62
	BI2	I intend to recommend Sanhe Ancient Town to others as a destination rich in cultural experiences.
	BI3	I am satisfied with my cultural heritage tourism experience in Sanhe Ancient Town and intend to spend more time engaging in similar cultural activities in the future.
Cultural heritage tourism tourist experience(CHTTE)	CHTTE1	My cultural tourism experience in Sanhe Ancient Town was highly satisfying and enjoyable.	62–64
	CHTTE2	The cultural experiences in Sanhe Ancient Town enhanced my understanding and appreciation of local heritage.
	CHTTE3	My cultural heritage experience in Sanhe Ancient Town inspired my interest in exploring similar heritage destinations in the future.
	CHTTE4	I felt deeply engaged and immersed in the cultural atmosphere during my visit to Sanhe Ancient Town.

All measurement items were originally developed in English and underwent standardized translation and back-translation procedures⁴⁸. Two bilingual researchers independently translated the items into Chinese, after which a separate bilingual expert back-translated them into English. Discrepancies between the original and back-translated versions were discussed and resolved to ensure semantic and conceptual equivalence.

Prior to formal data collection, a pre-test involving ten individuals familiar with Sanhe Ancient Town was conducted to examine item clarity, comprehensibility, and completion time through cognitive interviewing. Although this group was not intended to represent the target population, the process ensured that respondents interpreted each item as intended. Minor wording adjustments were made for improved clarity, and the final instrument demonstrated satisfactory readability.

Consistent with recent developments in heritage tourism research, the Place Attachment (PA) items were operationalized as attachment-infused facets that reflect how emotional bonding with Sanhe Ancient Town colours visitors’ evaluative, normative, and control-related tendencies. Specifically, PA1 items capture the social sharing of attachment (e.g., wanting significant others to appreciate the place), PA2 items tap value-laden attachment (e.g., seeing Sanhe as personally meaningful and worth recommending), and PA3 items reflect agency-related attachment (e.g., feeling able to revisit or contribute to the site). These items therefore refer explicitly to feeling attached to Sanhe as a place, whereas the TPB constructs are defined at a more general level: Attitude (AB) assesses overall evaluations of engaging in cultural heritage tourism, Subjective Norm (SN) assesses perceived social expectations regarding such engagement, and Perceived Behavioral Control (PBC) assesses perceived ease of carrying out the behaviour. PA is thus specified as an attachment-centred construct that provides an affective context for these evaluations, norms and control beliefs, rather than measuring them directly.

Given the self-reported nature of the questionnaire data, common method bias (CMB) was examined using Harman’s single-factor test. All measurement items were entered into an unrotated exploratory factor analysis. The first factor accounted for 40.87% of the total variance, which is below the recommended threshold of 50%⁴⁹. Furthermore, the factor structure did not show single-factor dominance, and multiple factors had eigenvalues greater than 1, suggesting that common method bias is unlikely to pose a threat to the validity of the findings.

Respondent

A systematic intercept sampling approach was used. Over 800 visitors were approached during the ten-day data collection period, of whom 506 consented to complete the questionnaire. After the data-screening procedure, 420 valid responses were retained for analysis. Although a substantial number of visitors declined to participate due to time constraints or lack of interest, the final sample size comfortably exceeds recommended thresholds for SEM and provides adequate statistical power.

Data were collected from June 22 to July 1, 2025, covering both weekdays and weekends. According to the Hefei Municipal Culture and Tourism Bureau’s event calendar, no major festivals, temple fairs, or ritual performances occurred during this period, so visitors’ experiences primarily reflected routine sightseeing and everyday cultural activities rather than large-scale ritual events. These contextual features characterize the ritual environment of Sanhe during the survey period and provide background for interpreting the findings in later sections.

Demographically, the sample consisted of 47.46% male and 52.53% female visitors. Roughly half of the respondents (52.38%) were non-local tourists, while the remainder were residents from Hefei or surrounding areas (Table 2). Independent-sample t-tests and chi-square tests comparing valid and excluded cases (n = 86) showed no significant differences across demographics, indicating that screening procedures did not introduce meaningful bias. Sensitivity analyses comparing local and non-local visitors also revealed no significant differences across PA, TPB constructs, or behavioral intentions.

Table 2.

Demographic characteristics of the respondents.

Attribute	Items	Numbers	Percentage (%)
Gender	Male	189	47.62%
	Female	231	52.38%
Age	Under 18 years old	21	5%
	18 ~ 25	107	25.48%
	26 ~ 30	70	16.67%
	31 ~ 40	47	11.19%
	41 ~ 50	48	11.43%
	51 ~ 60	45	10.71%
	Over 60 years old	82	19.52%
Profession	Students	44	10.48%
	Teachers	21	5%
	Healthcare	11	2.62%
	Marketing/Sales/Business	16	3.81%
	Purchasing	15	3.57%
	Administrators	9	2.14%
	HR	17	4.05%
	Product/Operations	21	5%
	Self-employed	29	6.9%
	Finance/Accounting/Cashier/Auditor	15	3.57%
	Business Managers	20	4.76%
	Lawyers/Legal Counsel	19	4.52%
	Designers	20	4.76%
	Service Industry	21	5%
	Technical Development/Engineers	15	3.57%
	Agricultural, Forestry, Animal Husbandry, and Fishery Workers	10	2.38%
	Workers	9	2.14%
	Full-time Housewife/Husband	21	5%
	Self-employed	14	3.33%
	Retired	54	12.86%
	Other	19	4.52%
Region	Local resident of Hefei or the surrounding areas	200	47.62%
	Tourists from other places	220	52.38%

Data collection

As shown in Fig. 2, surveys were administered using mobile devices at the main exit, visitor center, and several popular heritage landmarks. This multi-location strategy ensured variation in visitor experience stages (e.g., arrival, mid-visit, departure). All participants were informed of the study purpose and assured of anonymity and the voluntary nature of participation. All procedures adhered to ethical standards in tourism research.

Fig. 2.

Survey interview location and related information in Sanhe Ancient Town.

Data analysis

Structural equation modeling

SEM involves two main steps: first, the assessment of the measurement model through the reliability and validity of the constructs; second, the assessment of the structural model (Fig. 1) by testing the hypotheses.

The descriptive statistics for all measurement items are shown in Table 3. Responses covered the full scale range (1–5), with item means between 3.643 and 4.152 and standard deviations from 1.050 to 1.246, indicating sufficient variability. Skewness (− 1.151 to − 0.555) and kurtosis (− 0.777 to 0.514) fell within the acceptable thresholds of |3| and |10| (Moorthy et al., 2019), supporting approximate univariate normality.

Table 3.

Results of Descriptive statistics analysis and normality test.

Construct	Code	Min	Max	Mean	SD	Skewness	Kurtosis
SP	SP1	1	5	3.964	1.091	−0.938	0.144
SP	SP2	1	5	4.024	1.096	−1.040	0.383
SP	SP3	1	5	4.005	1.081	−0.977	0.264
RP	RP1	1	5	4.143	1.054	−1.136	0.514
RP	RP2	1	5	4.152	1.108	−1.151	0.350
RP	RP3	1	5	3.998	1.186	−1.014	−0.015
NP	NP1	1	5	3.710	1.242	−0.606	−0.698
NP	NP2	1	5	3.762	1.179	−0.627	−0.540
NP	NP3	1	5	3.786	1.157	−0.746	−0.280
PA	PA12	1	5	3.790	1.184	−0.689	−0.580
PA	PA13	1	5	3.714	1.164	−0.559	−0.620
PA	PA21	1	5	4.017	1.069	−0.881	−0.078
PA	PA22	1	5	3.945	1.050	−0.724	−0.330
PA	PA23	1	5	3.790	1.164	−0.598	−0.617
PA	PA31	1	5	3.800	1.149	−0.722	−0.289
PA	PA32	1	5	3.643	1.246	−0.555	−0.752
SN	SN1	1	5	3.748	1.182	−0.668	−0.438
SN	SN2	1	5	3.657	1.215	−0.570	−0.595
SN	SN3	1	5	3.793	1.178	−0.746	−0.282
AB	AB1	1	5	4.012	1.080	−0.994	0.260
AB	AB2	1	5	4.071	1.112	−1.126	0.428
AB	AB3	1	5	4.048	1.098	−1.008	0.146
PBC	PBC1	1	5	3.836	1.188	−0.682	−0.558
PBC	PBC2	1	5	3.883	1.186	−0.756	−0.508
PBC	PBC3	1	5	3.757	1.190	−0.598	−0.646
BI	BI1	1	5	3.957	1.139	−0.949	0.029
BI	BI2	1	5	4.010	1.088	−1.036	0.464
BI	BI3	1	5	3.993	1.128	−0.989	0.120
CHTTE	CHTTE1	1	5	3.912	1.083	−0.786	−0.066
CHTTE	CHTTE2	1	5	3.871	1.115	−0.709	−0.333
CHTTE	CHTTE3	1	5	3.876	1.109	−0.565	−0.771

Multivariate normality was further examined using Mardia’s test. Both multivariate skewness (β = 15.67, p < 0.001) and kurtosis (β = 89.42, p < 0.001) were significant, indicating deviation from multivariate normality. As such violations are common in Likert-type survey data, maximum likelihood estimation was retained, supplemented with bias-corrected bootstrapping to ensure robust standard errors and confidence intervals for structural paths.

Construct reliability and convergent validity were evaluated using standardized factor loadings, critical ratios (C.R.), Cronbach’s α, composite reliability (CR), and average variance extracted (AVE) (Table 4). Most factor loadings exceeded the 0.70 criterion, and the remaining moderate loadings (e.g., PA22 = 0.522; CHTTE3 = 0.622) were within the acceptable range for behavioral research. All freely estimated C.R. values were significant at the 0.001 level, indicating that each item contributes meaningfully to its latent construct.

Table 4.

Results of construct validity and reliability analysis.

Construct	Code	Mean	SD	Factor Loading	C.R.	P	α	AVE	CR
SP	SP1	3.964	1.091	0.683	–	–	0.742	0.498	0.748
	SP2	4.024	1.096	0.753	14.436	***
	SP3	4.005	1.081	0.678	11.399	***
RP	RP1	4.143	1.054	0.697	–	–	0.773	0.532	0.773
	RP2	4.152	1.108	0.731	14.71	***
	RP3	3.998	1.186	0.759	12.955	***
NP	NP1	3.710	1.242	0.762	–	–	0.830	0.622	0.832
	NP2	3.762	1.179	0.804	19.821	***
	NP3	3.786	1.157	0.800	16.673	***
PA	PA12	3.790	1.184	0.681	–	–	0.859	0.451	0.853
	PA13	3.714	1.164	0.718	17.305	***
	PA21	4.017	1.069	0.695	–	–
	PA22	3.945	1.050	0.522	11.366	***
	PA23	3.790	1.164	0.685	16.171	***
	PA31	3.800	1.149	0.704	–	–
	PA32	3.643	1.246	0.674	15.791	***
SN	SN1	3.748	1.182	0.735	–	–	0.781	0.542	0.780
	SN2	3.657	1.215	0.734	14.430	***
	SN3	3.793	1.178	0.739	14.525	***
AB	AB1	4.012	1.080	0.700	–	–	0.747	0.495	0.746
	AB2	4.071	1.112	0.708	12.802	***
	AB3	4.048	1.098	0.702	12.715	***
PBC	PBC1	3.836	1.188	0.757	–	–	0.806	0.582	0.807
	PBC2	3.883	1.186	0.759	15.1	***
	PBC3	3.757	1.190	0.773	15.384	***
BI	BI1	3.957	1.139	0.740	–	–	0.758	0.519	0.764
	BI2	4.010	1.088	0.742	14.674	***
	BI3	3.993	1.128	0.678	13.372	***
CHTTE	CHTTE1	3.912	1.083	0.710	–	–	0.697	0.434	0.696
	CHTTE2	3.871	1.115	0.641	10.431	***
	CHTTE3	3.876	1.109	0.622	10.202	***

Internal consistency reliability was satisfactory, with Cronbach’s α ranging from 0.697 to 0.886 and CR from 0.696 to 0.907, both above recommended thresholds. AVE values for NP (0.622), PBC (0.582), BI (0.519), and SN (0.542) exceeded the 0.50 benchmark. Although SP (0.498) and PA (0.451) were slightly below 0.50, their CR values (> 0.74 and > 0.85, respectively) justify their retention, consistent with the criterion that high CR can compensate for marginally lower AVE⁶⁵. For CHTTE, the AVE is more clearly below the conventional threshold, which reflects the broad, outcome-oriented nature of this construct.

Overall, the results indicate acceptable reliability and convergent validity for the present structural modelling purposes; however, effect sizes and path strengths involving PA and especially CHTTE should be interpreted with some caution, and future work is encouraged to refine these scales.

Discriminant validity was further evaluated using the Heterotrait–Monotrait Ratio (HTMT), which has been shown to outperform the traditional Fornell–Larcker criterion in detecting construct overlap⁶⁶. After removing items with poor loadings (PA33 and CHTTE4) and re-estimating the measurement model, all HTMT values fell below the conservative threshold of 0.95, ranging from 0.474 to 0.953 (Table 5). The 95% bias-corrected bootstrap confidence intervals for all construct pairs did not include 1.00, indicating that no pair of constructs exhibited redundancy.

Table 5.

Results of the HTMT.

SP	SP	RP	NP	PA	SN	AB	PBC	BI	CHTTE
	1.000	0.747	0.474	0.670	0.578	0.723	0.588	0.624	0.667
RP	0.747	1.000	0.480	0.615	0.525	0.628	0.554	0.645	0.662
NP	0.474	0.480	1.000	0.953	0.850	0.726	0.810	0.787	0.606
PA	0.670	0.615	0.953	1.000	0.951	0.888	0.878	0.880	0.664
SN	0.578	0.525	0.850	0.951	1.000	0.800	0.872	0.890	0.573
AB	0.723	0.628	0.726	0.888	0.800	1.000	0.759	0.854	0.666
PBC	0.588	0.554	0.810	0.878	0.872	0.759	1.000	0.875	0.592
BI	0.624	0.645	0.787	0.880	0.890	0.854	0.875	1.000	0.781
CHTTE	0.667	0.662	0.606	0.664	0.573	0.666	0.592	0.781	1.000

Although several correlations (e.g., PA–SN and NP–PA) were relatively high, this pattern is not unexpected in cultural-memory- and TPB-based models, where place attachment can function as an affective precursor to TPB determinants and narrative processing can serve as a cognitive antecedent of attachment. In the present study, PA, SN and NP are retained as separate constructs because they remain distinguishable according to conventional HTMT and Fornell–Larcker criteria and are theoretically defined with different focal content. At the same time, their relatively high correlations indicate that these dimensions are conceptually proximate, which we treat as a measurement caveat and an important focus for more fine-grained scale development and validation in future work.

The Fornell–Larcker criterion was also used to assess discriminant validity (Table 6). For most constructs, the square root of AVE exceeded the corresponding inter-construct correlations, indicating acceptable discriminant validity. One exception occurred between PA and SN, where the correlation slightly exceeded √AVE, suggesting only partial discriminant validity for this pair. Given the conceptual relatedness of place attachment and subjective norm in social–psychological models, some degree of overlap is not surprising. Taken together with the HTMT results and the absence of perfect redundancy, we consider the level of discriminant validity adequate for the present model, while noting that PA and SN in particular should be interpreted with caution and would benefit from further refinement in future scale development work.

Table 6.

Results of the correlation analysis and the discriminant validity comparison among dimensions.

Construct	SP	RP	NP	PA	SN	AB	PBC	BI	CHTTE
SP	0.705	0.737	0.457	0.633	0.564	0.716	0.576	0.606	0.665
RP	0.737	0.731	0.485	0.578	0.525	0.626	0.553	0.642	0.675
NP	0.457	0.485	0.789	0.866	0.846	0.725	0.808	0.789	0.612
PA	0.633	0.578	0.866	0.699	0.900	0.864	0.812	0.845	0.675
SN	0.564	0.525	0.846	0.900	0.737	0.799	0.872	0.894	0.578
AB	0.716	0.626	0.725	0.864	0.799	0.704	0.759	0.847	0.675
PBC	0.576	0.553	0.808	0.812	0.872	0.759	0.763	0.874	0.588
BI	0.606	0.642	0.789	0.845	0.894	0.847	0.874	0.763	0.771
CHTTE	0.665	0.675	0.612	0.675	0.578	0.675	0.588	0.771	0.719

Model fit indices are summarized in Table 7. The chi-square ratio (CMIN/DF = 3.161) is within the acceptable range. Incremental fit indices—GFI (0.835), NFI (0.825), CFI (0.872), IFI (0.873), and TLI (0.854)—meet or closely approach recommended thresholds, providing evidence of adequate comparative fit. RMSEA (0.072) and SRMR (0.067) fall within the desirable < 0.08 range, indicating satisfactory approximate fit. Although AGFI (0.799) and PCLOSE are marginal, the overall pattern of indices supports the adequacy of the model for structural analysis.

Table 7.

Model fit indices.

Model fit index	Value	Recommended threshold	Evaluation
CMIN/DF	3.161	1–5	Acceptable
GFI	0.835	> 0.80	Acceptable
AGFI	0.799	> 0.80	Marginal
TLI	0.854	> 0.85	Acceptable
CFI	0.872	> 0.85	Acceptable
NFI	0.825	> 0.80	Acceptable
IFI	0.873	> 0.85	Acceptable
RMSEA	0.072	< 0.08	Good
PCLOSE	0.000	> 0.05	Not supported
RMR/SRMR	0.067	< 0.08	Good
AIC	1465.688	—	—
ECVI	3.498	—	—
HOELTER (0.05)	149	> 200 ideal	Moderate

The structural model results are shown in Table 8. SP (β = 0.319, p < 0.001) and NP (β = 0.612, p < 0.001) significantly predicted PA, supporting H1 and H3, whereas RP had no significant effect, thus H2 was not supported. PA exerted strong positive influences on SN (β = 1.026), AB (β = 0.845), and PBC (β = 0.993), supporting H4–H6. All three TPB components significantly predicted BI (SN → BI: β = 0.306; AB → BI: β = 0.388; PBC → BI: β = 0.309), supporting H7–H9. Finally, BI strongly predicted CHTTE (β = 0.694), supporting H10. Given the high correlations between PA and some TPB components (especially SN and PBC), the relatively large standardized coefficients for PA → SN and PA → PBC should be understood as indicating a strong association under conceptual proximity, rather than as precise estimates of effect magnitude. The consistent significance of these pathways, together with bootstrap-based confidence intervals, provides overall support for the proposed theoretical framework, within the measurement limitations noted above.

Table 8.

Results of path coefficients hypotheses.

Hypothesis	Path	B	SE	CR	P	95% BC Bootstrap CI	Result
H1	SP → PA	0.319	0.069	4.642	***	[0.19, 0.44]	Supported
H2	RP → PA	0.045	0.062	0.724	0.469	[−0.08, 0.17]	Not supported
H3	NP → PA	0.612	0.039	15.531	***	[0.54, 0.69]	Supported
H4	PA → SN	1.026	0.063	16.222	***	[0.91, 1.14]	Supported
H5	PA → AB	0.845	0.059	14.250	***	[0.74, 0.94]	Supported
H6	PA → PBC	0.993	0.064	15.622	***	[0.87, 1.11]	Supported
H7	SN → BI	0.306	0.108	2.836	0.005	[0.09, 0.51]	Supported
H8	AB → BI	0.388	0.115	3.385	***	[0.18, 0.60]	Supported
H9	PBC → BI	0.309	0.084	3.669	***	[0.14, 0.47]	Supported
H10	BI → CHTTE	0.694	0.064	10.824	***	[0.57, 0.82]	Supported

Given the marginal AVE values observed for SP, PA, and CHTTE, we conducted a series of additional analyses to further examine the robustness of the measurement model. First, we ran a split-sample CFA by randomly dividing the full sample into two subsamples (n₁ ≈ n₂ ≈ 210). As reported in Supplementary Table S3, both subsamples yielded highly comparable factor loadings (typically between 0.65 and 0.80), AVE and CR values, and global fit indices (χ²/df ≈ 2.4–2.5, CFI ≈ 0.81–0.83, TLI ≈ 0.80–0.81, RMSEA ≈ 0.08). The same items (e.g., PA22, CHTTE3) consistently showed relatively weaker loadings across both subsamples, suggesting that the slightly sub-optimal AVE for SP, PA, and CHTTE reflects the broad conceptual scope of these constructs rather than instability in a particular split of the data.

Second, we carried out item-deletion sensitivity analyses focusing on the weakest indicators (Supplementary Table S4). When the lowest-loading PA item (PA22) was removed, the AVE of PA increased from 0.45 to approximately 0.48, whereas its composite reliability remained high (CR ≈ 0.85). Crucially, the key structural paths (SP → PA, NP → PA, PA → SN/AB/PBC, BI → CHTTE) preserved their directions and statistical significance, and the explained variance of CHTTE increased slightly (from about 0.54 to 0.58), with global fit indices at a similar level. This indicates that our main findings are not driven by a single weak PA indicator.

Third, we examined the effect of dropping the weakest CHTTE indicator (CHTTE3). In this two-item specification, the AVE of CHTTE remained around 0.42 while its composite reliability decreased to around 0.60, indicating that a reduced CHTTE scale did not meaningfully improve convergent validity and instead weakened internal consistency. Although the BI → CHTTE path became somewhat stronger and R² for CHTTE increased (to about 0.66), the overall fit indices remained comparable to the original model. Taken together, these analyses show that the core structural relationships are stable across different specifications and that the risk of spuriously significant paths due to a few marginal items is limited. At the same time, we explicitly acknowledge that SP, PA, and especially CHTTE are modelled here as relatively broad, outcome-oriented constructs; future work should therefore further refine and validate more site-specific and potentially multidimensional measures of cultural heritage tourism experience.

In summary, the structural model showed an overall fit pattern that is broadly consistent with the proposed Cultural–Affective–Behavioral Intention chain. Among the three cultural-memory pathways, symbolic (SP) and narrative (NP) cues are positively associated with place attachment, whereas the ritual pathway (RP) is not supported as a linear predictor in this dataset. Place attachment is positively associated with the three TPB antecedents (attitude, subjective norm, and perceived behavioral control), which jointly predict behavioral intention, and intention is positively associated with the overall heritage-tourism experience index; however, coefficients involving PA and CHTTE should be interpreted with caution given marginal convergent validity and conceptual proximity with TPB-related items. The following Random Forest analyses are reported as an exploratory robustness complement to assess relative predictive signal under potential collinearity and nonlinearity, rather than as a re-test of the SEM paths.

Random forest

To complement the SEM results and explore potential nonlinear effects and higher-order interactions, we fitted random-forest regressors with BI and CHTTE as dependent variables and the seven TPB and cultural-memory constructs (SP, RP, NP, PA, AB, SN, PBC) as predictors. Models were implemented in Python (scikit-learn 1.x) with 100 trees, a maximum depth of 10, minimum samples per split of 5 and per leaf of 2, bootstrap sampling, and out-of-bag (OOB) prediction enabled (Table 9). For BI, the random-forest model achieved an OOB R² of 0.59 and a 10-fold cross-validated R² of 0.59 (RMSE = 0.58; MAE = 0.41), indicating moderate predictive power on unseen data. The corresponding training R² was higher (0.88), which suggests some overfitting, a common feature of flexible tree ensembles, and reinforces the need to focus on OOB and cross-validated metrics when judging performance. For CHTTE, the model yielded lower predictive accuracy (training R² = 0.78, OOB R² = 0.36; 10-fold CV R² = 0.36, RMSE = 0.67; MAE = 0.51), consistent with the greater conceptual breadth and measurement complexity of overall tourist experience. Given the moderate sample size and correlated predictors, we therefore interpret the random-forest results as exploratory robustness checks that highlight relative patterns of predictor importance, rather than as high-precision predictive models or formal tests of the hypothesised paths.

Table 9.

Random forest hyperparameters and performance for BI and CHTTE.

Target	n_estimators	max_depth	min_samples _split	min_samples _leaf	max_features	bootstrap	OOB R²	CV R² (mean ± SD)	CV RMSE (mean ± SD)	CV MAE (mean ± SD)
BI	100	10	5	2	auto	True	0.592	0.593 ± 0.065	0.575 ± 0.058	0.410 ± 0.030
CHTTE	100	10	5	2	auto	True	0.363	0.364 ± 0.195	0.673 ± 0.097	0.505 ± 0.075

To quantify the relative importance of each predictor, we computed permutation importance scores by repeatedly shuffling each feature (10 independent repetitions × 10 permutations per feature, yielding 100 estimates per variable). For each predictor we then calculated the mean importance and its variability, and used one-sample t-tests against zero with Benjamini–Hochberg false discovery rate (FDR) adjustment to check whether the observed importances were clearly distinguishable from noise. As shown in Table 10, all seven predictors had importance values that were reliably greater than zero (all FDR-corrected p < 0.001). SN and PBC emerged as the strongest predictors of BI, followed by AB and PA. The three cultural-memory pathways (NP, RP, SP) also contributed positively, although with smaller effect sizes, suggesting that symbolic, reenactment, and narrative cues influence intention primarily through the TPB components rather than dominating them. Because several predictors are conceptually adjacent and moderately to strongly correlated, random-forest importance values are best interpreted at the level of clusters of constructs (e.g., the TPB block vs. the cultural-memory block) rather than as precise rankings of single variables. We therefore refrain from making strong claims about SN being the unique dominant determinant of BI, and instead emphasize that the TPB components jointly provide the strongest predictive signal.

Table 10.

Permutation-based feature importance for BI.

Rank	Feature	Mean importance	SD	t-value	p-value	FDR q-value	Sig.
1	SN	0.247	0.0208	37.55	< 0.001	< 0.001	***
2	PBC	0.218	0.0213	32.41	< 0.001	< 0.001	***
3	AB	0.130	0.0078	52.55	< 0.001	< 0.001	***
4	PA	0.130	0.0113	36.23	< 0.001	< 0.001	***
5	NP	0.112	0.0111	24.86	< 0.001	< 0.001	***
6	RP	0.108	0.0054	63.66	< 0.001	< 0.001	***
7	SP	0.055	0.0023	73.69	< 0.001	< 0.001	***

For CHTTE, permutation results again showed that all predictors were statistically important (all FDR-corrected p < 0.001; Table 11). Behavioral intention (BI) had the highest mean importance (0.31), followed by place attachment (PA; 0.25) and the narrative pathway (NP; 0.22). Reenactment and symbolic pathways (RP, SP) also contributed meaningfully, whereas attitude (AB), subjective norm (SN), and perceived behavioral control (PBC) showed smaller but still non-negligible contributions. This pattern suggests that visitors’ cultural-heritage tourism experience is enhanced primarily through an affective–behavioral mechanism: cultural memory pathways strengthen place attachment and TPB constructs, which in turn support more meaningful and satisfying heritage tourism experiences, rather than relying on any single pathway alone. This is consistent with the SEM results, where BI and PA occupy central positions in the structural model. Given the collinearity among several predictors and the moderate sample size, however, the importance scores should be interpreted as highlighting relative contributions of construct families (e.g., behavioral intention and attachment vs. specific cultural-memory pathways) rather than as precise, rank-ordered effect sizes for individual variables.

Table 11.

Permutation-based feature importance for CHTTE.

Rank	Feature	Mean importance	SD	t-value	p-value	FDR q-value	Sig.
1	BI	0.306	0.0183	52.99	< 0.001	< 0.001	***
2	PA	0.246	0.0188	41.45	< 0.001	< 0.001	***
3	NP	0.222	0.0196	35.67	< 0.001	< 0.001	***
4	RP	0.193	0.0058	80.51	< 0.001	< 0.001	***
5	SP	0.143	0.0070	54.67	< 0.001	< 0.001	***
6	AB	0.069	0.0026	84.25	< 0.001	< 0.001	***
7	SN	0.060	0.0065	46.48	< 0.001	< 0.001	***
8	PBC	0.050	0.0043	36.86	< 0.001	< 0.001	***

The SHAP summary plots and Boruta feature-selection results (Figs. 3, 4 and 5) showed a ranking pattern highly consistent with the permutation-based importance tests. For BI, all three methods converged in identifying the TPB components—especially subjective norm and perceived behavioral control, followed by attitude and place attachment—as the most influential predictors, while the three narrative pathways (SP, RP, NP) made smaller but non-negligible contributions. For CHTTE, BI and place attachment emerged as the dominant drivers, with the narrative pathways exerting moderate incremental effects and the remaining TPB components playing a secondary role. Taken together with the SEM results, these convergent machine-learning analyses suggest that visitors’ cultural-heritage tourism is more strongly associated with an affective–behavioral mechanism in which cultural-memory pathways—especially narrative cues—strengthen place attachment and the TPB constructs (attitude, subjective norm, and perceived behavioral control), rather than being driven by any single symbolic or ritual design element alone.

Fig. 3.

Boruta feature importance for Behavioral Intention (BI).

Source: Generated by Boruta R library.

Fig. 4.

Boruta feature importance for Cultural Heritage Tourism Experience (CHTTE).

Source: Generated by Boruta R library.

Fig. 5.

a. Left)SHAP feature importance for BI (tuned Random Forest). b. (Right༉SHAP feature importance for CHTTE (tuned Random Forest).

Source: Generated by Boruta R library.

Discussion

This study examines mechanisms through which cultural memory and place attachment relate to the formation of cultural heritage tourism experiences, and how these relationships can inform design-oriented hypotheses. Data were collected via an on-site questionnaire survey at Sanhe Ancient Town (Anhui Province, China) and analysed using structural equation modelling (SEM) and an exploratory Random Forest approach. Importantly, we treat the proposed model as an analytically transferable scaffold (i.e., a testable template that can be adapted and re-estimated across heritage contexts), while interpreting the present coefficients as context-bound and not statistically generalizable beyond this single-site sample.

Based on the findings of this study, RQ1 was supported. Cultural memory, through its symbolic, ritual, and narrative dimensions, is positively associated with visitors’ place attachment to cultural heritage tourism sites, and this attachment is positively associated with higher levels of self-reported tourism experience. The results also supported RQ2: place attachment is indirectly related to visitors’ behavioural intentions by directly influencing the three variables of the Theory of Planned Behavior (TPB)—attitude toward the behaviour, subjective norms, and perceived behavioural control—which in turn show positive associations with the broad index of cultural heritage tourism experience used in this study. These empirical patterns offer the conceptual foundation for considering RQ3 at a theoretical level. In the following sections, we therefore draw on the observed pathways between cultural memory, place attachment, and TPB constructs to propose visual language and narrative strategies for cultural heritage tourism experience design, treating RQ3 as a set of theory-informed design hypotheses rather than as an additional empirically evaluated outcome of the present survey.

This chapter discusses the research findings, including significant and non-significant paths, unexpected results, related theoretical models, and limitations.

Place attachment

Place attachment emerges in this study as a core socio-emotional mechanism that integrates visitors’ cognitive evaluations, affective connections, and behavioral dispositions during their heritage tourism experience. The findings show that place attachment is positively associated with all three TPB components, namely attitude toward the behavior (AB), subjective norm (SN), and perceived behavioral control (PBC). This pattern aligns with Vada, Prentice, and Hsiao, who argue that emotional connection with a destination enhances positive behavioral tendencies and strengthens evaluative judgments³⁸.

Recent scholarship also positions place attachment within broader social and sensory domains of tourism behavior. Ramkissoon illustrates that interpersonal trust and perceived social impacts contribute to stronger attachment, which subsequently drives pro-social and pro-environmental intentions⁴⁰. Similarly, Ramkissoon et al. show that multisensory experiences, including visual symbols, soundscapes, tactile cues, and ambient cultural elements, reinforce both place loyalty and civic participation through attachment processes⁶⁷.

Extending these insights to a cultural-heritage context, our findings indicate that symbolic, narrative, and ritual cues collectively facilitate visitors’ emotional bonding with Sanhe Ancient Town. These cues help transform interpretive encounters into meaningful experiences that sustain heritage appreciation and long-term engagement. Symbolic cues support cultural identity recognition, narrative cues encourage personal meaning-making, and ritual cues evoke embodied participation, jointly forming a holistic attachment mechanism.

In interpreting the structural relationships, we also note several relatively high correlations among theoretically adjacent constructs, particularly between place attachment (PA) and subjective norm (SN) and between the narrative pathway (NP) and PA. Such proximity is theoretically plausible in heritage tourism, where emotional bonds to a site often co-occur with perceived social approval and narrative engagement. At the same time, these correlations indicate only partial discriminant validity between some constructs. While PA, SN and NP remain distinguishable according to conventional HTMT and Fornell–Larcker criteria, their overlap suggests that the corresponding path coefficients should be interpreted with caution and that more fine-grained, less overlapping scales are needed in future research.

Taken together, these results demonstrate that place attachment not only shapes visitors’ immediate affective engagement but also fosters a sense of collective responsibility and sensory belonging. This dual role has important implications for cultural-heritage site management, as emotionally bonded visitors are more likely to support heritage conservation, adopt sustainable destination practices, and participate in repeat visitation.

Cultural memory

Symbolic dimension and the narrative dimension of cultural memory

This study examined the symbolic and narrative dimensions of cultural memory and their effects on place attachment through structural equation modeling. The findings confirm that both dimensions significantly strengthen visitors’ emotional bonds with heritage sites.

The symbolic dimension refers to visible elements such as architecture, landmarks, and motifs that act as cultural markers. These features are not only visually distinct but also carry collective memory and cultural identification^18,26. By embodying symbolic meaning, they evoke shared recognition and contribute to place attachment. This supports Scannell and Gifford, who argued that culturally meaningful symbols sustain emotional connections between individuals and places²⁵. Sensory experiences, such as walking through historic streets or observing material traces, allow visitors to engage directly with cultural stories, enhancing identity and experiential depth.

The narrative dimension emphasizes the role of myths, rituals, and historical events in shaping memory. Narratives serve as formative forces that transform the past into meaningful experiences and continually activate remembrance²⁹. Through storytelling, visitors integrate visual cues, displays, and historical fragments into personalized interpretations. This aligns with Chronis, who noted that cultural narratives connect history, symbols, and personal experience, deepening immersion and emotional involvement¹². Resonating with local narratives enables tourists to perceive themselves as part of the cultural storyline, thereby fostering exploration and reinforcing place attachment.

Ritual dimension of cultural memory

Contrary to our initial hypothesis, the ritual dimension of cultural memory did not exhibit a statistically significant linear effect on place attachment in the SEM results. This finding diverges from influential work that highlights the role of ritual practices in sustaining cultural memory and social bonding^27,28. At the same time, the Random Forest analyses suggested that ritual-related cues still contribute to variations in behavioral intention and overall experience in a nonlinear, context-dependent manner.

Several contextual features of Sanhe Ancient Town may help to explain this pattern, although they cannot be taken as definitive tests of any mechanism. First, large-scale ritual events and performances are concentrated around major holidays, and no such events occurred during our data-collection period. For most visitors in this sample, ritual encounters may therefore have been limited to small-scale or incidental observations rather than salient, shared experiences. Second, recent festival activities at Sanhe, such as dragon-boat races or river-lantern ceremonies, are broadly similar to those found at other Chinese heritage destinations and may not be perceived as highly distinctive or tightly coupled to Sanhe’s specific historical narratives. Third, current ritual programming seldom foregrounds the site’s core historical themes (e.g., its military significance in the Three Kingdoms period and the Taiping Rebellion), which may weaken the perceived coherence between ritual experiences and the destination’s identity.

These observations are regarded as working hypotheses rather than firm causal explanations. Future research could examine them using qualitative approaches—such as brief exit interviews, informal conversations, or focused field observation—to document how visitors actually notice, participate in, and interpret ritual activities at Sanhe. Such evidence would help clarify the conditions under which rituals are experienced as generic festive entertainment versus as place-specific practices that meaningfully strengthen attachment to the site.

VR narrative prototype for sanhe ancient town

To address Research Question 3 (RQ3), which asks how cultural memory and place attachment can inform visual and narrative strategies in heritage tourism design, this study specifies a conceptual VR narrative prototype for Sanhe Ancient Town. The prototype is not analyzed as an additional empirical study; instead, it is presented as a theory-informed design illustration that translates the cultural-memory pathways (symbolic, ritual, narrative) and attachment mechanisms into concrete visual, interactive, and narrative tactics. In this sense, the prototype functions as a conceptual design artifact and a set of design hypotheses that can be evaluated in future user studies, rather than as evidence derived from the present survey data.

A design mapping framework translating the conceptual model into VR strategies

To clearly demonstrate how theoretical constructs inform design decisions, we developed a conceptual design mapping through a structured mapping process. Table 12 summarizes how each cultural memory pathway is aligned with specific VR design strategies, the corresponding attachment mechanisms, and the hypothesized TPB targets (design intent).

Table 12.

Mapping Cultural Memory Pathways to VR Visual and Narrative Strategies.

Cultural Memory Pathway	VR Design Strategy	Intended PA Mechanism (design intent)	Hypothesized TPB Target (design intent)
Symbolic Pathway(SP)	Visual symbol mapping; culturally meaningful architectural cues; color pattern overlays	Intended to support cognitive familiarity and place identity	Hypothesized to support more favorable attitudes toward the behavior (AB)
Ritual Pathway(RP)	Embodied incense lighting interaction; gesture-based participation	Intended to elicit emotional resonance and place dependence	Hypothesized to support perceived behavioral control (PBC)
Narrative Pathway(NP)	A story-driven boat trip; spatial audio; environmental changes synchronized with time.	Intended to support affective bonding and meaning-making	Hypothesized to target subjective norms (SN) by foregrounding shared narratives

This framework makes explicit how the theoretical model can be translated into hypothesis-generating design guidance for immersive cultural heritage experiences. It also illustrates how cultural memory is not only interpreted cognitively but enacted through visual, behavioral, and narrative modalities within a VR environment.

Conceptualizing cultural memory pathways in the VR prototype

As shown in Fig. 6, The VR narrative prototype comprises three illustrative experiential scenes, each corresponding to one cultural memory pathway. These scenes demonstrate how symbolic, ritual, and narrative cues can be embedded in immersive environments to support place attachment.

Fig. 6.

VR Narrative Prototype Based on Cultural Memory Pathways (SP–RP–NP): Visual Symbolic Mapping, Ritual Interaction, and Story-Driven Immersion in Sanhe Ancient Town.

Symbolic pathway scene: Night-market street immersion

The symbolic pathway is conceptualized as a virtual reconstruction of Sanhe’s traditional night-market streets. In the concept, users would wander through narrow alleys illuminated by lanterns and surrounded by distinctive historical façades. Architectural colors, wooden structures, decorative motifs, and spatial proportions are displayed as visual symbol overlays, prompting recognition of cultural identity cues. Interactive text prompts provide knowledge about architectural origins and historical significance, intended to support cognitive familiarity and place identity, the key mechanisms of symbolic-driven attachment.

Ritual Pathway Scene: Incense-Lighting at Xianyue Pavilion.

The ritual pathway is conceptualized as an embodied experience set in front of the iconic Xianyue Pavilion. In the concept, users would approach an incense burner, lift virtual incense sticks using the VR controller, and perform a simplified incense-lighting gesture. This reenactment is intended to allow users to sense ritual rhythm and bodily engagement.

Narrative Pathway Scene: Historical Storytelling on a River-Cruise Boat.

The narrative pathway is presented through a river-cruise experience along the ancient moat. The wooden boat moves slowly while users listen to spatial audio narrating Sanhe’s major historical events, including the Three Kingdoms military encounters and later wartime episodes. Overlay text appears at key moments to guide attention and is designed to help users integrate environmental cues with narrative content. This scene is designed to support meaning-making, a core mechanism of narrative-driven place attachment.

How the prototype activates place attachment and TPB constructs

Within the proposed framework, the prototype is intended only as an illustrative example of how cultural memory pathways might collectively foster place attachment and, in turn, relate to the TPB antecedents. Symbolic cues are configured to enhance cognitive recognition and place identity; ritual participation is framed to encourage emotional involvement and a sense of agency; and narrative immersion is used to deepen personal meaning-making and perceived social relevance. Taken together, these design choices outline a plausible way in which a VR experience could support more favourable attitudes, subjective norms, and perceived behavioural control, and thereby potentially encourage intentions such as revisiting, recommending, or engaging in future cultural learning. In this study, these links are presented as design hypotheses rather than as empirically tested effects of the prototype.

Contribution to RQ3

Through this integrated design illustration, the prototype offers one possible way in which cultural memory pathways and place attachment might inform visual, interactive, and narrative strategies in heritage tourism design. In doing so, it addresses RQ3 at the level of design hypotheses and provides a transferable template that other heritage sites can adapt and empirically evaluate in subsequent VR experiments or user studies.

Context-specific considerations in cultural heritage tourism

The results of this study indicate that several measurement items under the Cultural Heritage Tourism Experience (CHTTE) construct exhibited relatively modest factor loadings, ranging from 0.48 to 0.69. These values are lower than the commonly cited 0.70 benchmark achieved by most of the other latent variables in the model⁶⁸. Consistent with this pattern, the convergent validity of CHTTE was relatively weak, with an AVE below the conventional 0.50 threshold and a coefficient of determination (R²) of 0.541⁶⁵. The random-forest models in Sect. 3.2.2 likewise showed that CHTTE is more difficult to predict than behavioral intention, which is in line with its broader conceptual scope and measurement noise.

One plausible explanation lies in the generic nature of the questionnaire items, which were adapted from established tourism-experience scales but not originally developed for a single heritage site. The four items were designed to tap emotional resonance, interactive participation, willingness to revisit, and social sharing of experiences. These dimensions are appropriate for capturing broad aspects of tourism experience^69,70, yet within the specific context of cultural heritage tourism they may not be sufficiently sensitive to key elements such as cultural learning, perceived authenticity of the setting, or the depth of engagement with historical narratives. In other words, the indicators used here describe visitors’ overall enjoyment and engagement in a relatively coarse way and do not differentiate among more fine-grained heritage-specific facets.

Taken together, these considerations suggest that, in the present study, CHTTE functions as a relatively broad, site-specific indicator of visitors’ overall heritage experience rather than as a finely tuned latent construct. Accordingly, the structural paths involving CHTTE are best read as describing the general direction and relative magnitude of associations under these measurement conditions, rather than as highly precise benchmarks of effect size. Future studies that develop and validate more differentiated, multidimensional scales for heritage experience—for example, separate dimensions for cultural learning, perceived authenticity, and immersive atmosphere—will be able to estimate these relationships with greater precision and test the robustness of the current patterns across different cultural heritage destinations.

Implications for related theoretical models

This study yields two main empirical insights. First, place attachment was found to be a key affective mechanism linking cultural memory to visitors’ reported experiences in a heritage setting, increasing their willingness to revisit, recommend, and further explore similar destinations. Second, structural equation modeling indicated that the symbolic and narrative dimensions of cultural memory exert strong positive effects on place attachment, while the ritual dimension did not show a significant linear path to attachment in this single-site context. Complementary Random Forest analyses suggested that ritual-related cues still carry predictive signal for behavioral intention and overall experience, albeit in a more complex, nonlinear manner and with only moderate out-of-sample accuracy.

Taken together, these results show that symbolic, narrative, and ritual dimensions of cultural memory, in combination with place attachment, can provide a useful lens for thinking about visual and narrative strategies in cultural heritage tourism. In the case of Sanhe Ancient Town, symbolic and narrative cues emerged as the most robust correlates of attachment, whereas the current ritual programming appeared less tightly connected to visitors’ attachment reports. Rather than overturning prior work that emphasizes the importance of ritual practices^27,28, this pattern nuances those expectations by indicating that the relative salience of memory pathways may be strongly context-dependent and contingent on how rituals are designed and communicated at a given site.

Relative to dominant TPB extensions and PA–TPB models in tourism^36,43,45,47, the present framework differs in three respects. First, it explicitly positions cultural-memory pathways as upstream cognitive–symbolic inputs that precede both place attachment and TPB components, whereas most prior studies begin from attitudes, satisfaction, or general destination image. Second, place attachment is modelled as an affective hub that can influence all three TPB antecedents simultaneously, rather than being treated solely as a parallel predictor of behavioral intention or as a simple mediator between satisfaction and intention. Third, the model extends the traditional TPB outcome space by linking these mechanisms not only to behavioral intention but also to a broad index of cultural-heritage tourism experience (CHTTE). These extensions do not aim to replace existing PA–TPB formulations, but instead offer a complementary, memory-sensitive specification that is particularly suited to heritage contexts where symbolic and narrative cues are central to experience formation.

Beyond these theoretical connections, the integration of place attachment and the Theory of Planned Behavior with a multidimensional account of cultural memory—comprising symbolic, ritual, and narrative pathways—helps clarify how cognitive interpretation, emotional resonance, and shared meanings jointly shape behavioral tendencies in cultural heritage settings. In this way, the framework broadens the explanatory reach of TPB- and PA-based models into domains where cultural representation and storytelling are central design variables.

Methodologically, the combined use of Structural Equation Modeling and Random Forest provides a pragmatic form of cross-validation. SEM offers an interpretable, hypothesis-driven account of directional relations among latent constructs, whereas Random Forest contributes an exploratory, pattern-detection perspective on predictive importance under potential nonlinearity and interaction. Given the moderate out-of-bag performance and the collinearity among several predictors, the machine-learning results are best viewed as robustness checks and as a way to compare the relative contribution of construct families (e.g., attachment, TPB components, and memory pathways), rather than as a basis for fine-grained variable ranking.

Collectively, these theoretical and methodological contributions illustrate how cultural memory, place attachment, and planned behavior can be jointly mobilized to inform the analysis and design of cultural heritage tourism. At the same time, given the marginal AVE values for PA and CHTTE, the corresponding path coefficients should be interpreted as approximate indicators of direction and relative strength rather than as definitive estimates; future studies that employ refined, site-specific experience scales and multi-site samples will be important for corroborating the magnitude and generalizability of these relationships.

Contributions

Taken together, the findings of this study suggest three main contributions.

First, at the theoretical level, the results provide empirical support for an integrated Cultural–Affective–Behavioral Intention framework that links symbolic, ritual, and narrative pathways of cultural memory to place attachment, the three TPB components (attitude toward the behaviour, subjective norm, perceived behavioural control), behavioural intention, and cultural-heritage tourism experience. By treating cultural-memory pathways as upstream symbolic and narrative inputs, and by extending the TPB outcome space to include both behavioural intention and a broad index of heritage experience (CHTTE), the model supplements existing PA–TPB and memory-based tourism approaches rather than replacing them. It offers one context-sensitive specification of how cultural cues and place-based emotions may jointly shape intention formation in heritage settings.

Second, the study clarifies the role of place attachment as a central socio-emotional mechanism within this extended framework. The SEM results indicate positive associations between PA and all three TPB antecedents. These patterns are theoretically consistent with recent attachment-based models of tourist behaviour, in which emotional bonding colours attitudes, perceived social expectations, and perceived agency. At the same time, as noted in the measurement and discussion sections, the convergent validity indices for PA (and CHTTE) are marginal; the corresponding path coefficients should therefore be interpreted as indicative of direction and relative strength rather than as precise effect sizes, and future work with refined, site-specific scales will be needed to corroborate their magnitude.

Third, the research translates these mechanisms into a conceptual design for VR-based heritage interpretation. The SP/RP/NP pathways are linked to illustrative visual and narrative tactics—such as symbolic architectural cues, embodied ritual interactions, and story-driven spatial navigation—and to their proximal psychological targets in the TPB framework. In this study, this pathway–mechanism–TPB matrix is presented solely as a theory-informed design proposal and a set of testable design hypotheses, not as an empirically evaluated VR intervention. It is intended to provide a structured basis that future studies can implement and experimentally assess in dedicated VR experiments, thereby offering a replicable bridge between behavioural theory and practical experience design in cultural-heritage tourism.

Limitations and future directions

Despite the contributions of this study, several limitations should be acknowledged. To begin with, the demographic composition of the sample presents certain imbalances. The age distribution is bimodal, with a higher proportion of younger visitors and older adults, while middle-aged groups are comparatively underrepresented. Although this structure may constrain the statistical generalizability of parameter estimates and effect magnitudes to the broader population of cultural heritage tourists, it largely reflects the typical visitor composition of Sanhe Ancient Town. The site is routinely frequented by younger tourists and older adults, whereas mid-aged groups account for a smaller share of its regular tourism flow. Moreover, because Hefei is not a major national tourism city, local leisure-oriented visitation has historically constituted a substantial proportion of total visitors. These contextual factors help explain the demographic pattern observed in the dataset, while still acknowledging limitations for statistical generalizability beyond this site and sampling period.

In addition, the survey did not include socioeconomic variables such as education level and household income. These characteristics can influence cultural participation, cognitive engagement with heritage information, and travel-related behavioral intentions. Their absence restricts the ability to examine how socioeconomic status may moderate the relationships proposed in the conceptual model.

Furthermore, data were collected during a non-peak tourism period, which may have increased the proportion of local visitors from Hefei and nearby regions. Although sensitivity analyses indicated no significant differences between local and non-local participants on place attachment, TPB variables, or behavioral intentions, seasonal variation remains an important consideration for future research aiming to achieve a more representative visitor structure.

A further limitation concerns the measurement of several latent constructs, especially the cultural heritage tourism experience (CHTTE). Although the scales exhibit acceptable internal consistency, the AVE values for SP, PA, and particularly CHTTE are slightly below the conventional 0.50 threshold, and CHTTE is operationalized as a relatively broad, generic, single-factor outcome index rather than a finely tuned, site-specific, multidimensional scale. As shown in our split-sample CFA and item-deletion sensitivity analyses, the core structural relationships remain stable across alternative specifications, but simple item removal does not substantively improve convergent validity and may weaken internal consistency. Future research should therefore develop and validate more site-specific and multidimensional measures of heritage tourism experience (e.g., explicitly capturing immersion in Sanhe’s historical narratives, perceived authenticity, and cultural learning), and cross-validate these refined scales across different cultural heritage destinations.

Given these limitations, future studies are encouraged to conduct multi-period or year-round data collection to improve robustness across seasons and visitor compositions. Incorporating socioeconomic indicators would also deepen understanding of visitor heterogeneity and behavioral drivers. Integrating qualitative approaches, such as in-depth interviews or narrative analysis, may further enrich the interpretation of cultural memory processes and reveal more context-specific dimensions of heritage experience. By examining diverse heritage settings and combining quantitative and qualitative methods, future research can develop more context-sensitive and theoretically grounded frameworks for cultural heritage tourism design.

Conclusion

This study investigated how cultural memory, place attachment, and Theory of Planned Behavior (TPB) constructs jointly relate to visitors’ experiences and intentions in a cultural heritage tourism context. Using survey data from visitors to Sanhe Ancient Town, we proposed and empirically examined an integrated Cultural–Affective–Behavioral Intention framework in which symbolic, ritual, and narrative pathways of cultural memory feed into place attachment, which in turn influences attitude, subjective norm, perceived behavioral control, behavioral intention, and an overall index of cultural-heritage tourism experience (CHTTE). Structural equation modeling (SEM) was used to test the hypothesised directional relationships, while Random Forest analysis served as an exploratory, pattern-detection tool to probe potential nonlinearities and higher-order interactions.

The SEM results indicate that symbolic and narrative cultural-memory pathways are positively associated with place attachment, and that place attachment exerts positive effects on all three TPB antecedents—attitude toward the behaviour, subjective norm, and perceived behavioural control. The TPB antecedents are positively associated with behavioural intention, and behavioural intention is positively associated with the CHTTE index. Random Forest analyses broadly converged with the SEM findings at the pattern level: TPB constructs and place attachment jointly provided the strongest predictive signal for behavioural intention, while behavioural intention, place attachment, and narrative cues were comparatively more important for CHTTE, with symbolic and ritual elements making smaller but detectable contributions. Given the moderate out-of-bag and cross-validated R² values and the conceptual proximity among several predictors, these important patterns are interpreted at the level of construct families (TPB vs. cultural-memory blocks) rather than as precise rank-orderings of individual variables, and they function as robustness checks rather than as high-precision predictive models.

Theoretically, the study proposes and offers context-specific empirical evidence consistent with an integrated framework that extends dominant PA–TPB and TPB-based tourism models in three ways. First, it explicitly positions cultural-memory pathways as upstream cognitive–symbolic inputs that precede both place attachment and TPB components, rather than starting from attitudes, satisfaction, or service quality alone. Second, place attachment is modelled as an affective hub that can shape all three TPB antecedents simultaneously, rather than serving only as a parallel predictor of behavioural intention or as a simple mediator between satisfaction and intention. Third, the model extends the traditional TPB outcome space by linking these mechanisms not only to behavioural intention but also to a broad index of cultural-heritage tourism experience. The framework is intended to be analytically transferable—a theoretically grounded structure that can be tested, adapted, and refined in other heritage contexts—rather than statistically generalisable in its exact parameter estimates from Sanhe Ancient Town.

At the same time, several measurement features of the present study warrant a cautious interpretation of effect sizes. The AVE values for place attachment and CHTTE are modest, some CHTTE items exhibit relatively low loadings, and several theoretically adjacent constructs (such as PA–SN and NP–PA) display relatively high correlations despite meeting conventional HTMT and Fornell–Larcker criteria. These properties suggest that the latent variables are statistically separable but conceptually proximate. Consequently, path coefficients involving PA and CHTTE are best viewed as approximate indicators of direction and relative strength rather than as definitive estimates of magnitude, and the development of more fine-grained, site-sensitive scales for cultural-memory pathways, attachment facets, and heritage experience remains an important task for future research.

Practically, the findings underscore the importance of designing heritage environments that coordinate symbolic, ritual, and narrative elements in ways that foster place attachment and support TPB-related predispositions. In the case of Sanhe Ancient Town, symbolic and narrative cues showed stronger linear associations with attachment in this dataset, whereas ritual practices appeared more limited and context-dependent in the present setting. To translate these insights into a design-oriented format, the study outlines a VR narrative prototype in which the three cultural-memory pathways are linked to illustrative visual and interaction tactics (e.g., symbolic architectural cues, embodied ritual interactions, story-driven spatial navigation) and to their proximal psychological targets in the TPB framework. In this study, this pathway–mechanism–TPB matrix is offered solely as a theory-informed design proposal and a set of hypothesis-generating scenarios rather than as an empirically evaluated intervention. It is intended to serve as a testable design template that future work can implement and test in dedicated VR experiments, thereby providing a replicable bridge between behavioural theory and practical experience design in cultural heritage tourism.

Finally, several limitations define the scope of the conclusions and point to directions for future research. The data were collected at a single heritage destination within a limited time window, so seasonal dynamics, cross-site differences, and longer-term behavioural outcomes could not be examined. The cross-sectional design precludes strong causal claims about the temporal ordering among cultural-memory cues, attachment, TPB constructs, and experience outcomes. The measurement of CHTTE relied on relatively general items that may not fully capture key heritage-specific dimensions such as cultural learning, authenticity, and atmosphere, and qualitative insights into how visitors interpret ritual activities at Sanhe were beyond the scope of this survey. Subsequent studies could address these limitations by combining refined, site-specific measurement scales with mixed-method designs, multi-site or longitudinal surveys, experimental and quasi-experimental evaluations of VR and on-site interpretation, and qualitative work on ritual distinctiveness and narrative coherence. Within these boundaries, the present findings suggest that integrating cultural memory, place attachment, and planned behaviour offers a promising and analytically transferable lens for understanding—and ultimately designing—meaningful cultural heritage tourism experiences.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, Y.Q.; Methodology, Y.Q.; Software, X.P.; Validation, E.J.; Formal analysis, Y.Q. and X.P.; Investigation, Y.Q. and X.P.; Writing—original draft preparation, Y.Q.; Writing—review and editing, E.J.; Visualization, Y.Q. and X.P.; Supervision, E.J. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data availability

All data generated or analyzed during this study are included in this article. The raw data are available from the First Author Yanran Qian upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Informed consent

Informed consent was obtained from all participants involved in the study.

Institutional review board

Upon review by the Hanyang University Institutional Review Board, the experimental design and research plan have been deemed scientifically sound, fair, and impartial, posing no harm or risk to participants. Participant recruitment is based on the principles of voluntariness and informed consent, with full protection of participants ’ rights and privacy. The study involves no conflict of interest and does not violate ethical standards or legal regulations.