Comparing Emotional Brand Attachment in Small-Group and Network-Based Brand Communities

Abstract

Emotional brand attachment is a key driver of long-term consumer–brand relationships, yet limited research has examined how different brand community structures foster such attachment. This study compares the effects of participation in small-group versus network-based brand communities and investigates the moderating role of regulatory focus. Using an online survey of 1,773 consumers from two Chinese-speaking markets, hierarchical regression analyses were conducted while controlling for product involvement, self-congruence, and public self-consciousness. The results show that brand community participation enhances emotional brand attachment, with participation in small-group communities exerting a significantly stronger effect than participation in network-based communities. Promotion-focused consumers develop stronger emotional attachment through small-group participation, whereas prevention-focused consumers exhibit weaker attachment in network-based communities. This study contributes to the literature by demonstrating how community structure embeds psychological distance, thereby extending construal-level theory, and by identifying regulatory focus as a key boundary condition shaping emotional outcomes in brand communities. The findings also offer practical insights for designing brand community strategies that align with consumers’ motivational orientations.

Plain Language Summary

This Study Explores How People’s Experiences in Online Brand Communities Shape Their Emotional Connection to Brands

This study looks at how people’s participation in online brand communities influences their emotional connection to brands. Brand communities can take various forms, ranging from smaller, more interactive groups to larger, network-style groups. We wanted to know whether these different types of communities affect people’s feelings toward a brand in different ways, and whether personal motivation plays a role. We collected survey data from 1,773 participants, including students, professionals, and government employees. We measured their participation in brand communities, their emotional attachment to brands, and their personal motivational style, known as “regulatory focus.” A promotion-focused mindset emphasizes growth and achievement, while a prevention-focused mindset emphasizes security and avoiding mistakes. We also considered other factors, such as how involved people are with the product and how much they care about their public image. Our results show that joining brand communities generally strengthens people’s emotional bond with brands. However, the type of group and the individual’s motivational style matter. People with a promotion focus feel more strongly attached to brands when they are active in small, interactive groups. In contrast, prevention-focused individuals do not show the same increase in attachment, especially in large network-style groups. These findings suggest that companies should design brand communities with different consumer mindsets in mind. For example, creating small, engaged groups may be more effective for customers who seek growth and achievement, while larger communities may not deliver the same benefits for prevention-focused individuals. In short, our study underscores the importance of tailoring brand community strategies to align with individuals' motivations, enabling brands to foster stronger and more meaningful consumer relationships.

Introduction

Emotional brand attachment is a central determinant of enduring consumer–brand relationships and a key driver of brand loyalty and advocacy (Aboulnasr & Tran, 2020; Grisaffe & Nguyen, 2011; Junaid et al., 2019). Consumers who develop strong emotional ties to a brand are more likely to engage in repeat purchase, tolerate price fluctuations, and maintain long-term relational commitment, even under adverse conditions (Dunn & Hoegg, 2014; Jang et al., 2015; Thomson et al., 2005). Despite its well-documented outcomes, prior research on emotional brand attachment has largely focused on internal psychological antecedents, such as self-congruence, product involvement, and public self-awareness, while paying comparatively less attention to firm-driven relational environments, including brand communities (Malär et al., 2011). As firms increasingly invest in long-term relationship-building strategies, understanding how marketing instruments such as brand communities shape emotional attachment has become an important theoretical and managerial issue (Jeong et al., 2021).

Brand communities refer to consumer collectives organized around shared enthusiasm for a brand (Albert et al., 2008) and play an important role in enhancing consumer engagement and relational outcomes. Digital brand communities facilitate the dissemination of brand-related information (Kim & John, 2008; Simon & Tossan, 2018), strengthen social bonds among members (McAlexander et al., 2002), and contribute to the development of durable consumer–brand relationships (Algesheimer et al., 2005). Prior research documents both short-term outcomes, such as accelerated word-of-mouth (Brodie et al., 2013; Brown et al., 2003), and longer-term outcomes, including heightened engagement (Baldus et al., 2015; Muñiz & Schau, 2005; Wirtz et al., 2013) and increased brand loyalty (Chang et al., 2013; Franke & Shah, 2003; Schouten et al., 2007). Although these studies demonstrate the relational benefits of brand communities, most research has focused on behavioral outcomes rather than examining emotional brand attachment as the affective mechanism through which community participation shapes long-term consumer-brand relationships. Consequently, existing findings remain fragmented, and limited research has systematically investigated how participation in brand communities contributes to the formation of emotional attachment, a gap that the present study seeks to address.

Brand communities differ substantially in structure and interaction intensity. Dholakia et al. (2004) identify multiple community forms and broadly distinguish between small-group communities characterized by close, repeated interaction and network-based brand communities characterized by diffuse, low-frequency interaction. Prior studies suggest that community structure shapes consumer motivation and participation patterns (Ma & Agarwal, 2007; Reid & Hogg, 2005; Scarpi, 2010; Zaglia, 2013). Recent research has further emphasized the importance of understanding interaction dynamics in digital brand communities, including online community behavior (Efendioğlu, 2024), the effects of social media content (Goh et al., 2013), and brand interaction processes (Zhang & Zhang, 2023). Despite these advances, existing studies have not directly compared whether participation in small-group versus network-based brand communities results in different levels of emotional brand attachment, even though theory suggests that these structures differ in intimacy, psychological distance, and relational depth. This lack of systematic comparison represents a substantive gap in the brand community literature that the present study addresses.

To further explain when and for whom brand community participation enhances emotional brand attachment, this study incorporates regulatory focus as a moderating mechanism. Regulatory focus theory distinguishes between promotion-focused and prevention-focused motivations, which systematically shape individuals’ goals, cognitive processing, and emotional responses (Higgins, 1998; Leonardelli et al., 2007). Promotion-focused consumers are oriented toward aspirations, growth, and ideal-self attainment, whereas prevention-focused consumers emphasize security, responsibility, and the avoidance of negative outcomes. These motivational orientations are closely linked to two self-related processes: self-enhancement and self-verification (Cowart et al., 2008; Leonardelli et al., 2007). Self-enhancement aligns with promotion focus and ideal-self expression, whereas self-verification aligns with prevention focus and actual-self consistency. Although these mechanisms are well established in psychology, they have rarely been integrated into explanations of how different brand community structures shape emotional brand attachment. By examining regulatory focus as a boundary condition, the present study provides a theoretically grounded account of how community structure and motivational orientation jointly influence emotional attachment.

Although prior brand community research often draws on social identity theory to explain consumer attachment, identity-based explanations typically treat communities as structurally homogeneous, thereby overlooking how differences in interaction density, intimacy, and psychological distance shape the development of emotional brand attachment. As a result, existing studies cannot explain whether small-group and network-based communities activate distinct self-related processes, such as self-enhancement or self-verification, nor can they predict whether these structures generate different emotional outcomes. By conceptualizing community structure as a psychological and motivational context rather than merely a social grouping, the present study moves beyond traditional identity frameworks and offers a distinct theoretical perspective.

Despite growing interest in brand communities, three gaps remain. First, no study has systematically compared how participation in small-group versus network-based communities shapes emotional brand attachment. Second, although regulatory focus is central to consumer motivation, its moderating role within brand communities remains unclear. Third, prior research has not integrated self-enhancement and self-verification mechanisms into explanations of how different community structures generate emotional brand attachment.

Accordingly, this study addresses these gaps by pursuing three objectives: (1) to examine whether brand community participation enhances emotional brand attachment; (2) to compare the differential effects of small-group versus network-based participation; and (3) to investigate how promotion and prevention regulatory focus moderate these effects.

This study contributes to the literature in three important ways. First, it extends construal-level theory by demonstrating that psychological distance is embedded in community structures, with small-group participation providing a close, concrete environment that strengthens emotional attachment. Second, it integrates self-enhancement and self-verification mechanisms into the brand community context, illustrating how different community structures activate distinct motivational processes. Third, it advances regulatory focus theory by identifying promotion and prevention orientations as boundary conditions for the emotional outcomes of community participation.

The remainder of this paper is structured as follows: Section “Literature Review” reviews the relevant literature; Section “Conceptual Framework and Hypotheses” develops the hypotheses; Section “Method” describes the methodology; Section “Results” presents the empirical results; Section “Discussion” discusses theoretical and managerial implications; and Section “Conclusion” concludes the paper.

Literature Review

Brand Community Participation

Brand community participation refers to consumers’ active involvement in brand-related community activities and interactions with other members (Tsai et al., 2012). A brand community represents a specialized, non-geographically bound social group organized around shared admiration for a brand (Muñiz & O’Guinn, 2001). Through participation in community traditions, rituals, and interactions, members reinforce shared meanings, sustain the community’s culture, and strengthen their perceived connection to both the brand and fellow members (Tsai et al., 2012). Prior research suggests that brand community participation facilitates knowledge dissemination and provides emotional and informational support, thereby deepening consumer–brand relationships (Koh & Kim, 2004). Participation enables consumers to express their identities, engage in social signaling, and reinforce self-related meanings through brand-related interactions, making participation a critical antecedent of emotional brand attachment.

Brand communities can be conceptually classified using sociological distinctions between neighborhood solidarities and social networks (Wellman, 1999). Neighborhood solidarities are characterized by dense interpersonal ties and close, repeated interaction, whereas social networks consist of loosely connected individuals with weaker relational bonds. A similar distinction appears in social psychology between common-bond groups and common-identity groups (Prentice et al., 1994; Sassenberg, 2002). Common bond groups derive cohesion from interpersonal relationships, while common identity groups are held together by shared identification with the group as a whole. This conceptual distinction provides a useful theoretical foundation for distinguishing between small-group brand communities, characterized by close interpersonal interaction, and network-based brand communities, characterized by diffuse, low-intensity interaction. By distinguishing participation across these two community structures, prior classification frameworks offer a basis for examining how different forms of participation generate distinct relational and emotional outcomes.

Members of virtual brand communities are typically defined by their online presence rather than by deep interpersonal ties with specific individuals (Zheng et al., 2023). Such communities consist of geographically distributed individuals connected through loosely structured and dynamic networks centered on shared thematic interests, corresponding to network-based virtual communities in the literature. In contrast, small-group-based virtual communities are rooted in preexisting interpersonal relationships rather than being defined solely by the digital platform. These communities comprise tightly connected individuals who maintain close interpersonal networks and use online interactions to coordinate shared goals while reinforcing existing offline relationships. This distinction highlights meaningful variation in structural cohesion and interpersonal embeddedness across virtual communities and provides a theoretical rationale for differentiating between network-based and small-group brand communities in consumer research.

Emotional Brand Attachment

In this research model, emotional brand attachment is the dependent variable, referring to the deep emotional ties and affection consumers hold toward a given brand (Thomson et al., 2005). It comprises three core feelings: affection, passion, and connection (Malär et al., 2011): affection reflects warm feelings toward a brand, passion captures more intense and energizing emotions, and connection represents a psychological bond and brand self-linkage (Shimul, 2022; Thomson et al., 2005). Prior research suggests that repeated brand experiences and meaningful interactions strengthen attachment over time (Escalas & Bettman, 2005; Lastovicka & Sirianni, 2011; Park et al., 2010; Thomson et al., 2005), and emotional attachment is crucial for consumption (Xu et al., 2022). For consumers, such ties facilitate identity differentiation and provide social belonging, self-expression, and enhanced self-efficacy (Chen et al., 2017; Joseph et al., 2021; McLeod et al., 2020; K. Y. Wang et al., 2024). Emotional brand attachment also drives repeated purchases, brand loyalty, and deeper customer engagement (Bahri-Ammari et al., 2016; Hwang et al., 2021; Pedeliento et al., 2016; Shimul, 2022) and is associated with improved financial performance and firm value (O’Neill & Carlbäck, 2011). Prior studies identify multiple antecedents of emotional brand attachment, including customer trust (Hwang et al., 2021), brand reliability (Japutra et al., 2019), brand distance and brand-self prominence (Park et al., 2013), and self-congruence, product involvement, and public self-awareness (Malär et al., 2011). Nevertheless, most work emphasizes direct consumer-brand interactions, while relational dynamics within virtual brand communities, where consumers interact with one another in addition to the brand, remain insufficiently examined.

Although emotional brand attachment is often discussed alongside relational constructs such as brand trust and brand commitment, it is conceptually distinct. Emotional attachment reflects a deep affective bond characterized by affection, passion, and psychological connection to the brand (Park et al., 2010; Thomson et al., 2005). In contrast, brand trust refers to consumers’ confidence in a brand’s reliability and integrity, whereas brand commitment reflects an intention to maintain a long-term relationship (Burke & Stets, 1999). Because attachment is more identity-relevant and affect-laden, it is especially responsive to social and symbolic environments such as brand communities where interpersonal meaning-making occurs. Clarifying this distinction supports treating emotional brand attachment as a unique construct in this study and helps explain why community structure may differentially shape its formation.

Regulatory Focus

This study examines the moderating role of regulatory focus in the relationship between brand community participation and emotional brand attachment, providing insight into how different community environments shape consumers’ psychological responses. According to regulatory focus theory, individuals pursue goals and regulate their behavior based on distinct motivational orientations that may operate as chronic dispositions or be situationally activated (Higgins & Pinelli, 2020). Regulatory focus can also manifest at the group level, referred to as collective regulatory focus (Faddegon et al., 2008). Originally proposed by Higgins (1998), regulatory focus theory distinguishes between two self-regulatory systems: promotion and prevention. A promotion focus is oriented toward aspirations, growth, and the pursuit of positive outcomes, whereas a prevention focus emphasizes security, responsibility, and the avoidance of negative outcomes (Jing et al., 2024). These two motivational orientations function as independent dimensions rather than opposite ends of a single continuum (Johnson et al., 2015), allowing individuals or groups to exhibit high levels of both, only one, or neither (Guo et al., 2024; Shin et al., 2016).

These distinct motivational orientations are critical for understanding how consumers interpret and respond to different brand community structures. Small-group communities, which offer intimacy and relational depth, may resonate more strongly with promotion-focused consumers seeking relational enrichment and opportunities for growth. In contrast, the diffuse and less predictable nature of network-based communities may be less appealing to prevention-focused consumers who prioritize stability, safety, and risk avoidance. Accordingly, regulatory focus provides a theoretically grounded basis for predicting heterogeneous effects of brand community participation on emotional brand attachment.

Prior research further suggests that promotion- and prevention-focused consumers differ systematically in information processing and decision-making. Promotion-focused consumers tend to pursue advancement and positive gains and rely more on heuristic and rapid judgments, whereas prevention-focused consumers emphasize obligation fulfillment and vigilance, evaluate products through a safety-oriented lens, and engage in more systematic and effortful information processing (Freitas & Higgins, 2002; Pula et al., 2014; Roy et al., 2024; X. Wang et al., 2022; Werth & Foerster, 2007). These differences imply that consumers with distinct regulatory orientations may respond differently to the relational depth and social cues embedded in various forms of brand community participation.

Control Variables

Three control variables were included in this study: product involvement, public self-awareness, and self-congruence. Product involvement reflects the personal relevance and perceived importance of a product; consumers with higher involvement tend to evaluate brand attributes more carefully and exhibit stronger emotional responses to brands (Heitz-Spahn et al., 2024; Malär et al., 2011). Public self-awareness captures the extent to which individuals attend to how others perceive them. Consumers high in public self-awareness tend to make brand choices strategically to manage social impressions, whereas those low in public self-awareness are less influenced by social evaluation and are more likely to form brand attachment based on self-congruence (Fenigstein et al., 1975; Rasool et al., 2024). Self-congruence refers to the perceived match between consumers’ self-concept and the brand (Gecas, 1982; Shimul et al., 2024; Shimul & Phau, 2022), encompassing both actual self-congruence and ideal self-congruence (Malär et al., 2011). Prior research indicates that both dimensions contribute to the formation of emotional brand attachment.

Conceptual Framework and Hypotheses

Figure 1 illustrates the conceptual framework of this study. Brand community participation serves as the independent variable and is examined at both the overall and subgroup levels. The framework distinguishes between participation in small-group communities, characterized by dense interpersonal interaction, and participation in network-based communities, characterized by more diffuse and lower-frequency interaction. These two forms of participation are hypothesized to influence the dependent variable, emotional brand attachment. Accordingly, H1 tests the overall effect of brand community participation, whereas H1a and H1b compare the differential effects of small-group and network-based participation.

Figure 1.

Conceptual framework.

Regulatory focus is incorporated as a moderating variable. Promotion and prevention foci represent distinct motivational orientations that shape how consumers interpret and respond to community experiences. H2, H2a, and H2b propose that regulatory focus moderates the relationship between brand community participation and emotional brand attachment at both the overall and subgroup levels. Furthermore, H3, H3a, and H3b specify differential moderating effects for promotion versus prevention focus, reflecting the expectation that promotion-focused consumers benefit more from highly interactive small-group communities, whereas prevention-focused consumers exhibit weaker emotional attachment in diffuse network-based communities. Together, these hypothesized relationships provide an integrated framework for examining how community structure and motivational orientation jointly shape emotional brand attachment.

Main Effects of Small-Group and Network-based Brand Community Participation on Emotional Brand Attachment

This hypothesis is grounded in construal level theory (Trope & Liberman, 2010), which explains how psychological distance shapes individuals’ emotional responses. Because closer psychological distance induces low-level, concrete construals, the degree of relational closeness experienced within brand communities can meaningfully influence emotional brand attachment.

In addition, brand community participation can promote both self-enhancement and self-verification processes (Hammedi et al., 2015). Through self-enhancement, consumers reinforce aspects of their ideal self, whereas self-verification affirms consistency with their actual self-concept (Escalas & Bettman, 2005). Both mechanisms deepen psychological alignment with the brand and strengthen emotional attachment. Prior research consistently shows that self-enhancement and self-verification generate positive emotional responses toward brands, thereby enhancing emotional brand attachment (Malär et al., 2011).

Accordingly, participation in brand communities, by providing opportunities for both ideal-self expression and actual-self affirmation, is expected to enhance emotional brand attachment. Hence, we propose the following hypothesis:

• H1: Brand community participation (BCP: Sample) is positively associated with emotional brand attachment (EB).

Relative Effect of Small-Group Virtual Communities Participation and Network-Based Participation on Emotional Brand Attachment

Drawing on interpersonal relationship theory, self-enhancement is most likely to emerge in low-risk and low-cognitive-demand environments, such as small-group virtual communities (De Valck et al., 2009; Heinonen & Strandvik, 2009; Presi et al., 2014; Swann et al., 1990). When interaction partners pose minimal relational risk, individuals tend to prefer contexts that affirm their self-view (Hixon & Swann, 1993). Accordingly, consumers with strong self-enhancement motives are more inclined to initiate interactions in intimate online communities, where relational closeness provides a psychologically safe setting for identity expression (De Valck et al., 2009; Hanks et al., 2024; Heinonen & Strandvik, 2009; Presi et al., 2014).

Self-enhancement theory further suggests that individuals are motivated to elevate their sense of personal worth (Sedikides & Strube, 1997) and to move closer to their ideal selves to bolster self-esteem (Higgins, 1987). When a brand aligns with consumers’ ideal self, it facilitates aspirational identity expression and strengthens emotional attachment. Small-group communities are particularly conducive to this process because their interpersonal intimacy amplifies opportunities for ideal-self affirmation (Agyekum et al., 2025; Boldero & Francis, 2002; Grubb & Grathwohl, 1967).

Accordingly, participation in small-group communities should lead to stronger emotional bonding than participation in other community structures. Thus, we propose the following hypothesis:

• H1a: Small-group participation (BCP: Small-group) has a positive impact on emotional brand attachment (EB).

Drawing on self-expansion theory, individuals possess an inherent motivation to incorporate others, including brands, into their self-concept, such that emotional attachment to the brand strengthens as it becomes integrated into the self (Aron et al., 2005). Accordingly, emotional brand attachment is closely related to consumers’ self-concept (Kleine et al., 1993). Unlike small-group communities, network-based communities attract consumers primarily through topic relevance and informational value rather than interpersonal closeness (Dholakia et al., 2004; Prentice et al., 2019). In these settings, consumers rely on informational feedback from other users, even without personal familiarity, to evaluate and confirm aspects of their self-concept (Gilal et al., 2025), making network-based communities particularly conducive to self-concept evaluation and confirmation.

Self-verification theory further suggests that individuals are motivated to confirm and maintain their self-concepts by seeking environments that validate their self-perceptions and avoiding those that threaten them (Swann, 1983; Swann et al., 1992). Successful self-verification generates psychological benefits such as security, comfort, and predictability, which reinforce relational bonds (Burke & Stets, 1999) and promote consistency between self-perception and external feedback (Lecky, 1945). Consequently, self-verification provides psychological safety and stability that support emotional attachment (Robinson & Smith-Lovin, 1992; Swann & Read, 1981).

Prior research indicates that individuals tend to engage with and form connections with entities that validate their self-image, while distancing themselves from those that do not (Min et al., 2020; Shimul & Phau, 2023). During self-verification, individuals experience cognitive and affective responses as they evaluate whether feedback affirms their self-view (Leary, 2007). In virtual communities, emotions elicited by successful self-verification enhance interactions and strengthen the bond between the brand and the validating community environment, thereby fostering emotional attachment and reinforcing commitment (Burke & Stets, 1999).

One effective way to achieve self-verification is to engage with brands whose personalities align with consumers’ actual selves. Such alignment provides self-relevant affirmation, enhancing positive affect and deepening emotional attachment (Malär et al., 2011; Shimul, 2022). Accordingly, participation in network-based brand communities, where information exchange and self-concept confirmation are emphasized over interpersonal closeness, is expected to facilitate self-verification processes and strengthen emotional brand attachment. Thus, we propose the following hypothesis:

• H1b: Network-based participation (BCP: Network-based) has a positive effect on emotional brand attachment (EB).

If both H1a and H1b are supported, indicating that participation in both small-group and network-based brand communities enhances emotional brand attachment, an important theoretical question arises regarding the relative strength of these effects. Given the substantial differences between these community types in relational depth, interaction patterns, and underlying psychological mechanisms, it is necessary to determine which form of participation exerts a stronger influence on emotional brand attachment.

This comparison is grounded in construal level theory, which explains how psychological distance shapes mental construals (Liberman & Trope, 1998). According to this theory, greater psychological distance leads individuals to adopt abstract, high-level construals, whereas reduced distance results in concrete, low-level construals (Liberman et al., 2007; Saeed et al., 2024). When objects or events are psychologically distant across dimensions such as time, space, or social proximity, individuals rely on simplified, abstract representations; conversely, psychological closeness enables access to rich contextual information and promotes concrete construals (Liberman et al., 2007; Saeed et al., 2024).

Construal level theory further helps explain differences in emotional reactions across brand community types and has been widely applied in consumer behavior research (e.g., Kim & John, 2008). Importantly, self-enhancement and self-verification motives are systematically linked to psychological distance. Self-verification is generally experienced as psychologically closer, whereas self-enhancement is associated with greater psychological distance, leading to differences in construal level and emotional intensity (Liberman et al., 2007).

Small-group communities reduce psychological distance by fostering dense interpersonal interaction and relational intimacy, thereby facilitating both self-enhancement and self-verification. When self-enhancement opportunities are embedded in tightly knit communities, perceived psychological distance is reduced, particularly when ideal self-congruence is high, leading consumers to construe such communities at a lower, more concrete level (Chatterjee, 2025). In contrast, network-based communities are more likely to evoke abstract construals due to weaker interpersonal ties and greater psychological distance.

Prior research indicates that greater psychological distance attenuates emotional intensity, whereas psychologically close and concretely construed experiences elicit stronger affective responses (Williams & Bargh, 2008). Accordingly, differences in psychological distance and construal level imply that participation in small-group communities should generate stronger emotional brand attachment than participation in network-based communities. Hence, we propose the following hypothesis:

• H4: Small-group participation has a stronger effect on emotional brand attachment than network-based participation.

Moderating Effects

Prior research suggests that positive and negative emotions are associated with distinct behavioral tendencies (Frijda et al., 1989), which align with promotion- and prevention-focused motivational orientations (Higgins et al., 1997; Watson et al., 1999). In interpersonal contexts, promotion-focused goals emphasize the pursuit of positive outcomes such as growth, intimacy, and personal development, whereas prevention-focused goals emphasize avoiding negative outcomes such as conflict, rejection, or loss.

Regulatory focus similarly shapes how consumers interpret and respond to brand relationships. Promotion-focused consumers approach brands with growth-oriented motivations and tend to engage in self-enhancement and proactive interaction. In contrast, prevention-focused consumers adopt a vigilant orientation, emphasizing risk avoidance and consistency, and may distance themselves from brands when they perceive potential threats or incongruence. Prior studies indicate that favorable brand relationships are more likely to activate promotion-focused tendencies, whereas unfavorable or uncertain relationships can trigger prevention-focused responses aimed at self-protection (Aaker & Lee, 2001; Chernev, 2004).

Importantly, promotion and prevention foci are systematically linked to different forms of self-congruence. Promotion focus facilitates ideal self-congruence by emphasizing aspirations and growth, whereas prevention focus facilitates actual self-congruence by prioritizing safety, responsibility, and self-verification (Malär et al., 2011). Accordingly, when consumers’ regulatory focus aligns with a promotion orientation, achieving ideal self-consistency becomes more salient, thereby supporting self-improvement. Conversely, alignment with a prevention orientation strengthens the motivation to maintain actual self-consistency and reinforces self-verification processes.

Based on these arguments, regulatory focus is expected to moderate the relationship between brand community participation and emotional brand attachment differently across community types. Thus, we propose the following hypotheses:

• H2: The moderating effect of promotion focus (RF_PM) on the relationship between brand community participation (BCP: Sample) and emotional brand attachment (EB) will differ by community type.

• H2a: Promotion focus (RF_PM) strengthens the positive relationship between small-group participation (BCP: Small-group) and emotional brand attachment (EB).

• H2b: Promotion focus (RF_PM) weakens the positive relationship between network-based participation (BCP: Network-based) and emotional brand attachment (EB).

• H3: The moderating effect of prevention focus (RF_PF) on the relationship between brand community participation (BCP: Sample-group) and emotional brand attachment (EB) will differ by community type.

• H3a: Prevention focus (RF_PF) weakens the positive relationship between small-group participation (BCP: Small-group) and emotional brand attachment (EB).

• H3b: Prevention focus strengthens (RF_PF) the positive relationship between network-based (BCP: Network-based) and emotional brand attachment (EB).

Collectively, these hypotheses explain how community structure and motivational orientation jointly shape emotional brand attachment. Small-group brand communities are expected to strengthen emotional attachment through self-enhancement and ideal-self expression, whereas network-based communities foster attachment through self-verification and actual-self alignment. Promotion and prevention foci are therefore expected to moderate these relationships in opposite directions across the two community types. The next section describes the research design and measurement procedures used to test these predictions empirically.

Method

Data Collection and Sample

Primary data were collected through an online questionnaire administered between December 1 and December 31, 2024. The survey was hosted on Wenjuanxing, which also supported respondent recruitment, and additional distribution took place via LINE and WeChat to broaden coverage across Chinese-speaking markets. Participants received a unique access link directing them to a restricted survey page. A total of 2,923 responses were collected. After removing incomplete or invalid cases, 1,773 valid questionnaires remained and were used for subsequent analyses (Table 1). Among these respondents, 74.22% resided in China and 25.78% in Taiwan. The sample characteristics, gender, age, education, occupation, and monthly income are summarized in Table 1.

Table 1.

Demographic Characteristics of the Samples.

Category	Item	Abbr.	Sample (N = 1,773)		Small ( N = 1,252)		Network ( N = 407)
			n	(%)	n	(%)	n	(%)
Product type	Fast moving consumer goods	Pdt_1	630	35.53	421	33.63	144	35.38
	Durable consumer goods	Pdt _2	618	34.86	454	36.26	137	33.66
	Service industry	Pdt _3	465	26.23	327	26.12	119	29.24
	Agriculture, forestry, fishing, and animal Husbandry Goods	Pdt _4	60	3.38	50	3.99	7	1.72
Gender	Male	Male	1088	61.36	788	62.94	248	60.93
	Female	Female	685	38.64	464	37.06	159	39.07
Age	Under the age of 20 years	Age _1	405	22.84	290	23.16	90	22.11
	21∼30 age group	Age _2	218	12.30	141	11.26	57	14.00
	31∼40 age group	Age _3	394	22.22	264	21.09	98	24.08
	41∼50 age group	Age _4	549	30.96	401	32.03	117	28.75
	51∼60 age group	Age _5	204	11.51	155	12.38	43	10.57
	61 years old or above	Age _6	3	0.17	1	0.08	2	0.49
Educational	Junior high school and below	Edu_1	88	4.96	73	5.83	12	2.95
	High school/ Technical	Edu _2	435	24.53	319	25.48	92	22.60
	Undergraduate	Edu _3	1,015	57.25	714	57.03	236	57.99
	Master’s degree or above	Edu _4	235	13.25	146	11.66	67	16.46
Occupation	Manufacturing industry	Occ_1	416	23.46	321	25.64	82	20.15
	Financial industry	Occ_2	155	8.74	94	7.51	48	11.79
	Information industry	Occ_3	114	6.43	73	5.83	31	7.62
	Service industry	Occ_4	391	22.05	266	21.25	100	24.57
	Public servant	Occ_5	236	13.31	173	13.82	48	11.79
	Students	Occ_6	442	24.93	312	24.92	95	23.34
	Other occupations	Occ_7	19	1.07	13	1.04	3	0.74
Monthly income	TWD 0–9,999	Inc_1	395	22.28	282	22.52	83	20.39
	TWD 10,000–19,999	Inc_2	182	10.27	141	11.26	30	7.37
	TWD 20,000–29,999	Inc_3	479	27.02	354	28.27	101	24.82
	TWD 30,000–39,999	Inc_4	411	23.18	290	23.16	94	23.1
	TWD 40,000–49,999	Inc_5	180	10.15	111	8.87	52	12.78
	TWD 50,000 above	Inc_6	126	7.11	74	5.91	47	11.55
Regions	China	CHN	1316	74.22	994	79.39	253	62.16
	Taiwan region	TWN	457	25.78	258	20.61	154	37.84

Note. Abbr. = abbreviation; TWD = New Taiwan Dollar.

Part of the sample is drawn from the Taiwan region.

To identify brand community participants, respondents were first asked whether they had joined any brand-related online community within the past 3 months (e.g., email lists, announcement boards, online forums, synchronous chat groups, browser-based community rooms, and multi-user virtual environments). Based on this screening question, 1,654 respondents reported community participation, while 119 reported no such participation.

Following Dholakia et al. (2004), participating respondents were classified according to their dominant interaction pattern within the focal community. Respondents who reported usually interacting with the same group of people and indicated a high level of interaction (scores of 4 or 5 on the interaction item) were classified as small-group community participants (n = 1,252; 62.94% male). Respondents who reported interacting with multiple groups and reported similarly high interaction levels were classified as network-based community participants (n = 407; 60.93% male).

Five respondents who reported participation in both community types were retained in the full sample but excluded from subgroup analyses to maintain group distinctiveness. Accordingly, the final dataset comprised 1,773 respondents, including 1,252 small-group participants, 407 network-based participants, and 119 non-participants. These classifications form the basis for testing H1a, H1b, and the proposed moderating hypotheses.

Reliability, Validity, and Common Method Variance Tests

All analyses were conducted using STATA 17.0, SPSS 20.0, and AMOS 23.0. Because brand community participation was operationalized as a categorical classification based on interaction patterns rather than a latent construct, reliability and validity assessments were not applicable to this variable.

Internal consistency reliability was assessed using Cronbach’s α and composite reliability (CR), while convergent validity was evaluated using average variance extracted (AVE). Following established criteria, CR values above .70 and AVE values above .50 indicate satisfactory reliability and convergent validity (Fornell & Lacker, 1981; Hair et al., 2019). CR was calculated based on standardized factor loadings using the following formula:

$$\mathrm{CR}=\frac{{(\sum {\lambda }_{\mathrm{i}})}^2}{(\sum {\lambda }_{\mathrm{i}}{)}^2+\sum {\theta }_{\mathrm{i}}}$$ (1)

where ${\lambda }_{\mathrm{i}}$ denotes the standardized loading of indicator i, and ${\theta }_{\mathrm{i}}=1-{\lambda }_{\mathrm{i}}^2$ represents the corresponding measurement error variance. AVE was computed as:

$$\mathrm{AVE}=\frac{{(\sum {\lambda }_{\mathrm{i}})}^2}{(\sum {\lambda }_{\mathrm{i}}{)}^2+\sum {\theta }_{\mathrm{i}}}$$ (2)

An AVE value greater than .50 indicates that a construct explains more than half of the variance of its indicators.

As reported in Table 2, Cronbach’s α values ranged from .831 to .952, exceeding the recommended threshold of .80. CR values ranged from .897 to .961, further confirming strong construct reliability.

Table 2.

Reliability and Validity Results of the Questionnaire Items.

Variables (code name)	Operational definition	Measurement items	SFL	CR	AVE	Cronbach’s α
Emotional Brand Attachment (EB)	- Dependent variable (continuous variable).- Refer to the items in Thomson et al. (2005), and divide the sum of all measured items by the number of items.	EB1 I think this brand community is affectionate.	0.872	.947	0.653	.937
		EB2 I think this brand community is full of love.	0.860
		EB3 I think this brand community is peaceful.	0.859
		EB4 I think this brand community is friendly.	0.858
		EB5 I can be attached the brand community.	0.850
		EB6 I bond this brand community with me.	0.786
		EB7 I can connect with others through this brand community.	0.835
		EB8 I think the brand community is passionate.	0.851
		EB9 I think this brand community is loath.	0.268
		EB10 I think this brand community is captivating.	0.849
Regulatory Focus Promotion (RF_PM)	- Independent variable (continuous variable).- Refer to the items in Lockwood et al. (2002) and divide the sum of all measured items by the number of items.	RF_PM3 I frequently imagine how I will achieve my hopes and aspirations.	0.858	.959	0.721	.948
		RF_PM5 I often think about the person I would ideally like to be in the future.	0.860
		RF_PM6 I typically focus on the success I hope to achieve in the future.	0.835
		RF_PM8 I often think about how I will achieve academic success.	0.863
		RF_PM12 My major goal in school right now is to achieve my academic ambitions.	0.852
		RF_PM14 I see myself as someone who is primarily striving to reach my “ideal self”—to fulfill my hopes, wishes, and aspirations.	0.864
		RF_PM16 In general, I am focused on achieving positive outcomes in my life.	0.861
		RF_PM17 I often imagine myself experiencing good things that I hope will happen to me.	0.809
		RF_PM18 Overall, I am more oriented toward achieving success than preventing failure.	0.839
Regulatory Focus Prevention (RF_PV)	- Independent variable (continuous variable).- Refer to the items in Lockwood et al. (2002) and divide the sum of all measured items by the number of items.	RF_PV1 In general, I am focused on preventing negative events in my life.	0.841	.961	0.733	.939
		RF_PV2 I am anxious that I will fall short of my responsibilities and obligations.	0.874
		RF_PV4 I often think about the person I am afraid I might become in the future.	0.821
		RF_PV7 I often worry that I will fail to accomplish my academic goals.	0.871
		RF_PV9 I often imagine myself experiencing bad things that I fear might happen to me.	0.853
		RF_PV10 I frequently think about how I can prevent failures in my life.	0.873
		RF_PV11 I am more oriented toward preventing losses than I am toward achieving gains.	0.840
		RF_PV13 My major goal in school right now is to avoid becoming an academic failure.	0.868
		RF_PV15 I see myself as someone who is primarily striving to become the self I “ought” to be—to fulfill my duties, responsibilities, and obligations.	0.862
Actual Self-Congruence (ASC)	- Control variable (continuous variable).- Refer to the items in Sirgy et al. (1997) and divide the sum of all measured items by the number of items.	ASC1 The personality of brand community is consistent with how I see myself.	0.925	.922	0.855	.831
		ASC2 The personality of brand community is a mirror image of me.	0.925
Ideal Self-Congruence (IDS)	- Control variable (continuous variable).- Refer to the items in Sirgy et al. (1997) and divide the sum of all measured items by the number of items.	IDS1 The personality of brand community is consistent with how I would like to be.	0.927	.924	0.859	.836
		IDS2 The personality of brand community is a mirror image of the person I would like to be.	0.927
Product Involvement (PI)	- Control variable (continuous variable).- Refer to the items in Trijp et al. (1996), and divide the sum of all measured items by the number of items.	PI1 Because of my personal attitudes, I feel that this is a product that ought to be important to me.	0.895	.897	0.655	.853
		PI2 Because of my personal values, I feel that this is a product that ought to be important to me.	0.896
		PI3 This product is not very important to me personally.	0.324
		PI4 Compared with other products, this product is important to me.	0.891
		PI5 I’m interested in this product.	0.878
Public Self-Consciousness (PSC)	- Control variable (continuous variable).- Refer to the items in Fenigstein et al. (1975) and divide the sum of all measured items by the number of items.	PSC1 I’m concerned about my style of doing things.	0.851	.943	0.703	.900
		PSC2 I’m concerned about the way I present myself.	0.877
		PSC3 I’m self-conscious about the way I look.	0.743
		PSC4 I usually worry about making a good impression.	0.870
		PSC5 One of the last things I do before I leave my house is look in the mirror.	0.869
		PSC6 I’m concerned about what other people think of me.	0.801
		PSC7 I’m usually aware of my appearance.	0.848

Note. SFL = Standardized factor loading; CR = Composite reliability; AVE = Average extracted variance.

Exploratory factor analysis (EFA) was conducted to examine the underlying factor structure. The Kaiser–Meyer–Olkin (KMO) measure was 0.902, and Bartlett’s test of sphericity was significant, indicating suitability for factor analysis. Principal component analysis extracted seven factors with eigenvalues greater than 1, and all retained items loaded above 0.50 without problematic cross-loadings. Standardized factor loadings for emotional brand attachment, regulatory focus (promotion and prevention), self-congruence (actual and ideal), product involvement, and public self-consciousness generally ranged from 0.653 to 0.859, supporting adequate convergent validity (Table 2).

One emotional brand attachment item (EB9) exhibited a lower standardized loading (0.268). Consistent with best-practice recommendations (Hair et al., 2019), indicators with low loadings may be retained when there is strong theoretical justification and overall construct validity is not compromised. EB9 is part of the original emotional brand attachment scale developed by Thomson et al. (2005), and its removal would alter the theoretical composition of this widely adopted construct. Importantly, the AVE for emotional brand attachment remained above 0.50, and re-estimation excluding EB9 produced substantively identical factor loadings, reliability coefficients, and hypothesis-testing results, indicating model stability.

A similar rationale applies to one product involvement item (PI3), which showed a standardized loading of 0.324. PI3 captures a value-driven dimension of involvement grounded in Trijp et al. (1996). Removing this item would narrow the construct’s conceptual domain, while AVE values remained above 0.50 and alternative estimations yielded substantively identical results. Retaining PI3 therefore, preserves conceptual completeness without compromising measurement robustness.

Because all data were collected using a single self-report survey, potential common method variance (CMV) was assessed. Procedural remedies were implemented, including ensuring anonymity and confidentiality, emphasizing that there are no correct or incorrect answers, and clarifying the academic purpose of the study (Podsakoff et al., 2003).

Statistically, Harman’s one-factor test was conducted. Seven factors with eigenvalues greater than 1 emerged, and the first factor accounted for 48.547% of the total variance, which does not exceed the commonly used 50% threshold. Accordingly, CMV is unlikely to pose a serious threat to the validity of the findings. The proportion of variance attributable to the first factor is calculated as:

$$\mathrm{CMV}(\%))=\frac{{\lambda }_{\mathrm{1}}}{\sum _{\mathrm{i}=1}^{\mathrm{k}}{\lambda }_{\mathrm{i}}}\times 100$$ (3)

where ${\lambda }_1$ denotes the eigenvalue of the first factor and $\sum _{\mathrm{i}=1}^{\mathrm{k}}{\lambda }_{\mathrm{i}}$ represents the sum of all eigenvalues. The computed value of 48.547% indicates that common method variance does not materially compromise the empirical results.

Description, Statistics, and Regression Model

The survey employed validated scales to measure emotional brand attachment, regulatory focus, self-congruence, product involvement, and public self-consciousness. All items were adapted from established international instruments and reviewed by subject-matter experts to ensure clarity and content validity. All constructs were measured using five-point Likert scales ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). Variable labels and abbreviations are reported in Table 3.

Table 3.

Descriptive Statistical Analysis.

Sample	N	Mean	SD	Min	Max	Skewness	Kurtosis
EB	1,773	3.437	0.781	1.000	5.000	−0.692	3.684
RF_PM	1,773	3.056	0.878	1.333	5.000	0.366	2.135
RF_PV	1,773	3.376	0.796	1.000	5.000	−1.075	3.996
ASC	1,773	3.380	0.916	1.000	5.000	−0.559	3.139
IDS	1,773	3.393	0.911	1.000	5.000	−0.554	3.149
PI	1,773	3.460	0.758	1.000	5.000	−0.493	3.357
PSC	1,773	3.325	0.636	1.429	5.000	−0.607	3.522
Small	N	Mean	SD	Min	Max	Skewness	Kurtosis
EB	1,252	3.699	0.440	1.800	5.000	0.709	4.724
RF_PM	1,252	2.712	0.746	1.333	5.000	0.960	3.437
RF_PV	1,252	3.677	0.402	1.444	5.000	0.176	4.312
ASC	1,252	3.692	0.592	1.000	5.000	0.085	3.469
IDS	1,252	3.691	0.594	1.500	5.000	0.146	3.261
PI	1,252	3.677	0.468	1.800	5.000	0.496	3.776
PSC	1,252	3.523	0.370	2.286	5.000	0.686	3.886
Network	N	Mean	SD	Min	Max	Skewness	Kurtosis
EB	407	2.795	1.117	1.000	5.000	0.477	2.076
RF_PM	407	3.954	0.583	1.889	5.000	−0.153	2.841
RF_PV	407	2.583	1.076	1.000	5.000	0.355	2.033
ASC	407	2.613	1.207	1.000	5.000	0.596	2.343
IDS	407	2.635	1.210	1.000	5.000	0.559	2.247
PI	407	2.924	1.101	1.000	5.000	0.487	2.022
PSC	407	2.799	0.890	1.429	5.000	0.542	2.189

Note. 1. The valid sample includes 1,773 respondents: 1,252 from small-group communities and 407 from network-based communities. Five respondents participated in both types of communities, and 119 respondents belonged to neither community type.

2. Diagnostic analyses of measurement variables are reported in Appendix 1.

Brand community participation was operationalized following Dholakia et al. (2004). Respondents identified the virtual brand community they engaged with most frequently and provided descriptive information, including community name, joining duration, interaction partners, and participation motives. They further indicated whether interactions typically occurred with the same members or with different groups, forming the basis for classifying participation as small-group or network-based, as described in Section “Data Collection and Sample.”

Emotional brand attachment was measured using the scale developed by Thomson et al. (2005), comprising three second-order dimensions, affection, connection, and passion, captured through ten indicators. Promotion and prevention orientations were measured using the scale of Lockwood et al. (2002). Actual and ideal self-congruence were assessed using the scale proposed by Sirgy et al. (1997). Product involvement was measured using items adapted from Trijp et al. (1996), supplemented with value-based statements. Public self-consciousness was measured using the seven-item scale developed by Fenigstein et al. (1975).

Descriptive statistics for the full sample and the two participation subgroups are presented in Table 3. Across the full sample (N = 1,773), emotional brand attachment had a mean of 3.437 (SD = 0.781). Promotion and prevention regulatory focus exhibited means of 3.056 (SD = 0.878) and 3.376 (SD = 0.796), respectively. Actual self-congruence (M = 3.380, SD = 0.916), ideal self-congruence (M = 3.393, SD = 0.911), product involvement (M = 3.460, SD = 0.758), and public self-consciousness (M = 3.325, SD = 0.636) all showed moderate to high levels across respondents.

Comparisons between community types revealed clear psychological differences. Small-group participants (N = 1,252) reported higher emotional brand attachment (M = 3.699, SD = 0.440) than network-based participants (M = 2.795, SD = 1.117). Small-group members also exhibited higher prevention focus (M = 3.677, SD = 0.402) and moderately high promotion focus (M = 2.712, SD = 0.746). In contrast, network-based participants (N = 407) showed higher promotion focus (M = 3.954, SD = 0.583) and lower prevention focus (M = 2.583, SD = 1.076).

Network-based participants also reported lower levels of actual self-congruence (M = 2.613, SD = 1.217) and ideal self-congruence (M = 2.635, SD = 1.210) than small-group participants. A similar pattern emerged for product involvement, with small-group members reporting higher involvement (M = 3.677, SD = 0.468) than network-based participants (M = 2.924, SD = 1.101). Overall, the descriptive statistics indicate systematic differences between small-group and network-based community participants, consistent with the proposed hypotheses (H1-H4).

Figure 2 presents nine scatterplots illustrating the bivariate relationships between emotional brand attachment (EB) and the three key predictors, brand community participation (BCP), promotion focus (RF_PM), and prevention focus (RF_PV), across the full sample, small-group participants, and network-based participants. Overall, the visual patterns suggest mild and heterogeneous associations across samples.

Figure 2.

Scatterplots of BCP, RF_PM, and RF_PV with Emotional Brand Attachment.

Specifically, BCP shows a slightly positive trend in the full sample and among small-group participants, while the association appears weaker among network-based participants. The relationship between RF_PM and EB is mixed, with generally weak or near-zero visual slopes across subsamples. In contrast, RF_PV exhibits a clearer positive association with EB in the full sample and the small-group subsample, though this pattern is less pronounced among network-based participants.

These scatterplots provide an initial exploratory overview of the data structure. However, because bivariate visualizations do not account for covariates or structural differences across community types, they offer only preliminary insights. Accordingly, the subsequent multivariate Ordinary Least Squares (OLS) models are required to formally estimate the net effects of BCP and regulatory focus while controlling for additional variables and sample heterogeneity.

The regression specifications used to test the proposed hypotheses are presented in Equations 4 to 6. Equation 4 estimates the main effects of brand community participation (BCP) on emotional brand attachment (EB) and is used to test H1, H1a, and H1b across the full sample, the small-group subsample, and the network-based subsample. Equation 5 incorporates promotion regulatory focus (RF_PM) and its interaction with BCP to examine the moderating effects proposed in H2, H2a, and H2b. Equation 6 similarly includes prevention regulatory focus (RF_PV) and its interaction with BCP to test the moderating effects specified in H3, H3a, and H3b. In all models, X and δ denote the vector of control variables, δ represents the corresponding coefficients, and ε is the stochastic error term.

$$\mathrm{E}{\mathrm{B}}_{\mathrm{i}}={\alpha }_0+{\alpha }_1\mathrm{BC}{\mathrm{P}}_{\mathrm{i}}+\delta X+{\epsilon }_{\mathrm{i}}$$ (4)

$$\begin{array}{c}\multicolumn{1}{c}{\mathrm{E}{\mathrm{B}}_{\mathrm{i}}={\beta }_0+{\beta }_1\mathrm{BC}{\mathrm{P}}_{\mathrm{i}}+{\beta }_2\mathrm{RF}\_\mathrm{P}{\mathrm{M}}_{\mathrm{i}}+{\beta }_3(\mathrm{BC}{\mathrm{P}}_{\mathrm{i}}*\mathrm{RF}\_\mathrm{P}{\mathrm{M}}_{\mathrm{i}})}\\ \multicolumn{1}{c}{+\delta X+{\epsilon }_{\mathrm{i}}}\end{array}$$ (5)

$$\begin{array}{c}\multicolumn{1}{c}{\mathrm{E}{\mathrm{B}}_{\mathrm{i}}={\gamma }_0+{\gamma }_1\mathrm{BC}{\mathrm{P}}_{\mathrm{i}}+{\gamma }_2\mathrm{RF}\_\mathrm{P}{\mathrm{V}}_{\mathrm{i}}+{\gamma }_3(\mathrm{BC}{\mathrm{P}}_{\mathrm{i}}*\mathrm{RF}\_\mathrm{P}{\mathrm{V}}_{\mathrm{i}})}\\ \multicolumn{1}{c}{+\delta X+{\epsilon }_{\mathrm{i}}}\end{array}$$ (6)

Results

Main Effects of Brand Community Participation

Table 4 reports the results of the hierarchical regression analyses examining the effects of brand community participation (BCP) on emotional brand attachment (EB). Model 1 shows that BCP is positively and significantly associated with EB (β = .108, SE = .027, p < .01), providing support for H1.

Table 4.

Results of Hypotheses Testing.

BCP	Sample			Small			Network
DV: EB	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
BCP	0.108*** (0.027)	0.191 (0.246)	−0.015 (0.110)	0.060*** (0.022)	−0.102 (0.098)	0.389*** (0.096)	−0.023 (0.023)	0.221** (0.106)	−0.178** (0.073)
ASC	0.193*** (0.016)	0.191*** (0.016)	0.178*** (0.017)	0.189*** (0.017)	0.192*** (0.017)	0.179*** (0.017)	0.196*** (0.016)	0.196*** (0.016)	0.184*** (0.016)
IDS	0.225*** (0.017)	0.223*** (0.018)	0.211*** (0.018)	0.220*** (0.018)	0.220*** (0.018)	0.206*** (0.018)	0.224*** (0.018)	0.222*** (0.018)	0.211*** (0.018)
PI	0.382*** (0.023)	0.385*** (0.024)	0.362*** (0.023)	0.388*** (0.023)	0.384*** (0.024)	0.358*** (0.024)	0.386*** (0.023)	0.385*** (0.024)	0.355*** (0.024)
PSC	0.192*** (0.024)	0.190*** (0.024)	0.126*** (0.026)	0.179*** (0.025)	0.178*** (0.025)	0.107*** (0.026)	0.184*** (0.025)	0.184*** (0.025)	0.114*** (0.026)
RF_PM		0.015 (0.067)			−0.033 (0.024)			0.000 (0.013)
BCP × RF_PM		−0.024 (0.068)			0.046* (0.028)			−0.063** (0.029)
RF_PV			0.072* (0.042)			0.153*** (0.027)			0.091*** (0.025)
BCP × RF_PV			0.040 (0.040)			−0.106*** (0.029)			0.060** (0.025)
Constant	−0.041 (0.040)	−0.089 (0.240)	0.090 (0.108)	0.068* (0.037)	0.204* (0.109)	0.022 (0.039)	0.071 (0.047)	0.081 (0.064)	0.189*** (0.069)
N	1,773	1,773	1,773	1,773	1,773	1,773	1,773	1,773	1,773
Adj R2	0.864	0.864	0.867	0.864	0.864	0.868	0.863	0.864	0.867

Note. 1. Robust standard errors in parentheses.

2. Model 1 verifies H1; Model 2 verifies H2; Model 3 verifies H3; Model 4 verifies H1a; Model 5 verifies H2a ; Model 6 verifies H3a; Model 7 verifies H1b; Model 8 verifies H2b ; Model 9 verifies H3b.

3. DV: Dependent Variable.

4. Diagnostic analyses of regression are reported in Appendix 2.

*

, **, and *** denote significance at the 90%, 95%, and 99% confidence levels, respectively

When the analysis is disaggregated by community type, distinct patterns emerge. In small-group communities, participation has a significant positive effect on EB (Model 4: β = .060, SE = 0.022, p < .01), supporting H1a. In contrast, network-based participation does not exhibit a significant direct association with EB (Model 7: β = −.023, SE = 0.023, n.s.), providing no support for H1b.

A comparison of coefficients across Models 4 and 7 indicates that the effect of participation is stronger in small-group communities than in network-based communities, consistent with H4.

Moderating Effects of Promotion Focus

We next examine whether promotion focus moderates the relationship between brand community participation (BCP) and emotional brand attachment (EB). At the aggregate level, the interaction between BCP and promotion focus is not statistically significant (Model 2: β = −.024, SE = 0.068, n.s.), indicating that promotion focus does not exert a uniform moderating effect across the full sample. Thus, H2 is not supported at the aggregate level.

When the analysis is disaggregated by community type, a differentiated moderating pattern emerges. In small-group communities, the interaction between participation and promotion focus is positive and statistically significant (Model 5: β = .046, SE = 0.028, p < .10), providing support for H2a. In contrast, in network-based communities, the interaction between participation and promotion focus is negative and statistically significant (Model 8: β = −.063, SE = 0.029, p < .05), supporting H2b.

Moderating Effects of Prevention Focus

We further examine whether prevention focus moderates the relationship between brand community participation (BCP) and emotional brand attachment (EB). At the aggregate level, the interaction between BCP and prevention focus is not statistically significant (Model 3: β = .040, SE = 0.040, n.s.), indicating that prevention focus does not exert a uniform moderating effect across the full sample. Thus, H3 is not supported at the aggregate level.

Subgroup analyses reveal a differentiated pattern of moderation across community types. In small-group communities, the interaction between participation and prevention focus is negative and statistically significant (Model 6: β = −.106, SE = 0.029, p < .01), providing support for H3a. In contrast, in network-based communities, the interaction between participation and prevention focus is positive and statistically significant (Model 9: β = .060, SE = 0.025, p < .05), supporting H3b.

Summary of Hypothesis Testing

The results of hypothesis testing are summarized as follows. H1 is supported, indicating that brand community participation (BCP) is positively associated with emotional brand attachment (EB). When community types are distinguished, H1a is supported, showing that small-group participation positively predicts EB, whereas H1b is not supported, as network-based participation exhibits no significant direct effect. Consistent with these findings, H4 is supported, indicating that the effect of participation on EB is stronger in small-group communities than in network-based communities.

Regarding promotion focus, the interaction with small-group participation is positive and statistically significant, providing support for H2a. The interaction between network-based participation and promotion focus is negative and statistically significant, supporting H2b. However, no significant moderating effect of promotion focus is observed at the aggregate level; therefore, H2 is not supported. For the prevention focus, no significant moderating effect is detected in the full-sample analysis, and thus H3 is not supported. Subgroup analyses reveal opposite moderating patterns across community types: prevention focus weakens the effect of small-group participation on EB (supporting H3a) and strengthens the effect of network-based participation on EB (supporting H3b).

In addition, we compare the effect of BCP on EB across small-group and network-based participants, using the moderators’ binary classification. By taking the first derivative of EB with respect to BCP in Equations 5 and 6, the simple slopes can be expressed as:

$$\frac{\partial \mathrm{E}{\mathrm{B}}_{\mathrm{i}}}{\partial \mathrm{BC}{\mathrm{P}}_{\mathrm{i}}}={\beta }_1+{\beta }_3(\mathrm{RF}\_\mathrm{P}{\mathrm{M}}_{\mathrm{i}})$$ (7)

$$\frac{\partial \mathrm{E}{\mathrm{B}}_{\mathrm{i}}}{\partial \mathrm{BC}{\mathrm{P}}_{\mathrm{i}}}={\gamma }_1+{\gamma }_3(\mathrm{RF}\_\mathrm{P}{\mathrm{V}}_{\mathrm{i}})$$ (8)

For Equation 7, the moderator $\mathrm{RF}\_\mathrm{P}{\mathrm{M}}_{\mathrm{i}}$ is evaluated at its high and low levels, defined as one standard deviation above and below its mean (High = 3.934; Low = 2.178). For Equation 8, the corresponding high and low levels of $\mathrm{RF}\_\mathrm{P}{\mathrm{V}}_{\mathrm{i}}$ are 4.172 and 2.579, respectively. The resulting simple-slope plots are shown in Figure 3.

Figure 3.

Slopes of BCP on EB for small-group and network-based communities.

The simple slope plots illustrate how the effect of BCP on EB varies across small-group and network-based communities when the moderators are evaluated at their high and low levels. When BCP increases from 0 to 1, the resulting slopes differ across community types and moderator values. The patterns shown in Figure 3 align with H2a, H2b, H3a, and H3b, indicating positive, negative, negative, and positive moderating effects, respectively.

Robustness Test and Heterogeneity Analysis

To assess the robustness of the main findings, three sets of additional analyses were conducted. First, all models were re-estimated after excluding respondents who simultaneously participated in both small-group and network-based communities. This procedure removed five overlapping cases, resulting in an effective sample of 1,768 observations, including 1,247 small-group members, 402 network-based members, and 119 non-members.

Second, the baseline Ordinary Least Squares (OLS) models were re-estimated using median (quantile) regression, with emotional brand attachment (EB) measured at the median. Bootstrap standard errors based on 500 replications were employed to ensure robust inference.

Third, stricter thresholds were applied to key variable constructions. For the dependent variable, the ninth emotional brand attachment item (EB9) was excluded, and the composite scale was recalculated. For community membership, small-group and network-based participation were coded as one only when respondents selected the highest category (5) on the corresponding 5-point scale, with all other responses coded as non-members. Under these stricter criteria, the effective sample consisted of 1,773 observations, including 550 small-group members, 150 network-based members, and 1,073 non-members. Across all three robustness checks, the coefficient estimates remain consistent with the baseline results. Detailed results are reported in Table 5.

Table 5.

Results of Robustness Test.

	Baseline OLS	Robustness 1: Excluding overlapping members	Robustness 2: Median regression	Robustness 3: Alternative thresholds
				Adjusted DV (EB)	Adjusted group	Adjusted DV + group
H1	0.108*** (0.027)	0.108*** (0.027)	0.082*** (0.025)	0.130*** (0.029)	0.034** (0.016)	0.040** (0.017)
H2	−0.024 (0.068)	−0.020 (0.068)	0.012 (0.092)	−0.061 (0.071)	0.034* (0.018)	0.022 (0.019)
H3	0.040 (0.040)	0.042 (0.040)	0.071 (0.046)	0.067 (0.045)	0.019 (0.017)	0.031* (0.019)
H1a	0.060*** (0.022)	0.061*** (0.022)	0.062** (0.026)	0.116*** (0.024)	0.053*** (0.019)	0.088*** (0.021)
H2a	0.046* (0.028)	0.050* (0.028)	0.086** (0.038)	0.033 (0.030)	0.089*** (0.024)	0.089*** (0.026)
H3a	−0.106*** (0.029)	−0.101*** (0.029)	−0.128*** (0.030)	−0.112*** (0.032)	−0.202*** (0.055)	−0.227*** (0.058)
H1b	−0.023 (0.023)	−0.016 (0.023)	−0.032 (0.031)	−0.074*** (0.025)	−0.022 (0.031)	−0.078** (0.033)
H2b	−0.063** (0.029)	−0.057** (0.029)	−0.088** (0.045)	−0.050 (0.031)	−0.098** (0.047)	−0.093* (0.049)
H3b	0.060** (0.025)	0.067*** (0.025)	0.084*** (0.025)	0.055** (0.026)	0.045* (0.024)	0.031 (0.026)

Note. 1. Robust standard errors are reported in parentheses.

2. In Robustness 1, we exclude the five respondents who simultaneously belong to both the small-group and network-based communities. The resulting effective sample consists of 1,768 observations (1,247 small-group members, 402 network-based members, and 119 non-members).

3. Robustness 2 re-estimates the models using median (quantile) regression with bootstrap standard errors based on 300 replications to ensure robust inference, following Koenker and Hallock (2001).

4. In Robustness 3, the adjusted DV is the EB scale after dropping the ninth item with the lowest standardized factor loading. The adjusted group thresholds code Small-group and Network-based as 1 only when respondents choose 5 on the 5-point scale. Under these stricter thresholds, the effective sample for the GROUP-adjusted models is 1,773 observations (550 small-group members, 150 network-based members, and 1,073 non-members).

*

, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

Following the first two robustness checks, excluding overlapping group members and using median regression, the estimated coefficients remain highly consistent with the baseline results in both sign and statistical significance. Under the stricter variable thresholds applied in Robustness Check 3, a limited number of hypotheses exhibit changes in significance levels; however, no systematic reversals in coefficient signs are observed. Overall, the main results remain stable across alternative sample definitions, estimation methods, and threshold specifications, supporting the robustness of the baseline findings.

In addition, a heterogeneity analysis was conducted based on the baseline regression models. The sample was partitioned by product type, gender, age, monthly income, and region to examine potential subgroup differences. As reported in Table 6, the estimated effects are broadly consistent across most subgroup classifications, with no systematic changes in coefficient direction. Notable heterogeneity is observed primarily across product types, where the same hypothesis yields both positive and negative significant coefficients for different product categories. In contrast, subgroup analyses by gender, age, income, and region show variations mainly in statistical significance rather than in coefficient direction, suggesting relatively limited heterogeneity across these characteristics.

Table 6.

Results of Heterogeneity Analysis.

	Product type				Gender		Age						Monthly income						Regions
	Pdt_1	Pdt_2	Pdt_3	Pdt_4	Male	Female	Age _1	Age _2	Age _3	Age _4	Age _5	Age _6	Inc_1	Inc_2	Inc_3	Inc_4	Inc_5	Inc_6	CHN	TWN
	(N = 630)	(N = 618)	(N = 465)	(N = 60)	(N = 1,088)	(N = 685)	(N = 405)	(N = 218)	(N = 394)	(N = 549)	(N = 204)	(N = 3)	(N = 395)	(N = 182)	(N = 479)	(N = 411)	(N = 180)	(N = 126)	(N = 1,316)	(N = 457)
H1	0.095** (0.040)	0.088** (0.040)	0.167*** (0.061)	−0.094 (0.090)	0.084** (0.033)	0.131*** (0.041)	0.160** (0.078)	0.186** (0.076)	0.070 (0.047)	0.084** (0.038)	−0.028 (0.106)	0.000 N/A	0.163** (0.067)	0.072 (0.087)	0.099** (0.050)	0.106** (0.053)	0.087 (0.070)	0.018 (0.085)	0.068*** (0.025)	0.159*** (0.056)
H2	−0.156 (0.102)	0.044 (0.094)	0.106 (0.147)	−2.272*** (0.822)	−0.058 (0.082)	−0.018 (0.104)	−0.191 (0.300)	0.268 (0.208)	−0.055 (0.085)	−0.319*** (0.118)	−0.137 (0.185)	0.000 N/A	−0.132 (0.246)	0.746* (0.398)	0.023 (0.214)	−0.064 (0.119)	−0.001 (0.116)	−0.056 (0.125)	−0.076 (0.087)	−0.102 (0.090)
H3	0.070 (0.050)	−0.011 (0.057)	0.086 (0.085)	−0.220 (0.349)	−0.008 (0.064)	0.056 (0.054)	−0.006 (0.100)	0.310*** (0.097)	−0.013 (0.052)	0.253*** (0.089)	−0.158 (0.114)	0.000 N/A	0.042 (0.099)	0.375** (0.149)	0.185** (0.091)	−0.040 (0.097)	0.084 (0.087)	−0.140 (0.144)	0.412*** (0.099)	−0.090 (0.096)
H1a	0.100*** (0.035)	0.025 (0.039)	0.079* (0.041)	0.157 (0.198)	0.065** (0.028)	0.054 (0.035)	0.049 (0.056)	0.004 (0.056)	0.067* (0.037)	0.135*** (0.049)	0.088 (0.076)	−0.967 N/A	0.103** (0.051)	−0.041 (0.065)	0.012 (0.055)	0.097** (0.043)	0.068 (0.056)	0.049 (0.068)	0.302*** (0.028)	0.017 (0.037)
H2a	−0.040 (0.046)	0.074 (0.047)	0.080 (0.051)	0.360*** (0.115)	0.071** (0.033)	0.009 (0.046)	−0.023 (0.060)	0.154* (0.090)	0.020 (0.053)	0.061 (0.048)	−0.005 (0.080)	−0.242 N/A	0.009 (0.055)	0.122* (0.064)	0.098 (0.062)	−0.001 (0.055)	0.053 (0.067)	−0.038 (0.107)	0.061* (0.034)	−0.031 (0.077)
H3a	−0.132*** (0.047)	−0.243*** (0.049)	0.053 (0.052)	−0.152 (0.110)	−0.105*** (0.036)	−0.110** (0.047)	−0.170*** (0.057)	−0.050 (0.090)	−0.040 (0.063)	−0.110** (0.052)	−0.195** (0.085)	−0.242 N/A	−0.148*** (0.054)	−0.069 (0.078)	−0.207*** (0.059)	−0.152** (0.061)	0.071 (0.091)	−0.042 (0.117)	−0.142*** (0.028)	−0.031 (0.075)
H1b	−0.039 (0.034)	−0.003 (0.041)	−0.052 (0.041)	−0.305* (0.181)	−0.043 (0.028)	0.007 (0.037)	0.019 (0.058)	0.079 (0.057)	−0.054 (0.038)	−0.088* (0.046)	−0.124* (0.074)	0.967 N/A	−0.020 (0.050)	0.085 (0.052)	0.030 (0.055)	−0.048 (0.041)	−0.043 (0.058)	−0.074 (0.069)	−0.194*** (0.027)	0.031 (0.039)
H2b	0.013 (0.044)	−0.088* (0.050)	−0.091* (0.054)	−0.272*** (0.089)	−0.091*** (0.035)	−0.022 (0.049)	0.012 (0.056)	−0.209* (0.109)	−0.064 (0.058)	−0.056 (0.046)	−0.031 (0.082)	0.249 N/A	−0.024 (0.052)	−0.127* (0.065)	−0.091 (0.063)	−0.013 (0.055)	−0.114 (0.075)	0.032 (0.120)	0.014 (0.032)	−0.068 (0.077)
H3b	0.103*** (0.035)	0.152*** (0.045)	−0.077* (0.044)	0.188* (0.096)	0.041 (0.030)	0.084** (0.041)	0.131** (0.057)	0.091 (0.067)	−0.015 (0.047)	0.051 (0.043)	0.069 (0.087)	0.251 N/A	0.115** (0.047)	0.138*** (0.052)	0.179*** (0.052)	0.056 (0.052)	−0.062 (0.060)	−0.013 (0.120)	0.081*** (0.027)	−0.049 (0.079)

Note. 1. Robust standard errors in parentheses.

2. N/A = Variance cannot be computed due to insufficient sample size (N = 3).

*

, **, and *** denote significance at the 90%, 95%, and 99% confidence levels, respectively.

Discussion

This study examines how structurally distinct brand communities shape emotional brand attachment and how regulatory focus conditions these effects (Muniz & O’Guinn, 2001; Schau et al., 2009). By jointly considering community structure and motivational orientation, the findings offer a more differentiated understanding of when and why brand community participation translates into emotional attachment.

Theoretical Implications

First, this study contributes to brand community research by demonstrating that community structure fundamentally conditions emotional outcomes rather than merely amplifying them. Although prior studies generally report positive associations between community participation and relational outcomes such as loyalty or attachment (Algesheimer et al., 2005; Brodie et al., 2013), they rarely distinguish between structurally different community forms. Our findings show that emotional attachment emerges primarily in contexts characterized by repeated interaction, relational stability, and mutual recognition—conditions more typical of small-group communities (Dholakia et al., 2004). Beyond social identity processes, these structural features also enable complementary affective mechanisms. Frequent interaction and stable membership facilitate emotional contagion and perceived authenticity, which prior research identifies as key drivers of emotional engagement in brand communities (Schau et al., 2009; Wirtz et al., 2013). These mechanisms are offered as theoretically grounded explanations rather than formally modeled mediators.

Second, the results extend construal-level theory (CLT) by embedding psychological distance within brand community environments. CLT posits that reduced psychological distance leads to concrete, low-level construals that intensify affective responses (Liberman & Trope, 1998; Liberman et al., 2007). Consistent with this logic, small-group communities reduce social distance through dense interpersonal ties and thereby foster emotionally rich brand experiences, whereas network-based communities maintain greater psychological distance that constrains affective attachment. By situating CLT within digitally mediated brand communities, this study advances prior CLT applications that have focused primarily on temporal or spatial distance (Williams & Bargh, 2008).

Third, this study advances regulatory focus theory by showing that promotion and prevention orientations operate as context-dependent moderators rather than universal drivers of emotional outcomes. Promotion focus emphasizes aspirational motives and ideal-self expression (Aaker & Lee, 2001; Higgins, 1998), yet our findings indicate that its positive effects emerge mainly in small-group settings. In contrast, promotion focus weakens the effect of network-based participation on emotional attachment, suggesting a misalignment between aspirational motives and the low-commitment, information-oriented nature of network-based communities. This pattern responds directly to calls for more nuanced applications of regulatory focus in consumer research (Chernev, 2004; Malär et al., 2011).

Fourth, the moderating effects of prevention focus highlight the role of self-verification mechanisms in shaping emotional attachment. The positive moderation observed in network-based communities suggests that these environments may facilitate self-verification through informational clarity and reduced interpersonal demands, consistent with prior work on actual-self congruence (Higgins et al., 1997; Malär et al., 2011; Sirgy et al., 1997). Conversely, the negative moderation in small-group communities indicates that dense relational expectations may heighten vigilance and dampen emotional attachment for prevention-focused individuals.

Taken together, these findings reposition brand communities as psychological infrastructures that activate distinct motivational and self-regulatory processes, rather than as uniformly beneficial engagement platforms. The results should also be interpreted in light of their cultural context. Because the sample draws from China and the Taiwan region, contexts characterized by stronger relational orientation and sensitivity to social context, the emotional advantages of small-group communities may be particularly salient. In such settings, repeated interaction, interpersonal familiarity, and relational stability are culturally congruent mechanisms for fostering emotional attachment. By contrast, network-based communities, defined by looser ties and lower relational obligation, may function differently in more individualistic or low-context cultures. Accordingly, the effects of community structure and regulatory focus identified in this study should be understood as context-dependent rather than universal, highlighting cultural orientation as an important boundary condition for future research.

Managerial Implications

From a managerial perspective, the findings suggest that firms should strategically align community design with their intended relational outcomes. Small-group communities, such as closed chat groups, brand ambassador programs, or invitation-based fan groups, are particularly effective in fostering emotional brand attachment. These formats enable repeated interaction, interpersonal familiarity, and mutual recognition, thereby strengthening affective bonds between consumers and the brand, consistent with prior research on interaction intensity and relational bonding (Algesheimer et al., 2005; Schau et al., 2009).

In contrast, network-based communities serve complementary rather than substitutive functions. The results indicate that network-based communities are less effective for promotion-focused consumers but more suitable for prevention-focused consumers, who tend to value lower relational risk, informational clarity, and reduced social obligation. Managers can therefore segment community strategies based on consumers’ motivational orientations, allocating emotionally intensive formats to consumers seeking relational depth while offering information-oriented platforms to those who prefer cautious or exploratory engagement.

More broadly, the findings caution against treating brand communities as one-size-fits-all engagement tools. Instead, managers may benefit from designing layered community architectures, in which small-group formats function as relational cores that cultivate emotional attachment and long-term commitment, while network-based platforms operate as accessible entry points for information exchange and risk-averse consumers (Schau et al., 2009; Wirtz et al., 2013). Such differentiated designs allow firms to optimize community investments by matching community structures to heterogeneous consumer motivations.

Limitations and Future Research

Several limitations of this study should be acknowledged. First, the cross-sectional research design limits causal inference. Future research could adopt longitudinal designs to examine how emotional brand attachment develops over time as consumers enter, exit, or transition between different types of brand communities (Brodie et al., 2013). In addition, experimental or quasi-experimental approaches could be employed to manipulate perceived community structure, thereby allowing for stronger causal tests of the psychological mechanisms proposed in this study (Aaker & Lee, 2001).

Second, although the sample includes respondents from China and the Taiwan region, cultural moderators were not explicitly modeled. Consequently, the generalizability of the findings to other cultural contexts should be interpreted with caution. Prior research suggests that cultural dimensions such as collectivism and uncertainty avoidance may influence both community engagement and regulatory focus (Hofstede, 2001), highlighting an important direction for future cross-cultural investigations.

Conclusion

This study examines how brand community participation relates to emotional brand attachment by distinguishing between small-group and network-based brand communities and considering consumers’ regulatory focus orientations. The findings demonstrate that emotional brand attachment does not emerge uniformly across community types. Rather, attachment is more strongly associated with participation in small-group communities characterized by stable and repeated interpersonal interactions, whereas network-based communities exhibit weaker direct emotional effects. Moreover, promotion and prevention orientations condition these relationships in distinct ways, underscoring the importance of motivational fit between individuals and community structures.

Overall, this study contributes to a more nuanced understanding of brand communities by highlighting community structure and individual motivation as key boundary conditions in the formation of emotional brand attachment. By integrating structural and psychological perspectives, the research clarifies why some community environments are more conducive to emotional bonding than others and provides a refined empirical foundation for future studies on consumer–brand relationships in digitally mediated communities.