Do parasocial relationships with micro- and mainstream celebrities differ? An empirical study testing four attributes of the parasocial relationship

ABSTRACT The aim of this study was to examine to what extent parasocial relationships with micro-celebrities and mainstream celebrities differ in terms of perceived reciprocity, authenticity, intimacy, and parasocial interaction. A within-person comparison of survey responses collected among 402 college students in China showed that participants perceived their relationships with micro-celebrities as more reciprocal than their relationships with mainstream celebrities. However, contrary to our expectations, participants perceived a lower degree of authenticity and intimacy for micro-celebrities than for mainstream celebrities. Moreover, we found no significant difference between the parasocial interactions with both types of celebrities. We additionally investigated the relative contribution of perceived reciprocity, authenticity, and intimacy to the parasocial interaction experienced with both micro- and mainstream celebrities. To that end, we performed multiple linear regression analyses run separately for each type of celebrity. We found that in the regression models for both micro- and mainstream celebrities, perceptions of reciprocity and intimacy positively predicted parasocial interaction, suggesting that these attributes are especially important in fostering parasocial relationships. Interestingly, perceptions of authenticity did not predict greater parasocial interaction with mainstream nor micro-celebrities, which suggests that the role of authenticity in fostering parasocial interactions may be different from what is commonly assumed. The implications of this study shed light on the changing nature of parasocial relationships in the context of micro-celebrities operating on social media.


Introduction
carried out by fans. In other words, the relationship is considered para-social because fans tend to be aware of the activities, anecdotes, and even personalities of their favourite celebrities, whereas celebrities barely know anything about their fans.
The concept of the parasocial relationship was developed in the 1960's, in the context of audiences' reactions to traditional mainstream celebrities who appeared in movies and on television shows (Horton and Wohl 1956, Rubin et al. 1985, Rubin and McHugh 1987. Mainstream celebrities in the mass media era typically ascended to a status of fame by receiving a lot of airtime and thus visibility in the 'one-to-many' broadcasts of mass media like film and TV. With the advent of social media, however, online micro-celebrities, such as famous beauty bloggers or vloggers, have emerged as a new type of celebrity. Because these micro-celebrities are mainly active on social media platforms, interactive communication between audiences and micro-celebrities is possible. This raises questions as to how the parasocial relationship with this new type of celebrity might evolve (Aguiar et al. 2019, Kumar and Benbasat 2001, Baek et al. 2013, Kim and Song 2016, Ferchaud et al. 2018, Kurtin et al. 2018. In our smartphones, micro-celebrities are never too far away; instead, they actually 'rest in our pockets and sleep in our beds' (Oliver et al. 2014, p. 152). The interaction between fans and micro-celebrities occurs through subscriptions, likes, shares, comments, and of course, feedback. In this vein, the relationship between followers and micro-celebrities has evolved; it is no longer strictly limited to a one-way conversation (Morgan 2017) but is, in fact, mutual and interactive.
Over the past decades, various studies have explored the nature of one-sided parasocial relationships in the context of following mainstream celebrities (e.g., Rubin et al. 1985, Ryff 1989, Ashe and McCutcheon 2001. To our knowledge, however, few studies have investigated differences in the parasocial relationship that people build with microand mainstream celebrities on social media. The first aim of this study is to examine if the parasocial relationship with micro-celebrities is perceived as stronger than with mainstream celebrities. To that end, we explore whether attributes of the parasocial relationship that have previously been identified for mainstream celebrities are established more strongly in micro-celebrities than in mainstream celebrities. Based on previous literature (Horton and Wohl 1956, Rubin et al. 1985, Rubin and Perse 1987, Rubin and McHugh 1987, we have extracted four attributes of the parasocial relationship: parasocial interaction, reciprocity, authenticity, and intimacy. The first research question guiding this study tests if these four attributes are perceived as stronger for micro-celebrities than for mainstream celebrities: RQ1: Do people perceive greater parasocial interaction, reciprocity, authenticity, and intimacy in their relationship with micro-celebrities than in their relationship with mainstream celebrities? The second aim is to explore the importance of reciprocity, authenticity, and intimacy in fostering parasocial interaction with micro-and mainstream celebrities on social media. Earlier studies have asserted that reciprocity does not play an essential role in fostering parasocial relationships with mainstream celebrities, while authenticity and intimacy do (e.g., Horton andWohl 1956, Rubin et al. 1985). Some recent studies have argued, however, that reciprocity does enhance parasocial relationships (e.g., Baek et al. 2013, Chen 2014, while authenticity might not be prioritised (e.g., Ferchaud et al. 2018). In short, the mixed evidence indicates that there remains a paucity of evidence on the critical factors that significantly predict parasocial relationships on social media, and whether these factors work differently for micro-and mainstream celebrities. Hence, a second research question that this study aims to answer is whether perceived reciprocity, authenticity, and intimacy predict parasocial interactions with micro-and mainstream celebrities on social media platforms. We examine parasocial interaction as the outcome measure, as a longer-term parasocial relationship might emerge after repeated experiences of parasocial interaction (Horton andWohl 1956, Perse andRubin 1989): RQ2: Do perceived reciprocity, authenticity, and intimacy predict experiences of parasocial interaction with celebrities on social media platforms, and do these factors work differently for micro-and mainstream celebrities?
To address the above research questions, we draw from cross-sectional survey data gathered from a sample of 402 Chinese students. Of these students, 160 individuals indicated following both micro-and mainstream celebrities, enabling us to make a withinperson comparison of students' perceptions of reciprocity, authenticity, intimacy and parasocial interaction for micro-celebrities and mainstream celebrities (RQ1). Multiple linear regression analyses were used to test whether reciprocity, authenticity, and intimacy predict experiences of parasocial interaction among followers of micro-celebrities (N = 200) and mainstream celebrities (N = 353).

The parasocial relationship and parasocial interaction
The concept of the parasocial relationship is well established in the field of media and communication studies. Initially, Horton and Wohl (1956) suggested that a parasocial relationship exits when there is an illusory relationship, seemingly close to a friendly or intimate relationship, between a media user and a media performer. This relationship is restricted to media users following their favourite media performers through viewing their content. Therefore, the parasocial relationship is regarded as a faux, one-sided relationship that is directed from the audience (i.e., media user) to the media performer, but not mutually developed (Horton and Wohl 1956, Rubin et al. 1985, Giles 2002, Bond 2016. The term parasocial interaction refers to the faux conversational behaviour from the audience towards a media performer (Horton andWohl 1956, Giles 2002). This interaction can occur in reaction to the media figure's performance, gesture, or voice (Horton and Wohl 1956). For example, via techniques such as close-up shots TV characters can address the audience and convey their emotions to them (Horton andWohl 1956, Rubin et al. 1985). Audiences may engage themselves through this close-up lens, facing the actor closely, and interacting with the performer intimately, despite being aware that it is only an imaginary scenario. To summarise, parasocial interaction can be regarded as a media audience's reaction to a media performer in such a way that the audience experiences having an intimate relationship with this performer.
It is important to distinguish the concept of parasocial interaction from that of the parasocial relationship, as they are often conflated. Parasocial interaction usually starts during the viewing experience (Horton and Wohl 1956, Rosengren and Windahl 1972, Hedinsson 1981. It concerns a one-way interpersonal involvement initiated by the viewers and directed to the celebrities on screen (Rubin et al. 1985, Perse andRubin 1989). This involvement can extend itself beyond the viewing experience when individuals seek guidance and make friends with the performers they like, start mimicking their behaviours, or even attempt to meet them in person. When this occurs, we can speak of a parasocial relationship. Thus, while parasocial interaction refers to an immediate sense of mutual awareness with a celebrity (i.e., media performer), the parasocial relationship is a long-term imaginary association (Dibble et al. 2016) between audiences and media performers. Parasocial interaction, as an essential attribute, contributes to the formation and strength of the parasocial relationship Rubin 1989, Ballantine andMartin 2005).

The mainstream celebrity and micro-celebrity
When a media performer is popular to a wide range of fans and is able to use certain public-related resources to make their achievements seen, we may say that this performer qualifies as a celebrity (Marshall 1997, Turner 2013, Hearn and Schoenhoff 2016. The fame of mainstream celebrities is built upon both the representation of their personalities at a cultural level and them becoming a commercial interest for media, advertisement, and entertainment industries. In most cases, celebrities are involved in a well-resourced industry in terms of 'distinct self-referential, attention-seeking and market-aware practices' (Hearn and Schoenhoff 2016, p. 196). These celebrities, such as Beyoncé or Yao Ming, who have been granted strong public identities outside of social media because of certain talents and/or efficient resources for self-promotion, are commonly referred to as mainstream celebrities (Senft 2008, Marwick and Boyd 2011, Hearn and Schoenhoff 2016, Khamis et al. 2017. The advent of social media, however, has given birth to another type of celebrity, the micro-celebrity (Senft 2008, Gamson 2011, Marwick and Boyd 2011, Turner 2013, Van Dijck 2013, Hearn and Schoenhoff 2016, Jerslev 2016, Khamis et al. 2017, Harvey 2018. A microcelebrity refers to an internet user whose fame is derived from the strategic maintenance of an interactive relationship with a group of fans on a social media platform (Senft 2008, Gamson 2011, Marwick and Boyd 2010. Micro-celebrities use social media for online selfpresentation (Ellison et al. 2006, Colapinto and Benecchi 2014, Khamis et al. 2017) and for direct interaction with their fans (Hearn and Schoenhoff 2016). They are self-motivated celebrities, who adopt unique forms of exhibition, bypass the strict control of a middleman, and interact intimately with fans, in order to become famous as a blogger, social media content producer, online gamer, etcetera (Gamson 2011, Hearn andSchoenhoff 2016).
Previous studies have already compared mainstream and micro-celebrities. Senft (2008) and Marwick and Boyd (2010), Marwick and Boyd (2011), for example, argue that similar to mainstream celebrities, a micro-celebrity may endeavour a kind of organised performance or practice to run a successful social media account. However, the differences are that micro-celebrities gain their fame by implementing a'do-it-yourself'strategy (Gamson 2011(Gamson , p.1065 whereas mainstream celebrities achieve their success through industrial operations (Hearn and Schoenhoff 2016). For this reason, the parasocial relationships with micro-celebrities may differ from mainstream celebrities in a number of key attributes. In this sense, when comparing parasocial relationships with these two types of celebrities, issues often arise not only with respect to their differences in fame, but also in the reciprocity perceived by fans, in the judgement from the fake to the real, and in the extent of the intimate interaction fans experience with celebrities. Therefore, this study examines to what extent parasocial relationships with micro-celebrities and mainstream celebrities differ in their perceived reciprocity, authenticity, intimacy, and parasocial interaction.

Perceived reciprocity in the parasocial relationship
The first research question of this study asks whether individuals who follow both microand mainstream celebrities differ in their perception of attributes contributing to the parasocial relationship. A first factor that contributes to the parasocial relationship and that may differ between micro-and mainstream celebrities is perceived reciprocity. Horton and Wohl (1956) stated that a crucial distinction between regular interpersonal relationships and parasocial relationships lies in the lack of 'effective reciprocity' (p.215). Before the advent of social media platforms -in the era of mass media -no matter how much the audience interacted with the media performer, the relationship that the audience perceived was never more than an illusion because the media performer could not really interact back. In other words, while the responses of celebrities could be arranged or rehearsed in advance of their shows, these responses would never be as genuine as those given in a face-to-face interaction (Kim and Song 2016). Given that audiences could nonetheless perceive a relationship with celebrities, earlier scholars argued that reciprocity might not be an essential attribute for the development of parasocial relationships (Rubin and McHugh 1987).
Social media, however, provide audiences with the possibility to experience a more interactive conversation with celebrities. This is true for both mainstream and microcelebrities that are active on social media. For mainstream celebrities, users usually do not expect responses, and especially not responses that are provided in a near-synchronous manner (Hartmann 2008, Baek et al. 2013. For micro-celebrities, however, users may experience more back-and-forth interaction and even a 'behind-the-scenes peek' into micro-celebrities' everyday lives (Gamson 1994, p.49, Bond 2016. As a result, users may experience stronger parasocial interactions with a micro-celebrity, especially on social media platforms such as YouTube where reciprocity is central to the user experience (Chen 2014). Therefore, the current study hypothesises that: H1a: Followers of micro-and mainstream celebrities on social media perceive a higher degree of reciprocity with micro-celebrities than mainstream celebrities.
The second research question of this study asks which attributes contribute to experiences of parasocial interaction, and whether these attributes work similarly for mainstream and micro-celebrities. Reciprocity was originally considered non-essential in fostering parasocial relationships in the mass media era (Horton andWohl 1956, Rubin andMcHugh 1987). However, on social media, it supposedly contributes to establishing a relationship between fans and both micro-and mainstream celebrities. This means that the higher the degree of reciprocity that followers perceive from celebrities on social media, the more they might develop a sense of celebrity-follower closeness. In this sense, for both mainstream and micro-celebrities, it may be easier to trigger audiences' parasocial interactions on social media (Baek et al. 2013, Kurtin et al. 2018. Accordingly, the current study posits that: H1b: Both for mainstream and micro-celebrities, perceived reciprocity positively predicts parasocial interaction.

Perceived authenticity in the parasocial relationship
A second attribute that may be perceived differently for micro-and mainstream celebrities (RQ1), yet that may contribute to the development of parasocial relationships (RQ2) is perceived authenticity. Rubin et al. (1985) stated that 'perceived realism' (p.174) is one of the main factors that afford an experience of parasocial interaction: The more authentic that celebrities are perceived, the more willing their fans are to attempt interpersonal interactions with them (Rubin and Perse 1987, Geraghty 1991, Cohen 2001, Giles 2002. On social media, micro-celebrities are considered to be more authentic than mainstream celebrities (Senft 2008, Marwick Alice 2010, Reinecke and Trepte 2014, Ferchaud et al. 2018. Two reasons explain this difference. First, microcelebrities, who engage in direct interactions with fans, are perceived as more genuine and approachable (Creswell 2008, Kim and Song 2016, Ferchaud et al. 2018. Second, the activities of mainstream celebrities are usually controlled by a commoditized third party (e.g., agent companies; Gamson 2011, Hearn andSchoenhoff 2016) or institutional middleman (Marwick Alice 2010). The image that mainstream celebrities build on social media may therefore be constructed by a third party, and not exactly represent who they really are. This phenomenon is considered to reduce the perceived authenticity of mainstream celebrities. Therefore, microcelebrities are supposed to be more authentic than mainstream celebrities. The current study hypothesises that: H2a: Followers of micro-and mainstream celebrities on social media perceive microcelebrities as more authentic than mainstream celebrities.
In relation to our second research question, whether perceived authenticity play a role in fostering parasocial interaction in social media context, extant research suggests it does. Fans who perceive the social media posts of their favourite celebrities as more authentic experience stronger feelings of connectedness (Bond 2016). This is especially the case for celebrities active on social media platforms such as Twitter (Bond 2016, Cohen andTyler 2016) on which posts often seem as 'official statements' (Bond 2016, p. 659), and YouTube (Kurtin et al. 2018, Ferchaud et al. 2018) on which audiences feel that media performers are vivid, approachable, and down-to-earth (Cohen 2001, Tsay-Vogel andSchwartz 2014). Cohen and Tyler (2016) further emphasised the importance of nonrobotic and genuine communication, which facilitates parasocial interactions on social media. Overall, the available evidence indicates that perceptions of authenticity play an important role in fostering and maintaining parasocial interaction with both mainstream and micro-celebrities. Hence, the following hypothesis is proposed: H2b: Both for mainstream and micro-celebrities, perceived authenticity positively predicts parasocial interaction.

Perceived intimacy in the parasocial relationship
A third attribute that may matter, is perceived intimacy. In the context of the parasocial relationship, perceived intimacy covers all instances where a media user displays a liking for a media performer (Giles 2002). In parasocial relationships with mainstream celebrities, intimacy can develop when audiences experience intimate emotions in relation to celebrities when viewing their shows (Horton and Wohl 1956). These intimate emotional experiences are comparable to how audiences feel when having intimate interactions with their flesh-and-blood friends (Perse and Rubin 1989). This 'one-sided' intimacy can be enhanced by celebrities through the repeated performance of their personalities, by merging their personalities with their on-stage roles, by addressing audiences directly from the screen, and through the familiar physical set-up of the show in which celebrities appear (Horton and Wohl 1956).
With micro-celebrities, intimacy may develop differently, as they often engage in intimate communicative strategies online (Abidin 2015;Jerslev, 2016, Khamis et al. 2017. When moving online, mainstream celebrities may also adopt these microcelebrity practices and communicate intimately with followers by posting personal pictures and videos, addressing rumours, and sharing personal information (Marwick and Boyd 2011). Nonetheless, although micro-celebrities and mainstream celebrities can both perform intimacy on social media, the degree of intimacy that audiences develop with them may still be more limited, as the mechanisms through which intimacy develops remain different. One such mechanism is similarity. Qualitative studies show that microcelebrities' personae are built upon a sense of everydayness (Jerslev, 2016), whereas mainstream celebrities embrace a meritocracy discourse, emphasising achievement, success and glamour (Gamson, 2011;Hou 2019). This suggests that followers may find more commonalities between their life and that of micro-celebrities than that of mainstream celebrities. Research suggests that individuals are attracted to online media performers that are similar to them in terms of physical appearance, personality, interests, and thoughts. This is because when a media user finds commonalities, this may trigger feelings of resonance and emotional attachment with the celebrity (Rojek, 2015;Lou and Kim 2019). An additional mechanism that may foster intimacy is reciprocity, as already discussed above. Mainstream celebrities' media personae are usually controlled by a celebrity industry (Turner 2013), compared to the relatively higher degree of agency of micro-celebrities, who work with Multi-Chanel Networks (Lobato 2016). This may allow higher degree of direct back-and-forth interactions with micro-celebrities, which helps in fostering greater intimacy with them. Therefore, the hypothesis is proposed: H3a: Followers of micro-and mainstream celebrities on social media perceive greater intimacy in their relationship with micro-celebrities than in their relationship with mainstream celebrities.
As for the contribution of intimacy to experiences of parasocial interaction (RQ2), Horton and Wohl (1956) stated that audiences form close emotional relationships with media performers they like, and that perceptions of intimacy are crucial for parasocial relationship maintenance. In the mass media age, mainstream celebrities in TV shows imitated the gestures and conversational style used in face-to-face interactions when interacting with other cast members and live studio audiences. These intimacy strategies contribute to the casual and easily approachable sense of the celebrity, thus facilitating audiences to engage in a one-sided interaction towards them. Social media platforms provide ample opportunities for such intimate practices; micro-celebrities are even characterised by their performance of intimacy (Abidin 2015;Jerslev, 2016, Khamis et al. 2017. The fundamental goal of these practices is to achieve accelerated intimacy by reducing uncertainty and improving familiarity (Bond 2016). As perceived intimacy is inherently intertwined with parasocial interaction (Horton and Wohl 1956), perceived intimacy with media performers, is the heart of the parasocial relationship (Rojek, 2015). Thus, we posit that: H3b: Both for mainstream and micro-celebrities, perceived intimacy positively predicts parasocial interaction.

Parasocial interaction in the parasocial relationship
Based on the hypotheses above, we expect that followers perceive greater reciprocity, authenticity, and intimacy in their relationship with micro-celebrities than that with mainstream celebrities. With regard to the experienced parasocial interaction with celebrities itself, we may also find a difference. Tsai and Men (2013) argued that social media facilitate a higher level of parasocial interactions between the users and media performers. Further, Chen (2014) proposed that the users may perceive a stronger parasocial interaction with a micro-celebrity (e.g., a famous YouTuber) than with a mainstream celebrity (e.g., a famous TV character). This is because the perceived intimacy is enhanced in terms of a more frequent reciprocal communication with micro-celebrities on social media (Lou and Kim 2019). Additionally, authenticity, as a footstone of the parasocial relationship (Rubin et al. 1985), has been argued to be perceived stronger for microcelebrities than for mainstream celebrities (Reinecke andTrepte 2014, Ferchaud et al. 2018). When all the attributes of parasocial relationships with micro-celebrities are considered to be enhanced, we may expect that: H4: Followers of micro-and mainstream celebrities on social media experience greater parasocial interaction with micro-celebrities than with mainstream celebrities.

Gender differences
In addition to exploring the differences in relationships with micro-and mainstream celebrities, we also investigated whether perceptions of the attributes of parasocial relationships with micro-and mainstream celebrities vary according to gender. Recent research indicates that there may be such differences. A study shows, for example, that teenage boys imagine the celebrities they follow -who are more often athletes -more as authority figures or mentors, while girls regard the celebrities they follow more as their friends (e.g. Gleason, Theran and Newber, 2017). To date, few studies have systematically explored associations between gender and perceptions of the attributes of the parasocial relationship. Hence, in this study, we additionally explore whether such gender differences present themselves (cf. RQ1), and include gender as a control variable when examining whether reciprocity, authenticity and intimacy predict parasocial interaction (cf. RQ2).

Sample and procedure
Chinese college students were recruited to fill out an anonymous online questionnaire in February 2019. Of the respondents (N = 402, 62.9% female), 98.4% were freshmen or sophomores whose ages ranged from 17 to 23 years (M = 19.02, SD = 2.02). To allow for within-person comparison, the survey consisted of similar questions about mainstream celebrities and micro-celebrities. We collected informed consent from the participants before the survey.
At the beginning of the online questionnaire, mainstream celebrities were defined as 'famous people who already have strong public identities and the resources to selfpromote, such as Yao Ming, Beyoncé'. Micro-celebrities were defined as 'online users who become popular through social media performance or practice, such as Sister Furong, Brother Caipeng' (Chinese famous blogger and vlogger). Next, participants were asked to indicate whether they followed mainstream and micro-celebrities on social media platforms. In total, 194 (48.26%) participants reported following only mainstream celebrities, 48 (11.94%) only micro-celebrities, and 160 (39.8%) followed both types of celebrities. The participants were then instructed to name a their favourite celebrity. In case the participant followed both mainstream and micro-celebrities, this question was asked for both types of celebrities. For the remainder of the questionnaire, a logic was set so that the scale items included the name of the nominated celebrities.
To compare micro-and mainstream celebrities in terms of their perceived reciprocity, authenticity, intimacy and parasocial interaction, we performed dependent t-tests on the answers of the subgroup of participants (N = 160) who follow both micro-and mainstream celebrities. To examine whether the degree of perceived reciprocity, authenticity, and intimacy predict the perceived parasocial interaction with mainstream and microcelebrities, we performed multiple linear regression analyses for both the group of participants who follow mainstream celebrities (N = 354) and the group of participants who follow micro-celebrities (N = 208).

Measures Perceived parasocial interaction.
Participants were asked to what extent they experienced parasocial interactions with the celebrities. We used the Experience of Parasocial Interaction Scale (EPSI-Scale: Hartmann and Goldhoorn 2011), in which 6 items were used to measure the agreement ranging from 1 (I do not agree at all) to 5 (I totally agree). The last item was slightly adjusted in order to reflect the friendship that the participants perceived. Exemplary items are '[name of the celebrity] is aware of me,' '[name of the celebrity] knows I am aware of him/her,' '[name of the celebrity] knows that I react to him/ her.' A factor analysis showed that all 6 items loaded on one factor and formed a onedimensional scale. For the items assessing parasocial interaction with mainstream celebrities, the Cronbach alpha for the scale was .951 (M = 2.08, SD = 1.08, the KMO statistic is 0.904, the Bartlett's test of sphericity is 0.000); For the items assessing parasocial interaction with micro-celebrities, the Cronbach alpha was .961 (M = 2.22, SD = 1.06, the KMO statistic is 0.899, Bartlett's test of sphericity is 0.000).
Perceived reciprocity. We measured perceived reciprocity with a self-constructed 10item scale to assess how frequently the celebrities give feedback to their fans and to the participant-in-question on social media. The scale included five personal and five general (i.e., fan base) statements:  A factor analysis of these items showed two eigenvalues greater than 1, together loading 66.88% of the variance. We removed the 5th item (M = 3.87, SD = 0.946) because the component analysis showed that this item did not fit the cluster, and it was not significantly related to the main factor. The remaining 8 items formed a one-dimensional scale. For the items assessing mainstream celebrities, Cronbach's alpha was .905 (M = 3.58, SD = 0.76, the KMO statistic is 0.925, the Bartlett's test of sphericity is 0.000); in the cohort of micro-celebrity fans, Cronbach's alpha was .938 (M = 3.40, SD = 0.87, the KMO statistic is 0.919, the Bartlett's test of sphericity is 0.000).

Perceived intimacy.
We built a 7-item scale based on Descutner and Thelen (1991) Fearof-Intimacy Scale and Miller and Lefcourt (1982) Social Intimacy Scale. We slightly adapted the items to match the context of social media interaction. Examples of the items were: '[name of the celebrity] When I am visiting celebrity's account, I feel s/he is accompanying me,' '[name of the celebrity] I would miss her/him when s/he is offline.' A factor analysis showed that all items loaded on a one-dimensional scale. The response categories ranged from 1 (not true to me at all) to 5 (totally true to me). For the items assessing mainstream celebrities, Cronbach's alpha was .734 (M = 2.61, SD = 0.774, the KMO statistic is 0.898, the Bartlett's test of sphericity is 0.000); For the items assessing micro-celebrities, Cronbach's alpha was .921 (M = 2.42, SD = 0.952, the KMO statistic is 0.875, the Bartlett's test of sphericity is 0.000).

Statistical controls
Demographic data and online celebrity following behaviours were controlled for in the multiple regression analyses. These variables were age, gender, the time spent on and the frequency of receiving feedback from celebrities online. The relevant descriptive statistics appear in Table 1. Table 2 shows the zero-order correlations between the main variables and the control variables. The results show that, both for mainstream and micro-celebrities, perceived parasocial interaction was positively related to perceived reciprocity (r MS = 0.352, p < 0.001; r MC = 0.392, p < 0.001). Perceived intimacy was positively related to feedback frequency (r MS = 0.285, r MC = 0.245), reciprocity (r MS = 0.430, r MC = 0.643), parasocial interaction (r MS = 0.375, r MC = 0.447) and authenticity (r MS = 0.373, r MC = 0.265), all ps < 0.001.

Correlations and control variables
Significant differences in gender were observed for the time spent, perceived reciprocity, perceived parasocial interaction, and perceived intimacy. Specifically, female students reported spending more time on following mainstream celebrities than male students (t MS = 2.665, p = 0.049), whereas male students perceived more reciprocity from mainstream celebrities than females (t MS = −3.163, p = 0.002). Male students also perceived a higher degree of intimacy from micro-celebrities than females (t Mc = −2.522, p = 0.013). In addition, male students reported a higher degree of perceived parasocial interaction with both mainstream and micro-celebrities (t MS = −2.037, p = 0.043; t Mc = −2.536, p = 0.012).

Hypotheses testing
The first aim of this study was to test to what extent the parasocial relationships with micro-and mainstream celebrities differ regarding four attributes: perceived reciprocity, authenticity, intimacy and parasocial interaction. A dependent samples t-test was performed on the subgroup of the participants who follow both mainstream celebrities and micro-celebrities (N = 160). The means, standard deviations, and the t values of all the main variables can be found in Table 3.
The first hypothesis that fans perceive a higher degree of reciprocity with microcelebrities than mainstream celebrities, was supported. Participants reported perceiving a lower degree of reciprocity from mainstream celebrities (M = 2.07, SD = 0.76) than from micro-celebrities (M = 2.35, SD = 0.82), t = −4.13, p < .001, BCa 95% CI [−0.41, −0.14], r = .32. The second and third hypotheses stated that participants perceive micro-celebrities as more authentic (H2a) and intimate (H3a) than mainstream celebrities. Counter to our expectations, however, participants perceived mainstream celebrities as more authentic than micro-celebrities. This difference, 0.22, BCa 95% CI [0.06, 0.37], was significant (t = 2.817, p = .006) and represented a small-sized effect d = 0.26. A similar result was found for the perception of intimacy. A higher degree of intimate relationships was perceived with mainstream celebrities than micro-celebrities (t = 3.6, BCa 95% CI [0.1, 0.36], p = .001). Therefore, H2a and H3a were rejected. With respect to the fourth hypothesis, followers perceive a higher degree of parasocial interaction with microcelebrities than with mainstream celebrities. The result shows that the difference (t = −0.257, p = .797, BCa 95% CI [−0.21, 0.16]) was not significant. H4 was not supported.
The second aim of this study was to test if the perceptions of reciprocity (H1b), authenticity (H2b) and intimacy (H3b) contribute to parasocial interactions that people experience with both mainstream and micro-celebrities. We tested these hypotheses with multiple hierarchical regression analyses. In the analysis, age, gender, time spent, and feedback frequency were used as control variables (Step 1), the perceptions of reciprocity, authenticity, and intimacy as predictors (Step 2), and parasocial interaction as the dependent variable (see Table 4).

Discussion
Although following celebrities on social media has become a common practice, research that examines differences in how people experience parasocial relationships with microcelebrities versus mainstream celebrities is sparse. This study expands the current literature by comparing parasocial relationships with micro-and mainstream celebrities in terms of audiences' perceptions of four key attributes of these relationships: reciprocity, authenticity, intimacy, and parasocial interaction (RQ1). Additionally, this study examined the roles of these attributes in fostering parasocial relationships, and explored whether they work differently for micro-and mainstream celebrities (RQ2). With respect to our first research question, the findings of our study support the first hypothesis: Greater reciprocity was perceived in the relationship with micro-celebrities than in the relationship with mainstream celebrities (H1a). This finding aligns with earlier observations that back-and-forth conversations between users and micro-celebrities are more likely to occur because social media platforms make it easier for ordinary people to access a micro-celebrity directly (Gamson 2011, Schouten et al. 2019, bypassing the barriers set by traditional media industries. In this way, the role of fans or audiences is elevated (Gamson 2011): On social media, their relationship with celebrities is no longer limited to a strictly unilateral relationship. This raises the question whether the term parasocial relationship is still apt in the context of micro-celebrity following, as a 'real' relationship with mutual interactions may develop.
Our hypothesis regarding the attribute authenticity (H2a) was rejected: Mainstream celebrities were perceived as more authentic than micro-celebrities. This result urges us to re-evaluate authenticity as an attribute characterising micro-celebrities. It is important to point out here that culture may play a role in how social media users evaluate the authenticity of especially micro-celebrities, as culture may influence social media production culture and aesthetics. In China, female micro-celebrities are often criticised for their indistinguishable appearance ('wanghong face') as a result of cosmetic surgery (Zhang and de Seta 2018). In general, Chinese micro-celebrity culture is criticised for playing the 'edge ball' strategy, by expressing oneself in ways that challenge the borderline of aesthetic and cultural norms, for instance through a livestream of 'erotic banana-eating' (Craig et al. 2021). These cultural aspects of Chinese micro-celebrity culture may affect audiences' perceptions of authenticity.
Our hypothesis concerning intimacy (H3a) was also rejected: Participants perceived a lower degree of intimacy for micro-celebrities than for mainstream celebrities. While we anticipated that micro-celebrities as ordinary people may share more similarities with followers resulting from their everydayness (Jerslev, 2016) and lesser control over their media personae by a media industry (Lobato 2016), our findings did not imply a higher level of perceived intimacy. We suggest that micro-celebrity culture is characterised by abundance, meaning that one user can follow numerous micro-celebrities instead of devoting their intimate attachment only to one performer. Also, the increasingly diversified content genres on social media may play a role Craig 2019, Craig et al. 2021) as several micro-celebrities feature talent display, stunts, and scripted performances in their content. In these content genres, intimacy practices are less important. To conclude, we suggest including the perceived similarity as a relevant difference between micro-and mainstream celebrities in future studies. Additionally, we advise examining the scope of micro-celebrities that people follow, as well as which content genres they produce. Finally, we recommend conducting cross-cultural content analyses of micro-celebrity content to examine to what extent culture affects content production.
The fact that the participants of our study perceived micro-celebrities as less authentic and intimate than mainstream celebrities might explain why our fourth hypothesis also failed: People indicated not experiencing greater parasocial interaction with micro-celebrities than with mainstream celebrities. Mainstream celebrities often use multiple communication channels, such as TV, film, or pop music as a vehicle to establish their celebrity personae (Gamson, 1994). Audiences' point of attachment, however, is ultimately one person, namely the celebrity that stars across all these channels. The increasingly diversified social media content genres again remind us that in many cases microcelebrities may only pay a role of producer of content instead of being the star in the content.
While our first research question looked for within-person comparisons in how individuals perceive and experience their relationship with micro-and mainstream celebrities, our second research question examined whether, for both types of celebrities, perceptions of reciprocity, authenticity, and intimacy positively predict experiences of parasocial interaction (RQ2). With respect to reciprocity (H1b), we found that perceived reciprocity positively predicts parasocial interaction, both for mainstream and micro-celebrities. This finding aligns with the findings of, among others, Baek et al. (2013), and remains of notice, given that earlier research asserted that reciprocity is not essential for experiencing parasocial interactions (e.g., Horton andWohl 1956, Rubin andMcHugh 1987). While that may still be true, our study suggests that reciprocity nonetheless matters. In agreement with also Chen (2014) and Kurtin et al. (2018), our study thus supports that reciprocity contributes to the development of parasocial relationships on social media. This finding implies that there may be need for a theoretical evolution of the parasocial relationship concept, specifically how it takes form on social media platforms.
In agreement with earlier studies (e.g., Horton and Wohl 1956, Rubin et al. 1985, Marwick and Boyd 2011, our study also found that, for both mainstream celebrities and micro-celebrities, perceived intimacy positively predicted parasocial interaction (H3b), suggesting that intimacy is still a crucial factor in the experience of parasocial interaction on social media.
As for authenticity as a predictor of parasocial interaction (H2b), however, our study found no support. The model tested for micro-celebrities even showed a marginally significant negative association between perceptions of authenticity and experiences of parasocial interaction. This finding requires us to carefully consider the concept of authenticity, and what it may mean in the context of an imaginary relationship. Scholars (e.g., Horton and Wohl 1956, Rubin and McHugh 1987, Giles 2002 have argued that viewers' parasocial experiences, albeit real in the mind, are eventually an imaginary process. This imaginary position is often held by the viewers themselves (Giles 2002). This implies that the imaginary process may involve audiences' creative interpretations of celebrity images. In this sense, whether a celebrity is authentic or not may become less relevant as audiences' can develop their own imagination.
An alternative, potentially concurrent explanation may be that while the concept of authenticity is deemed important for famous persons active in the field of marketing (Schoenmueller and Burhn, 2017) and political communication (Luebke 2020), it may matter less in general celebrity culture. Indeed, Marwick (2013) points out that authenticity is a historical and context-dependent concept covering the meanings of 'true to oneself', 'pre-or non-commercial' and 'distinguish oneself from others'. It is understandable when political and marketing communication strive for persuasive aims, the concept of authenticity become highly relevant and is operationalised within local contexts. As to parasocial interaction with celebrities, however, authenticity becomes only one type of celebrity attribute among others. Thus, our findings call for a more context-dependent conceptualisation of authenticity in micro-celebrity culture. When micro-celebrities are characterised by their status and visibility seeking practices (Senft 2008, Marwick 2013a, the goal to become popular through an inauthentic representation may not be read as that negative in the follower's imagined relationship with them.
Our control variable gender revealed a number of interesting associations. We found that male students spent less time on following celebrities, but perceived greater reciprocity and parasocial interaction with both mainstream and micro-celebrities than female students. These findings suggest that male students may establish parasocial relationships with celebrities more easily. Given that prior research shows differences in the types of celebrities followed, it would be interesting if future research could follow up on this gender difference, and explore whether the type of celebrity as well as other factors such as perceived similarity explain things.

Limitations and suggestions for future research
This study is subject to certain limitations. First, micro-and mainstream celebrities' communicative strategies can overlap. In the age of social media, the instrumental exchange (e.g. Hearn andSchoenhoff 2016, Khamis et al. 2017) is a tactic in practicing celebrity, meaning that micro-and mainstream celebrities may draw on each other's methods to gain fame. For instance, some micro-celebrities may draw on the mainstream celebrities' practice and employ a third party to better manage their fan base, as a result, they appear like mainstream celebrities in TV shows, tabloid news, etcetera. In turn, some mainstream celebrities may try to use reciprocal strategies to present their 'authentic' self (Creswell 2008) by coming down from the big screen and communicating with fans directly on social media. Mainstream and micro-celebrities adopt good use of each other's advantages, making their boundaries vaguer. This remains an interesting research question on classifying celebrities on social media.
Second, given that that the parasocial relationship is conceptualised as a long-term imaginary relationship with the celebrity, cross-sectional data may not be sufficient to adequately reflect the underlying mechanisms which trigger the parasocial interaction. A longitudinal study is needed to verify the determinants of parasocial relationships, and explore the changing nature of parasocial relationships in a long-term period.

Conclusion
In conclusion, this study offered insight into the attributes of the parasocial relationship on social media by comparing the parasocial relationships with micro-and mainstream celebrities, and testing which attributes predict parasocial interaction. Concerning the first research question, the findings suggest that fans perceive their relationship with micro-celebrities as more reciprocal than their relationship with mainstream celebrities. However, contrary to what previous studies suggest, participants perceived microcelebrities as less authentic than mainstream celebrities, and experienced their relationship with micro-celebrities as less intimate than that with mainstream celebrities. Perhaps in part because of the latter findings, we found no support for greater experiences of parasocial interaction with micro-celebrities on social media. For the second aim, we found that reciprocity and intimacy were positively related to parasocial interaction, whilst authenticity was not. At present, results based on causality are still hard to find in the literature of parasocial relationships on social media. Additional study based on longitudinal panel data is necessary to elucidate the factors in fostering parasocial relationships.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Notes on contributors
Yang Xu is a PhD candidate in the School of Humanities and Digital Sciences at Tilburg University.

Marjolijn Antheunis is Professor of Communication and Technology at the department of Communication and Cognition, Tilburg University.
Mariek Vanden Abeele is Associate professor of Online Culture at the department of Communication and Cognition and at the department of Culture Studies, Tilburg University.

Mingyi
Hou is Assistant professor of Culture Studies at the department of Culture Studies, Tilburg University.