Special Collection: Psychology of Live Streaming. Volume 4, Issue 3. DOI: 10.1037/tmb0000114
Vicarious interaction on live streams can occur when viewers experience what it is like to interact with a live stream persona by observing other viewers’ interactions with the persona. This study examined how vicarious interaction can cultivate parasocial relationships (PSRs) with live stream performers. Mechanical Turk workers were randomly assigned to view a clip of a live stream that either contained or did not contain instances in which the host addressed specific viewers in her audience. Results revealed an indirect effect of host–viewer interactions on PSR strength through participants’ imagination of how responsive the host seemed to other viewers. This effect was enhanced by parasocial interaction with the host. This study provides evidence of a secondhand, vicarious pathway to the development of PSRs.
Keywords: parasocial, vicarious interaction, responsiveness, live streaming
Funding: No external funding was provided for the completion of this study.
Disclosures: The authors have no known conflicts of interest to disclose.
Data Availability: The data that support the findings of this study are available from the corresponding authors upon reasonable request. The materials used in this study are available at this link.
Correspondence concerning this article should be addressed to Madison C. Martin, Department of Communication Studies, West Virginia University, P.O. Box 6293, Morgantown, WV 26506-6293, United States. Email: [email protected].
Live streaming is the act of broadcasting raw, real-time, video of an event, such as a video game, concert, or cooking segment over the Internet that is increasing in popularity. In the United States alone, the viewing audience for live streaming content grew from 126.7 million in 2019 to 158.2 million in 2022 (eMarketer, 2020). Part of the appeal of these broadcasts could be rooted in the ability of live stream hosts to facilitate feelings of parasocial relationship (PSR) closeness with their audience (Kneisel & Sternadori, 2022; Leith, 2021). Because live streaming success is determined by viewer attention, it behooves hosts to cultivate a sense of familiarity and closeness with their audience and keep them engaged. Live stream hosts accomplish this in ways that help their viewers get to know them, such as providing running commentary, disclosing details about their life, or displaying personal artifacts (Kneisel & Sternadori, 2022; Kowert & Daniel, 2021; Ruberg & Lark, 2020; Wohn & Freeman, 2020). Live stream hosts also forge connections by regularly interacting with their audience through greetings, requests for advice, and direct replies to notifications or comments left as chats or reactions (McLaughlin & Wohn, 2021; Wulf et al., 2021). However, even if a viewer never shares a one-on-one interaction with a live stream host, the possibility of interaction may be equally as important for cultivating PSR between viewers and hosts. As Kowert and Daniel (2021) pointed out, “the opportunity for reciprocity is omnipresent” during live streams, and this makes hosts seem inherently social.
Another hallmark of live streams that could facilitate PSR is a visible community that also engages with the host. Live stream viewers are regularly exposed to the hosts’ interactions with other viewers, and some have argued that the mere act of observing these social interactions between a media persona and their audience can give onlookers the opportunity to vicariously experience the hosts’ social warmth and responsiveness (Dai & Walther, 2018). By extension, it stands to reason that these imagined, secondhand social experiences with live streaming hosts can increase the magnitude of their PSR, vicariously. The present study utilized an experiment to examine the indirect effect of witnessing a live streamer host’s interaction with viewers on participants’ sense of PSR, through their imaginations of how responsive other audience members perceived the host to be. We also examined whether this perception interacts with parasocial interaction (PSI), a more direct intimacy-creating experience, to create a stronger viewer–host PSR.
A parasocial experience refers to a sense of quasi- or pseudo-social interactional or relational involvement with a media persona. These mediated connections are typically one-sided and do not provide opportunities for mutual interaction or reciprocal relational development between media performers and their audience (Horton & Wohl, 1956). Even in masspersonal environments (O’Sullivan & Carr, 2018), such as live stream platforms, where audiences have opportunities for limited, two-sided interaction with media personas, the one-to-many broadcasting model of these media events nonetheless maintains an imbalance in how socially available the performer seems to their audience compared to how socially available the audience seems to the performer. Kowert and Daniel (2021) referred to involvement with media personalities in these types of environments as “one-and-a-half-sided” parasociality. These experiences are not completely one-sided as they would be in noninteractive media contexts but given media performers’ higher status and their greater control over the communication affordances on the platform, they are not completely reciprocal either.
Two related, yet distinct, parasocial experiences are PSIs and PSRs (Dibble et al., 2016; Horton & Wohl, 1956). PSI refers to the illusion of two-way interaction with a media figure, a perception of mutual awareness, or the “simulation of conversational give and take,” (Horton & Wohl, 1956, p. 215). These experiences are confined to the live media experience (Dibble et al., 2016) and are enhanced when media figures make some acknowledgment of the audience through conventions such as breaking the fourth wall with direct verbal or bodily address (Hartmann & Goldhoorn, 2011; Ferchaud et al., 2018). PSRs, on the other hand, refer to feelings of relational attachment to media characters that endure beyond media exposure. Hallmarks of PSRs include interest, a sense of familiarity, and high levels of cognitive and emotional involvement with media figures (Cohen et al., 2021). As Horton and Wohl (1956) explained, this sense of bonding between members of an audience and media performers or public figures is illusionary in the sense that it is nonreciprocal and therefore not genuinely social. Despite the imaginary nature of both PSI and PSR, studies spanning several decades have demonstrated these reactions to media personas are psychologically normal (Reeves & Nass, 1996), and they operate similarly to nonmediated social encounters (see Tukachinsky Forster, 2023).
In their application of the model of interaction stages (Knapp et al., 2014) to understand PSR relationship development, Tukachinsky and Stever (2019) explained that PSI, or the “illusion of reciprocity and interactivity” can enhance PSRs (p. 305). Although the development of a PSR is not contingent on the experience of PSI or vice versa, the two concepts are usually correlated positively (Dibble et al., 2016; Hu, 2016). This could be because some media conventions can affect an audience’s sense of both PSI and PSR. For instance, if a media figure appears to make eye contact with their audience, this could facilitate both a feeling of mutual awareness and social attraction. PSI could also be a proxy for media figure responsiveness or how receptive the media figure would be as a relational partner.
Responsiveness can be conceptualized as a mix of interaction and relationship-related qualities, including attentiveness, listening, understanding, caring, a warm demeanor, and validation (Canevello & Crocker, 2010; Reis et al., 2017). Partner responsiveness has long been understood as an important ingredient in interpersonal relationship intimacy (e.g., Foot et al., 1977; Gadassi et al., 2016; Laurenceau et al., 2005). Davis and Perkowitz (1979) demonstrated experimentally that participants in a dyadic interaction perceived a confederate partner who responded to their disclosures more often as being more responsive (e.g., more attentive and sympathetic), and they perceived them as having greater friendship potential as a result.
Complimentary theorizing suggests that responsiveness also facilitates the development of these PSRs. Giles (2002) argued that parasociality was, in part, a function of what the chances were that an audience member could cultivate a genuinely social relationship with a media figure. He argued that “semiparasocial” contact occurs when there are chance meetings with media figures (e.g., a handshake at a fan political rally or a fan letter reply), but there is little chance that the media figure will maintain an interaction or relationship. Similarly, Hartmann (2008) suggested that the more a media figure seems relationally “within reach” or available for communication, the more likely they are to be an object of parasocial engagement (p. 184). Although the relationship between responsiveness and PSR has not been examined directly, Thorson and Rodgers (2006) found that the ease with which social media users felt they could communicate with a political candidate predicted their parasocial experiences with them. Additionally, Cohen and Tyler (2016) found that PSI with a public figure on Twitter was reduced if they had somebody else manage their account because it made the public figure seem less available for interaction. Most recently, McLaughlin and Wohn (2021) found that viewers who could recall being acknowledged by a live stream host also reported stronger PSI and PSRs. Collectively, these findings provide some evidence that the more communicatively receptive and responsive a media figure is, the more viable they should seem as parasocial “partners.”
Of course, although the opportunities to engage in two-way communication with media personas in new media environments such as live streams are always available, not all viewers take advantage of them. Besides appealing to viewers who come to converse with others, or leave reactions, comments, or questions for the performer, Speed et al. (2022) observed that Twitch (a popular live streaming platform) also attracts “loyal lurkers” who simply view live stream personas and follow the chat conversations. Although the digital affordances of these platforms encourage social interaction, arguably the main event of many of these broadcasts is the vicarious enjoyment of some other activity, such as a video game or interactions between other community members. For this reason, it makes intuitive sense that many viewers would prefer to simply observe live streamers and their audience as spectators, and that even the most socially active viewers would also spend some of their time watching interactions instead of participating in them directly. Yet it would be a mistake to think of these indirect, observational forms of viewership as a signal of disengagement with the host. Spectators can still interact with live stream hosts, albeit vicariously.
A year following the publication of the seminal description of parasocial experiences (Horton & Wohl, 1956), Horton and Strauss (1957) explored two additional modes of pseudointeraction specific to performances that encourage audience participation. One of these experiences was vicarious interaction, defined as a process in which audience members observe other audience member’s responses without taking part or being acknowledged by the media persona. In the context of television, conventions such as audience reaction shots and audience applause and laughter encourage vicarious spectatorship, in which viewers see the performance through the eyes and ears of other viewers. Notably, Horton and Strauss (1957) did not regard these secondhand audience experiences as being any less engaging or enjoyable than more direct experiences. They viewed these audience observations as a component of the overall entertainment experience that sometimes “shadows the parasocial performance itself” (p. 582).
It stands to reason that these vicarious interactions with performers through other audience members can enhance observers’ sense of PSR. Social cognitive theory explains how, rather than learning from direct, firsthand experience, people often learn how to respond in different situations by observing others, especially others with whom they feel similar (Bandura, 1986). One outcome of this social, vicarious, and learning process is that after watching a similar other engage in contact with another person, the observer can develop cognitive models of the interaction experience that they can apply to understand what it would be like if they were to engage in the same interaction (Harwood, 2021). By extension, it stands to reason that when viewers watch other audience members have positive interactions with a media persona, they can imagine their response to the performer’s attention and reciprocity. In doing so, by imagining how cared for others must feel during these exchanges, these vicarious experiences can positively impact on the observers’ own sense of PSR with the persona.
Dai and Walther (2018) explained this vicarious interaction process as a type of relationship forecasting, stating that “by observing the surrogate, the observer projects how the target person is most likely to act toward anyone—including him- or herself” (p. 322). In their study, they found that Twitter users who observe interactions between celebrities and other Twitter users increased what they termed, “parasocial intimacy,” with the celebrity. Providing evidence for the vicarious nature of this relationship, this effect was strengthened the more that observers felt bonded and similarity with the other Twitter users they observed.
What Dai and Walther (2018) conceptualized as parasocial intimacy could alternatively be labeled “relationship potential,” or how likely a celebrity would be to understand and care for followers if they were given the opportunity. They measured parasocial intimacy with an index of hypothetical partner responsiveness (Canevello & Crocker, 2010), and participants in their study reported on how attentive and sensitive they forecasted that the media figure might be to them, personally. This allowed them to demonstrate that observers were considering what it would be like to have a relationship with the celebrity, and that this perception was a function of how the celebrity interacted with others. However, although this study provided intriguing preliminary evidence that observation of others’ interactions with a media persona can enhance parasocial viewers’ feelings for them, it did not specifically show that this process was vicarious. That is, although Dai and Walther’s (2018) study is among the first to demonstrate that the observation of a media persona’s interaction increases perceptions of responsiveness (particularly if onlookers felt a kinship with other viewers), they did not investigate whether the imaginations of others’ experience was driving this effect. To extend their work, the present study examines the vicarious experience of a media figure’s responsiveness to others as the mediator of host–viewer interaction and strength of PSR with the live streamer.
Friendly live streamer interactions with their audience can positively affect their viewers’ PSR either firsthand, by demonstrating that they are responsive, or secondhand, by experiencing this responsiveness through others’ experiences with the live streamer. The first process, characterized by experiences with the media figure such as PSI, involves an audience member’s sense that they are engaged in social interaction. The second process is less direct and arguably more imaginative and immersive, as it involves an audience member’s considerations of what interacting with the media figure is like for other audience members. During vicarious interaction, observers consider the experience of other viewers to inform their own experience. Put differently, both paths to PSR development are imaginative, but one process involves imagining the media figure’s interactions with the self, while the indirect, vicarious process involves imagining the media figure’s interactions with others.
This study focuses on this vicarious PSR development process. Observing specific audience members interacting with a live stream host should increase how much those spectators imagine others’ perspectives and reactions in these interactions. The more that viewers vicariously experience the media persona’s responsiveness, the stronger their personal PSR with the host should be. Accordingly, we predicted the following:
Hypothesis 1: Compared to those not exposed to host–viewer interaction, those exposed to others’ interactions with a live streamer will report stronger PSRs with the persona, an effect that will be mediated by the media persona’s imagined responsiveness to others.
Of course, as previously discussed, one key characteristic of live streams that makes them ripe for parasocial connection is how the platform itself creates an illusion of relational reciprocity through the cultivation of PSI. Independent of what other viewers are experiencing, audience members’ own sense of mutual awareness or PSI with the host is an important ingredient in the cultivation of PSRs (Tukachinsky & Stever, 2019). We propose, however, that this direct, firsthand experience with the host works in concert together with less direct, vicarious experiences to strengthen PSRs with live streamers.
As previously discussed, imagined responsiveness is theorized to drive the process of PSR development through vicarious interaction because it is an indicator of relationship potential (Dai & Walther, 2018). This prediction is consistent with theorizing, which suggests that parasociality is a function of how many relationships potential media figures are perceived as having (Giles, 2002; Hartmann, 2008). Vicarious or direct, arguably all pseudosocial perceptions are rooted in an illusion of an opportunity for reciprocity (Horton & Wohl, 1956; Kowert & Daniel, 2021). Because both the direct path to PSR development through the self and the indirect path to PSR development through others are rooted in a perception of relational potential, they should work in tandem to enhance PSRs. That is, media persona engagement involving the self, such as PSI, plus vicarious engagement, such as imagined responsiveness to others, should contribute to the sense that a media performer is an accessible and responsive social actor. For this reason, we expect that perceptions of PSI should act as a contributory moderator of the effect of imagined responsiveness to others on PSR. More broadly, we expect that the entire process of vicarious interaction will be conditional on this moderator. Specifically, the positive indirect effect of host–viewer interaction on PSR through imagined responsiveness will positively increase as a function of perceptions of PSI. Accordingly, we proposed two additional hypotheses. Figure 1 illustrates the predicted model.
Hypothesis 2: The effect of imagined responsiveness to others on PSR will be conditional on PSI, such that the relationship will be stronger for those who have stronger PSI with the live stream host.
Hypothesis 3: The indirect effect of vicarious interaction on PSR through imagined responsiveness will increase with stronger PSI with the live stream host.
The study design was approved by the West Virginia University institutional review board (Protocol No. 2209654053). Participants were recruited on Amazon’s Mechanical Turk (AMT) with an advertisement for a study about “impressions of live streamers.” Participation was restricted to U.S. residents, aged 18 years or older, who had completed at least 500 AMT tasks with at least a 95% approval rating. Research has shown that AMT can potentially provide access to conscientious study participants and useful survey responses; however, the data need to be closely monitored for quality (Robinson et al., 2019). Although 367 participants began the study, data from eight participants needed to be deleted for submitting incomplete responses, and data from nearly half (47.54%) needed to be omitted because participants submitted nonsensical or incorrect answers to quality-assurance questions designed to gauge their attention (e.g., “True or False? Lemon Drop, the host of the live stream you watched had blond hair”). The remaining participants were compensated with 90 cents. There were 184 participants in the final sample, which was predominantly White (76.1%), male (58.2%), and highly educated, with over 70.5% having earned at least a bachelor’s degree.
The study was conducted online, using Qualtrics survey software. After reading a cover letter describing the study and consenting to participate, participants were randomly assigned to one of two versions of a live stream clip embedded in the online survey. They either watched a version in which the live streamer host addresses specific viewers (n = 91) or a version in which these acknowledgments were edited out (n = 93). After watching, participants completed a questionnaire that asked about their impressions of the live stream. Materials for this study can be viewed at https://osf.io/tdz5j/?view_only=1c12879df46441899ab90eaaf3f79d0d.
Although video gaming is by far the most popular category of live streaming content categories on Twitch, there are a variety of topic channels on the platform. We selected a live stream that focuses on cooking in hopes that it would have broad appeal to a variety of participants, regardless of whether they had an interest in gaming. The selected live stream was hosted by Lemon Drop (@ARecipeReborn), a female, biracial Asian–Canadian culinary arts student. Lemon Drop has 7.3 thousand followers on Twitch and hosts an average of three, 4-hr cooking streams per week (Twitch, n.d.).
Participants were randomly assigned to watch one of two edited clips of a stream that aired live during September 2022. During this segment, Lemon Drop prepares to cook gyros, discusses what she did over a holiday weekend, and her thoughts on a popular TV show. In the host–viewer interaction version of the live stream, she engages in several instances of direct interaction with her audience. During these interactions, she simply acknowledged audience members with short greetings (i.e., “Hello, Marie, thanks for the bits!”). The total length of this video clip, including the host–viewer interactions, was 4 min and 15 s. In the condition without any host–viewer interactions, these greetings were edited out. This necessarily reduced the length of the video in this condition to 2 min and 32 s.
Besides the inclusion of greetings to viewers in the host–viewer interaction condition and the final length of the clips, the videos were identical. Our objective was to manipulate the presence or absence of host–viewer interactions but hold all other variables constant. Thus, the greeting segments included in the host–viewer interaction condition were carefully edited to exclude any details or information about the host that was not revealed to participants assigned to watch the other version of the clip. Care was taken to ensure that no variables that could influence viewer perceptions of the host, including information about her, varied between the two conditions, and we confirmed that the differences in segment length did not affect viewer perceptions of the host’s self-disclosure. The results of this analysis are included in the preliminary analyses.
All the measures in this study utilized a 7-point scale, ranging from 1 = strongly disagree to 7 = strongly agree. Canevello and Crocker’s (2010) six-item scale of partner responsiveness in interpersonal communication was adapted to measure imagined responsiveness to others (M = 5.21, SD = 1.01, ω = .86). Because this study was concerned with a vicarious process, scale items were adapted in such a way that required participants to consider how the host made other audience members feel (e.g., “Lemon Drop made others feel comfortable about themselves,” “Lemon Drop made others feel valued as a person”). PSR strength was measured with eight items from Rubin and Perse’s (1987) parasocial interaction scale (M = 4.12, SD = 1.23, ω = .89; e.g., “I would like to meet Lemon Drop in person,” “I feel as though I have known Lemon Drop for a long time”). Hartmann and Goldhoorn’s (2011) six-item Experience of Parasocial Interaction scale was used to assess participants’ PSI with the live stream host (M = 3.97, SD = 1.79, ω = .97; e.g., “Lemon Drop was aware of me,” “Lemon Drop knew that I paid attention to her”). Finally, we also adapted seven items from Ma and Leung’s (2006) measure of Internet self-disclosure (e.g., “Lemon Drop is always honest in her self-disclosures”), so that we could determine whether or not the experimental manipulation inadvertently affected perceptions of the live stream host’s valence of self-disclosure in terms of honesty and positivity (M = 4.58, SD = 1.07, ω = .77).
A series of t tests examined the difference between participants assigned to the two experimental conditions and variables of interest. There was a significant difference between the two conditions in terms of responsiveness, t(182) = −5.106, p < .001, demonstrating that exposure to host–viewer interactions had the intended effect on participant’s perceptions of the host’s responsiveness to others. Viewers imagined the live streamer as being more responsive to others in the host–viewer condition (M = 5.57, SD = .81), compared to the condition in which these interactions were edited out (M = 4.86, SD = 1.05). The manipulation did not significantly affect any other perceptions of the live streamer; however, in terms of PSR strength, t(181) = .59, p = .55, PSI, t(182) = −.84, p = .40, or perceptions of disclosure valence, t(182) = −.17, p = .87. We also examined whether the edits made for the experimental manipulation affected the viewing experience at all (measured with single items). Participants did not report any differences between conditions in live stream enjoyment, t(181) = −.64, p = .52, production quality, t(181) = −1.35, p = .18, or how understandable it was, t(180) = −.50, p = .62.
Notably, many participants thought that they recalled seeing direct communication between the host and specific audience members, even if they were not exposed to it. After watching the live stream, participants were asked whether “Lemon Drop spoke directly to audience members (i.e., addressed them by name) who wrote to her in the chat.” This question was asked as an attention check. There was a significant relationship between the experimental conditions participants were assigned to and their response, χ2(1, N = 184) = 40.41, p < .001. Among participants in the host–viewer interaction condition, nearly all participants correctly recalled the host addressing other audience members (97.8%). However, among participants who were not exposed to any host–viewer interactions, less than half (40.9%) correctly indicated that no interactions occurred. Because there was a significant difference in imagined responsiveness between conditions even though participants in the no-interaction condition did not notice the lack of viewer–host interactions, this demonstrates that the study’s manipulation was more influential than participants’ recall. Most participants believed the host addressed specific audience members. Regardless of what they could remember, those that were not exposed to these interactions engaged in significantly lower levels of imagined responsiveness.
Table 1 displays zero-order correlations between the variables of interest. Neither participant age nor gender were related to any of the variables of interest. However, the frequency that participants watched live streams was positively correlated with perceptions of responsiveness, r = .27, p < .001, PSR strength, r = .26, p < .001, and PSI, r = .33, p < .001. Additionally, self-disclosure valence was also positively associated with perceptions of responsiveness, r = .36, p < .001, PSR strength, r = .30, p < .001, and PSI, r = .41, p < .001. For this reason, both live stream viewing frequency and self-disclosure valence were entered as a control variable in the remainder of analyses.
Zero-Order Correlations Between Variables of Interest
1. Gender (0 = male)
3. Frequency of live stream viewing
4. Self-disclosure valence
5. Imagined responsiveness to others
6. Parasocial interaction
7. Parasocial relationship
* p < .05. ** p < .001.
The first hypothesis predicted that there would be an indirect effect of the host–viewer interaction manipulation on PSR strength, through estimates of others’ perceptions of media figure responsiveness. To examine this, the PROCESS macro for SPSS (v4.0) was used to run Model 4 with 10,000 bootstrap samples for percentile bootstrap confidence intervals. The experimental condition (no host–viewer interaction = 0) was entered as the independent variable, PSR strength was entered as the dependent variable, and imagined responsiveness to others was entered as the mediator. This analysis showed that although the direct effect was not significant (c′ = −.21, BootSE = .17, p = .21, 95% CI [−.55, .12]), there was a significant indirect effect of host–viewer interaction on PSR strength, through imagined responsiveness, ab = .32, BootSE = .09, 95% BootCI [.17, .50]. H1 was supported.
Next, Model 14 of the PROCESS macro was used to examine PSI as a moderator of this mediated effect. This model was identical to the first model tested, except that PSI was entered as a second-stage moderator of the effect of responsiveness on PSR strength (the b path). The model was run with 10,000 bootstrapped samples for percentile bootstrap confidence intervals. In support of H2, the interaction term was statistically significant. PSI enhanced the effect of others’ perceptions of imagined responsiveness on PSR strength, B = .09, SE = .04, p = .015, 95% CI [.02, .17]. The results of this analysis are displayed in Table 2. The Johnson–Neyman procedure was used to probe this interaction (see Table 3). The greater viewers’ perception of parasocial interaction with Lemon Drop was the greater the effect of host–viewer interaction on PSR strength through the imagination of others’ perceptions of responsiveness. Furthermore, consistent with H3, the analysis also uncovered evidence of a conditional indirect effect. The index of partial moderated mediation quantifies the relationship between the moderator and the size of the indirect effect of the independent variable on the dependent variable. In this case, the index of partial mediated moderation for PSI (index of partial moderated mediation = .07, BootSE = .04, BootCI [.01, .15]) was significant, indicating that the indirect effect of host–viewer interaction on PSR strength was conditional on PSI. As displayed in Table 2, the size of the indirect effect of the host’s direct address on PSR strength through others’ perceptions of responsiveness increased with stronger perceptions of PSI.
OLS Path Model Coefficients
Imagined responsiveness to others
F(3, 178) = 25.75, p = .000, R2 = .302
Host–viewer interaction (0 = no direct interaction)
Frequency of live streaming viewing
Perceptions of disclosure
Parasocial relationship strength (PSR)
F(6, 175) = 19.636, p = .000, R2 = .402
Host–viewer interaction (0 = no direct interaction)
Imagined responsiveness to others
Parasocial interaction (PSI)
Imagined Responsiveness × PSI
Frequency of live streaming viewing
Perceptions of disclosure
Conditional indirect effect of host–viewer interaction on PSR strength
Host–viewer interaction → imagined responsiveness → PSR at 16th percentile of PSI (1.667)
Host–viewer interaction → imagined responsiveness → PSR at 16th percentile of PSI (4.667)
Host–viewer interaction → imagined responsiveness → PSR at 16th percentile of PSI (5.667)
Note. OLS = ordinary least squares; SE = standard error; LL = lower limit; UL = upper limit; CI = confidence interval.
The Conditional Effect of Parasocial Relationship Strength at Different Values of Parasocial Interaction
Note. The region of significance is displayed in boldface. The moderator value of parasocial interaction defining the Johnson–Neyman significance region is 1.095, with 86.813% of the values in this region. When parasocial interaction is below 1.095, the effect of parasocial interaction on the conditional effect of parasocial relationship strength is not significantly different from zero. SE = standard error; LL = lower limit; UL = upper limit; CI = confidence interval.
One of the defining features of most live stream broadcasts is viewer interactions with the host. By utilizing interactive chat and reaction affordances on live stream platforms, audience members can engage in “one-and-a-half-sided” conversations with the performer (Kowert & Daniel, 2021). Of course, not all viewers take advantage of these social opportunities, and even those that do may nonetheless spend much of their time observing others’ interactions with the live streaming host. As spectators of these interactions, observers can vicariously assume the role of other audience members and experience social encounters with the host, secondhand (Harwood, 2021; Horton & Strauss, 1957). Extending Dai and Walther’s (2018) finding that these vicarious interactions—the observation of host—viewer communication—can enhance parasocial intimacy, the present study utilized an experiment to examine the effect of a live streamer host’s direct communication with specific audience members on PSR. Imagined responsiveness to others’ was a mediator of this effect, and PSI, a variable that has a more direct influence on intimacy perceptions, was examined as a moderator of this effect.
As predicted, the host–viewer interaction manipulation had an indirect effect on PSR through imagined responsiveness to others. That is, participants exposed to the interactions reported higher levels of imagined responsiveness with the live stream host and, in turn, had stronger PSRs with her. Consistent with Dai and Walther’s (2018) claim that other social media users can effectively become surrogate social participants, giving observers an opportunity to sample what is like to have the attention of a media figure, this finding provides evidence of a vicarious pathway to developing PSRs. Seeing Lemon Drop’s acknowledgments and interactions with specific viewers prompted participants to consider how responsive she must have seemed to others. This process of imagining how she seemed from others’ perspectives increased observers’ personal PSR with her.
This finding underscores an unconventional reason that live stream platforms and cultures that encourage community engagement can facilitate deeper parasocial connections with the live stream host. Decades of research on parasocial experiences has established a long list of factors related to media personas (e.g., self-disclosure), media affordances (e.g., accessibility), and individual viewers (e.g., attachment style) that can predict the development PSRs (see Tukachinsky & Stever, 2019). The process of PSR development is typically assumed to occur firsthand, in which media personas initiate and intensify parasocial connections through their audiences’ direct exposure. This study demonstrates, however, that individuals’ feelings of parasocial closeness to media personalities are not always purely an outcome of their own experience. Sometimes, the process of PSR development can also be shared among a community of coengaged followers and fans who can experience the media figure as a pseudofriend with their own observations and experiences and through the imagination of others’ experiences.
This study also examined whether more direct, firsthand perceptions of PSI would work in concert with the vicarious interaction process to enhance PSR. Unlike vicarious interaction, which is an indirect imaginative process of connection with a media persona through others, PSI involves the self, directly. However, both perceptions are theorized to intensify PSRs because they are rooted in viewers’ regard for media figures as being accessible, reciprocal, and therefore viable relational actors (Hartmann, 2008). Therefore, we predicted that the increase in PSRs through the vicarious interaction process should be bolstered by concurrent, self-related perceptions of PSI.
As expected, we did find that PSI moderates the effect of vicarious interaction process on PSR. PSI complimented the positive effect of host–viewer interaction and imagined responsiveness on PSR. Live stream viewers’ feelings of intimacy and connection with the host likely draw on both types of experiences. In fact, it is plausible that viewers switch back and forth between direct and vicarious modes during viewing. Theoretically, the same techniques and that facilitate PSI, such direct bodily address (Hartmann & Goldboorn, 2011), can also operate on a vicarious level if observers imagine others’ parasocial reactions. Whether viewers had the sense that a media persona was communicating with them directly or whether they imagined her responsiveness vicariously, they would have likely perceived her to be a warm, familiar, and accessible relational partner.
Interestingly, on the whole, most participants in the present study were under the impression that Lemon Drop engaged in direct interactions with members of the audience, even if they did not actually witness it. This is notable for two reasons. First, it speaks to how normative host–viewer interactions are on live streams. These interactions are so normative that viewers assume they occur, even if they do not witness them. It is also possible that the style of these broadcasts is so effective at creating a sense of presence, or nonmediation, that viewers have difficulty distinguishing between media performers’ PSI with themselves and indirect interaction through others. In a masspersonal context (O’Sullivan & Carr, 2018) like live stream broadcasts, where the boundaries that define communication between one-to-many, one-to-one, and many-to-many sometimes bleed together, it should come as no surprise that these imaginative social experiences are fluid too. Second, this detail is worth noting because it suggests that viewers’ imaginations of responsiveness to others were not an artifact of people’s general impressions or expectations of the host but rather a genuine reaction to her observed rapport with others. After all, most participants thought that she was gregarious enough to directly address specific viewers. Yet, despite the fact, viewers in both conditions believed that she initiated host–viewer interactions, only the viewers who actually witnessed it appear to have engaged in vicarious interaction.
Though this study provides compelling evidence that vicarious interaction as a result of exposure to host–viewer communication can be one path to the development of PSR, the findings should be interpreted within the light of several limitations and open questions for future research. First, this research relied on a sample from AMT, which limits the generalizability of our results. AMT workers provide a diverse sample (at least relative to other convenience samples) but one that tends to be more educated and technologically savvy compared to the U.S. population at large. Furthermore, although we omitted data from participants who showed any evidence of disregard for the study protocol, this did certainly not necessarily guarantee that the remaining participants were steadfastly attentive.
The viewing experience was also quite artificial and unlike an authentic live streaming viewing experience. The edited clip of the live stream was not, in fact, live from the point-of-view of the participants. Consequently, participants did not have the ability to directly interact with the live stream host or with other viewers like they may have if they were watching a true live stream. The broadcast was also noticeably edited. The clip participants were asked to watch was quite short. Live streams are typically 3–4 hr in length, which promotes lengthy viewing sessions in which viewers and hosts have plenty of time to develop relationships with one another, whether they are social, parasocial, or something in between.
Furthermore, our failure to observe a direct effect of host–viewer interaction on PSR raises the possibility that the live streamer’s interactions with her viewers in our stimulus were perhaps not remarkable enough to elicit feelings of intimacy unless participants went to the extra cognitive step of becoming vicarious involved with the viewers she engaged with. The interactions participants were exposed to were quite superficial, consisting of only perfunctory salutations and cursory acknowledgments. These interactions were limited in the interest of experimental control, to ensure that the presence of host–viewer interactions did not introduce potential confounds such as differences in disclosure perceptions. During natural live stream viewing experiences, viewers are likely to observe host–viewer interactions that demonstrate greater rapport, familiarity, and shared history. Exposure to more interactions with greater depth may be able to foster PSR through direct and vicarious means, but this is a possibility that merits additional research. These are just a few of many ways that the live stream viewing experience examined in this study deviated from the experience that voluntary viewers of authentic live streams have. Future theorizing on vicarious interaction during live streams would be well-served by examining how vicarious interaction occurs during a natural viewing experience, perhaps by soliciting retrospective survey responses from live stream audiences.
The results of this study make several contributions to understanding the mechanics of live streaming and theorizing on parasocial experiences more broadly. First, these findings suggest that spectatorship might sometimes be as psychologically engaging as interaction. A somewhat unique affordance of live streams is the ability for viewers to watch other viewers’ reactions to the content during the stream, and these experiences give observers the opportunity to engage with live stream hosts vicariously, without any interaction on their part. This should come as reassuring news to live stream hosts who are likely able to cultivate a more dedicated group of subscribers and followers through interactions with relatively few viewers. Hosts apparently do not need to do much in order to foster vicarious interaction with their audience. This study suggests that even simple salutations and “shout-outs” are sufficient to increase onlookers’ perceptions of their responsiveness. These findings can also speak to the appeal of viewing live streams that are not actually live. It is notable that the live stream studied in this research was prerecorded and that viewers did not have an opportunity to synchronously interact with the host, yet this did not prevent them from engaging vicariously.
This research also contributes to theorizing on parasocial experiences by shedding light on an underexplored form of media figure involvement. Extending Horton and Strauss’s (1957) and Dai and Walther’s (2018) conceptualization, this study is among the first to provide empirical evidence of vicarious interaction with media figures. The phenomenon itself is not new—arguably any sort of coviewing media experience should lend itself to vicarious media consumption experiences, but the process and outcomes of vicarious interaction are not well researched. Yet, anecdotally it stands to reason that vicarious engagement with media figures should have increasing relevance in a digital media ecology where viewers are afforded more opportunities to broadcast themselves, interact with media performers, and observe those interactions from a distance. A more comprehensive understanding of how and why audiences form attachments to media personalities will require consideration of both direct and surrogate audience processes.