2.1 Foundations

Although it was the first theory applied to CMC, Short, Williams, and Christie’s (1976) social presence theory presented a dual focus, with the latter of the two aimed directly at the nature of interpersonal interactions. It specified that communication media differ in their capacity to transmit nonverbal communication code systems (kinesics, facial expressions, gestures, vocal cues, touch, proxemics, physical appearance and dress) in addition to verbal content. The theory argues, on this basis, that the fewer the number of cue systems a medium supports, the less social presence one experiences in interaction with others, that is, the less the salience of one’s communication partners to him or her. When there are fewer nonverbal code systems supported by a medium, the less one’s sense that others are there.

The second aspect of the original theory specified that the less social presence one experiences due to a medium’s characteristics, the less interpersonal warmth and pleasantness users experience with one another. Although the theory was developed to describe differences between FtF, videoconferencing, and audioconferencing, early adopters of the model applied it to CMC (e.g., Rice and Case 1983) predicting that CMC, because of its dearth of nonverbal cues, conveys less socioemotional content than other, multi-modal forms of communication.

A similar position appeared, with specific regard to CMC and not videoconferencing more generally, in the lack of social context cues hypothesis (Kiesler, Siegel, and McGuire 1984; Siegel et al. 1986). This approach specified that CMC blocks the nonverbal cues that signify the roles and the status of individuals participating in discussions. In this view, text-based messages make one individual’s communication impersonal. Devoid of cues that normally regulate interaction and inhibit anti-normative behavior in face-to-face (FtF) communication, it posits, individuals’ attention is shifted toward the task as opposed to the person, and they become less concerned about public evaluation and more resistant to social influence (Joinson 1998).

Numerous experiments supported the social presence and lack of social context cues contentions. Typically, small groups in CMC conditions or FtF conditions addressed some decision-making task for some fixed amount of time. Comparisons of the transcripts from these conditions revealed less socially oriented communication, more hostile communication, fewer agreement messages, and less group consensus in CMC groups than FtF groups (e.g., Dubrosky, Kiesler, and Sethna 1991; Hiltz, Johnson, and Turoff 1986; Siegel et al. 1986). Most of these experiments appeared in the 1980s and 1990s, when CMC was a relatively new phenomenon, and prior to the Internet’s widespread diffusion among the public. Nevertheless, the notion that CMC without nonverbal cues inhibits effective detection of others’ characteristics or the nuanced meaning of their messages, especially compared to audio- and video-based media, still emerges quite frequently in the literature. For instance, Friedman and Currall (2003) claimed that, coupled with diminished feedback opportunities preventing quick repairs due to its asynchronous nature, the reduced individuality inherent in email aggravates disputes and escalates conflict. Email negotiators were less likely to engage in rapport-building communication and more prone to adopt an aversive emotional style, negatively confronting and intimidating each other (i.e., squeaky wheel bias), compared to those negotiating FtF (Thompson and Nadler 2002). Similar positions about the inferiority of CMC to multimodal interfaces appear with regard to forming impressions (Epley and Kruger 2005) and establishing liking (Okdie et al. 2011).

The theories’ intuitive claims and the straightforward nature of the experiments supporting them made them very popular, especially at a time when academics and the public began to consider seriously what computerized communication would do to people. Publications with catchy titles such as “Don’t Forget Dr. Frankenstein” (Sproull and Kiesler 1992) fed suspicions about the dehumanizing potential of the mysterious new technology and concerns about machine-like perceptions of communication partners (i.e., mechanomorphism; Shamp 1991). At the same time, a number of anecdotal reports emerged in both the academic and popular press depicting the development of close relationships via CMC systems (e.g., McKenna and Bargh 1999). Confiding in strangers, developing intense friendships, and romantic infatuation online with people one did not know and had never seen were not only disconcerting in the public eye, but remained unexplainable by extant CMC theories. Since social presence theory suggests that social dynamics in CMC were determined by the structural properties of the medium, variations in the nature of CMC could not be accounted for by this approach.

2.2 Social presence theory today

Although social presence was originally conceptualized as perceived salience of the mediated partner, which in turn affected the quality of interpersonal interaction, the study of social presence in CMC has focused on how “sociable, warm, sensitive, and personal” individuals perceive their mediated partner to be (Short et al. 1976: 66), although the former aspect has maintained a strong focus in research on virtual reality systems and, more recently, online games. In CMC research the construct has heavily gravitated toward intimacy (social presence), with relative disregard for involvement or connectedness (social presence). As a result, the ultimate utility of the social presence construct to understand and explain CMC users’ experience is limited (cf. Burgoon et al. 2002). Moreover, the positive connotations implied by the dimensions of sociability, warmth, and personalness in the construct’s operational definition seemed to corroborate the taken-for-granted assumption that FtF is the gold standard up to which other, mediated, forms of human communication cannot measure.

The notion that people interact not only with others through a medium, but also with the medium itself, prompted human-computer interaction (HCI) researchers to redefine social presence to encompass non-human artificial objects. For example, social presence has been conceptualized as “the extent to which other beings (living or synthetic) coexist and react to you” (K. M. Lee 2004: 41), shifting the focus from perceived “humanness” toward “interactivity.” Employing the revised concept, studies have examined how social presence mediates (K. M. Lee et al. 2006) or moderates (Skalski and Tamborini 2007) the effects of the virtual actors’ characteristics, such as personality and physical attractiveness.

Other research has established that individuals interacting with computers sometimes experience a surprisingly high level of presence based on characteristics of the system’s interface. In stark contrast to the view that a wide variety of nonverbal cues are necessary to impute personality online, the computers are social actors (CASA) paradigm has focused on the reverse tendency in HCI: It posits that people’s responses to computers are fundamentally social, mirroring the way they interact with other people (Nass and Moon 2000; Reeves and Nass 1996). For example, people attribute gender-specific knowledge to computers that issue male voices versus female voices (Nass, Moon, and Green 1997); they prefer a computer the verbal instructions of which reveal a dominant or submissive personality that is similar to their own (Moon and Nass 1996); they reciprocate favors they receive from a computer by working harder on subsequent tasks in order to “help” the computer (Fogg and Nass 1997). Based on such findings, CASA researchers concluded that even the simplest cues to humanness, like the use of voice and language, are powerful enough to elicit social scripts and trigger social behaviors grounded in interpersonal relationships.

Parallel development has occurred in certain CMC research to resurrect the relatively overlooked core component of human communication, perceived involvement and connectedness with others. In more faithful appropriation of its original definition, “apparent distance of the other” (Short et al. 1976: 157), researchers equated social presence with how strongly one feels as if he or she were “with” the communication partner and engaged in direct, non-mediated communication (e.g., Biocca, Harms, and Burgoon 2003; Lombard et al. 2000). Social presence from this perspective is viewed as a “transient phenomenological state” (Biocca et al. 2003: 469) and treated in a more holistic manner, beyond perceived warmth and sociability of the mediated partner.

Renewed emphasis on perceived involvement (social presence) has several implications for CMC research. First, it opens up the opportunities to explore social presence in relation to other constructs developed and tested in more traditional communication contexts, such as parasocial interaction (PSI; Horton and Wohl 1956) and transportation (Green et al. 2008). For example, Hartmann and Gold-hoorn (2011) differentiated long-term identification with and friendship toward media performers (parasocial relationship) from users’ illusory engagement in real interaction with media performers during exposure (i.e., PSI) and proposed EPSI (experience of parasocial interaction) scale to measure one’s temporary perceptions of how much a TV performer is aware of, attentive, and responsive to him or her. Although PSI deals specifically with the interaction that occurs only in the imagination of media users, in that both PSI and social presence center on individuals’ subjective experience of “immediate, personal, and reciprocal” interaction (Horton and Strauss 1957: 580), the literature on PSI can inform presence research and vice versa. Likewise, transportation, which refers to the extent to which one creates vivid mental images of the settings and characters in a mediated experience and develops strong feelings toward the characters (Green and Brock 2000; Green et al. 2008), seems to overlap with social presence, since it also focuses on the “as-if” feelings of direct experience in mediated interpersonal encounters even in the absence of audio-visual cues. Recent findings that the proclivity to experience transportation into a narrative world (i.e., transportability) positively predicted the level of social presence one experienced while reading a politician’s Twitter messages (E.-J. Lee and Shin in press) supports this conjecture.

Second, focusing on perceived involvement and interactivity also helps to reframe the questionable assumption that social presence is inherently positive. When social presence refers to how closely one feels he or she is “with” the partner, there is no reason to expect that increased social presence must yield only positive relational outcomes. Rather, social presence is likely to facilitate either positive or negative effects, depending on characteristics of the source or the context, by increasing the cognitive and/or sensory salience of the mediated partner. In support of this speculation, although the participants’ prior attitudes toward a politician significantly affected the extent to which they agreed with his policy statements expressed in his tweets, social presence significantly amplified such effects (E.-J. Lee and Shin in press).

Third, the holistic approach to social presence prompts CMC researchers to adopt a less medium-centric view in exploring its antecedents. With social presence referring to how human-like the partner is mentally represented, as an object in the cues-filtered-out tradition, little has been discussed as potential predictors or determinants of social presence, except for multimodality. However, the extent to which people feel as if they engage in one-to-one conversation with a remote partner can vary depending on a host of factors other than technological attributes of the medium, such as purpose of communication, message content, relationship with the partner, and social context. For example, Walther and Bazarova (2008) found that the feeling of being connected with remote CMC partners (i.e., electronic propinquity) varied as a function of not only the channel bandwidth, but also the availability of richer media, information complexity, and the user’s communication skills. Similarly, in E.-J. Lee and Jang’s (2013) research, remediation of politicians’ Twitter messages in a news article weakened social presence, albeit only among those lower in affiliative tendency, even though the number and the kind of cues afforded by each medium (Twitter vs. news article) did not differ. These findings suggest that factors beyond channel bandwidth predict and explain social presence and further define and determine how communicative experiences are construed by their participants as a result of a sense of being together.

3.1 Foundations

The social information processing (SIP) theory of CMC (Walther 1992) articulated a new set of assumptions about what people do when they communicate using different channels, and how they respond to a reduction of cues in CMC. The theory posits that communicators adapt to a restriction in nonverbal cue systems by using whatever communicative codes remain in order to accomplish interpersonal goals. The theory also strove to account for previous CMC research findings that had supported the social presence approach, that is, to show how these earlier studies found what they found, but how the limitations and restrictions of their research methods had occluded the potential of CMC to exceed the impersonal qualities those experiments attributed to it. Alternatively, the theory seeks to explain how, with time, CMC users are able to accrue impressions of and relations with others online which achieve the level of development that is expected through offline communication.

The theory articulates several assumptions and propositions that propel these effects. First, it explicitly recognizes that CMC is devoid of most nonverbal communication cues that accompany FtF communication. Unlike other theories of CMC which argue that the lack of nonverbal cues impedes impressions and relations, or reorients users’ attention to impersonal states or to group-based forms of relating, this theory assumes that communicators are motivated to develop impressions and affinity regardless of medium. It further proposes that when nonverbal cues are unavailable, communicators adapt their interpersonal (as well as instrumental) communication to whatever cues remain available through the channel that they are using. Thus, in text-based CMC, the theory expects individuals to adapt the encoding and decoding of social information (that is, affective, socioemotional, or relational messages) into language and the timing of messages. Although many have interpreted this argument to refer to emoticons (typed-out smiles, frowns, and other faces; Derks, Bos, and von Grumbkow 2007), the theory implicates language content and style characteristics as more primary conduits of interpersonal information.

A second major contention of SIP is that CMC operates at a different rate than FtF communication in terms of users’ ability to achieve equivalent levels of impression and relational definition. Because verbal communication with no nonverbal cues conveys a fraction of the information of multimodal communication, communication functions should require a longer time to take place. CMC users need time to compensate for the slower rate in order to accumulate sufficient information with which to construct cognitive models of partners, and to emit and receive messages with which to negotiate relational status and its nature.

With respect to the first major theoretical contention, research demonstrated that communicators adapt social meanings into CMC language which they would otherwise express nonverbally. For example, Walther, Loh, and Granka (2005) had dyads discuss a controversial issue FtF or via real-time computer chat. One dyad member increased or decreased his or her friendliness toward the other individual at the researchers’ prompting. Recordings of the FtF confederates were analyzed for the kinesic, vocalic, and verbal behaviors and their correspondence to variations in friendliness ratings. A number of vocalic cues provided the greatest influence on relational communication, followed by a group of specific kinesic behaviors; the confederates’ language had no significant influence on perceptions of their friendliness. In contrast, in the CMC transcripts, several specific language behaviors had significant associations with differences in friendliness, and CMC confederates were rated no less friendly than those in the FtF condition. A more recent study found that CMC users transmit liking and disliking of their partners by praising or denigrating their partners’ preferences with regard to a topic (Walther et al. 2010). These two studies suggest that CMC users convey their attraction to others through agreement with partners, word choice, the way they compliment or contest others’ views of things. Whereas FtF communicators use nonverbal cues to express liking and attraction, CMC users appear to use a different form of indirect expression – statements that converge or diverge from partners’ opinions.

The SIP theory is somewhat more equivocal about the second major element, the temporal dimension. The primary theoretical explanation for the additional time CMC requires for impression development and relational management is that electronic streams of verbal communication without nonverbal accompaniments contain less information than multimodal FtF exchanges. Even in so-called real-time CMC chat communication, cues are not fully duplexed in terms of seeing a partner’s reactions at the same time they generate an utterance. Consequently, more time may be needed for relational effects to accrue in CMC because CMC often involves asynchronous media. Both approaches have been used in empirical research: Recent studies have added support for SIP using strictly asynchronous communication (Peter, Valkenburg, and Schouten 2005; Ramirez et al. 2007), or using real-time chat episodes repeated over several consecutive days (Hian et al. 2004; Wilson, Straus, and McEvily 2006).

3.2 SIP theory today

SIP theory has been expanded by researchers other than its original developer to incorporate other media than text-based CMC, although these formulations are tentative. Tanis and Postmes (2003), for instance, established that the presentation of partners’ photos or the preinteraction exchange of biographies work equally well in instilling interpersonal expectations in CMC settings. Previously, SIP research has been more oriented to verbal exchanges such as CMC users’ biographical disclosures, attitudinal statements, and style. So it is noteworthy that photographs appear to function similarly to biographical text. A recent examination of uncertainty reduction processes via social network sites focused explicitly on the potential obsolescence of SIP theory in light of new media characteristics providing information aside from the interactive exchanges on which SIP traditionally focuses. Antheunis, Valkenburg, and Peter (2010) argued that social network sites provide an abundance of asynchronous and unintrusive biographical, multi-modal (pictorial) and sociometric information about other people. Therefore, they predicted, these alternative forms of social information would serve as the primary sources of uncertainty reduction about others, without need of recourse to interactive communication via text. However, results showed that, despite the appeal of these newer forms of information display, interactive communication contributed the most to uncertainty reduction about another individual. It remains to be seen whether exposure to another person’s profile, in addition to interactive communication, accelerates relationship development. It may simply be that more information propels relationship development faster than less information does, or that exposure to profile information allows users to delve deeper and probe further on that which was disclosed in profiles, achieving greater depth of questions and disclosures.

4.1 Foundations

The hyperpersonal model of CMC (Walther 1996) proposes that a set of concurrent, theoretically-based processes explains how CMC may facilitate impressions and relationships online that exceed the desirability and intimacy that occur in parallel offline interactions. The model follows four common components of the communication process to address how mediation by technology may affect cognitive and communicative processes relating to message construction and reception: Effects due to receiver processes, message senders, attributes of the channel, and feedback. The model has received a great deal of attention in the literature and extensions and revisions to the model have been proposed. Certain aspects of the model remain under-researched – such as the holistic integrity of the four sub-components as well as the effects of reciprocal feedback – although some progress has been made with respect to these issues.

4.2 Hyperpersonal CMC today

In addition to research that has supported or challenged the hyperpersonal model, a variety of extensions to the model have been made, including applications to new social technologies. The model has provided utility in the study of online dating and deception in online dating profiles, self-disclosure and interpersonal attributions, the development of adolescents’ identity through online interaction, and other settings.

Research on the reciprocal feedback process has led to a reconsideration of the fourth component of the model, in that several CMC studies have generated findings consistent with a behavioral disconfirmation rather than behavioral confirmation effect (see Ickes et al. 1982; Burgoon and Le Poire 1993). Behavioral disconfirmation takes place when one individual anticipates an unpleasant interaction with a target person, and in order to avert unpleasantness, over-accommodates to improve the person’s demeanor. The first study which appears to have witnessed these events was the Walther (2007) study described above, in which participants anticipated online communication with a high-school-aged loner, a college student, or a professor. When communicating with same-sex partners, despite pretest indications that the high-schoolers were the least-desired communication partners, male subjects who believed they were communicating with a male high-schooler showed greater editing and affection than with a male peer or professor.

Research has tested specific hypotheses about behavioral confirmation and disconfirmation in CMC (Tong and Walther in press). Although behavioral confirmation was specified in the original articulation of the hyperpersonal model, it has only now been supported with a direct empirical test. This research confronted two major issues. The first was whether CMC is sufficiently potent to transact the kind of interpersonal influence by one person to alter systematically another person’s demeanor. Most theorists of behavioral confirmation and disconfirmation suggest that nonverbal cues are not only primary (Rosenthal 1993), but may be indispensable (Ickes et al. 1982) for the conveyance of interpersonal expectancies, leaving doubt whether text-based CMC can render similar effects at all. The second issue concerned the factors that lead perceivers into behavioral confirmation versus behavioral disconfirmation: the valence of the expectation about the partner, and the perceived malleability of the expected characteristic. An experiment involved dyads in a real-time chat, in which one partner was led to expect that the target partner either (1) had a pleasant personality, or (2) was in a good mood, or (3) had an unpleasant personality or (4) was in a bad mood. Target partners were actually randomly assigned and naïve to the experimental conditions.

Coders rated the participants’ social attractiveness and sociability from the CMC transcripts. Consistent with behavioral confirmation, perceivers acted pleasantly and induced their targets to do the same when they expected a pleasant personality or a good mood, compared to the condition in which perceivers expected the targets to have a negative personality. When expecting their targets to have a potentially malleable bad mood, however, perceivers acted relatively pleasantly and induced a pleasant reciprocal response from the targets, engendering behavioral disconfirmation. The results demonstrate that the employment of text in CMC is sufficient to convey interpersonal expectancy effects such as behavioral confirmation and disconfirmation, and that perceived malleability and affective valence determine which process gets activated.

One of the most dramatic applications of the hyperpersonal model has been to virtual groups comprised of potentially antagonistic members. This field experiment applied the hyperpersonal model’s notions about CMC relations to premises of Allport’s (1954) intergroup contact hypothesis, which predicts the reduction of prejudice as a result of making friends with members of one’s outgroup. In this study (Ganayem et al. 2011), students from different Israeli colleges participated in an online, year-long course in six-member virtual groups, with two students each from Muslim colleges, religious Jewish colleges, and secular Jewish colleges. Through their long-term, online interactions, many participants’ prejudicial attitudes changed positively between the inception and completion of the course, among those with the most negative original attitudes toward other religious groups.

5.1 Foundations

At one point, the social identity model of deindividuation effects, or the SIDE model (Lea and Spears 1991) was one of the most dominant theories of CMC. Modifications to the theory in response to empirical challenges and changes in communication technologies that bear on the theory’s central assumptions appear to have accompanied a marginal decline in its popularity and scope. In certain contexts, however, it still remains the most parsimonious and robust explanatory framework for CMC dynamics.

The SIDE model, like others, considers the absence of nonverbal cues in CMC as a deterrent to the expression and detection of individuality, but the impersonal nature of CMC is not necessarily a hindrance to the development of meaningful social relationships online. Rather than leave users with no basis for impressions or relations at all, it predicts that CMC shifts users toward a different kind of social relation based on group-based perceptions of others as well as self-categorization of oneself as a member of a social group.

The SIDE model specifies two factors that drive online behavior (see for review Postmes, Spears, and Lea 1999; Postmes et al. 2001). The first factor is the visual anonymity that occurs when CMC users rely solely on text-based communication. When communicators cannot see each other, their awareness of inter-individual differences becomes significantly attenuated. Drawing on principles of social identity and self-categorization theories (Tajfel 1978; Turner 1987), the model argued that visual anonymity leads to deindividuation, but not in the sense of “dissolution of identity and diminished capacity for self-regulation” (Prentice-Dunn and Rogers 1980: 104) as in classical deindividuation theory. In the identity vacuum rendered by a state of deindividuation, the second major factor in the theory comes into play, causing a shift in attention toward some salient social category or group (i.e., social identification), as opposed to the personal identity defined by one’s idiosyncratic characteristics left suspended by deindividuation. Submersion into the group identity predisposes CMC users to behave on the basis of intergroup dynamics in the form of in-group favoritism (Tanis and Postmes 2003), conformity to group norms (E.-J. Lee 2006), and social stereotyping (E.-J. Lee 2007; Postmes and Spears 2002). Such top-down categorization then drives users’ perceptions of online partners in gross terms, a deviation from the bottom-up process of uncertainty reduction through the painstaking gathering of personal information.

The model also specified, theoretically, that when a deindividuated CMC user oriented to personal identification rather than social identification, then the systematic group-based effects on similarity and attraction should not occur. The model views interpersonal attraction toward members of a group as an aggregation of randomly-distributed values a person attaches to each unique individual. That is, when perceiving others individually, the degree to which one is attracted to another fluctuates from person to person such that, on balance, it should average to some neutral level. Attraction to a group to which one belongs, in contrast, should be systematically positive, regardless of its members’ individual characteristics. This difference in the root of attraction marks a key distinction between a group-based and interpersonal approach to social dynamics of CMC (Lea, Spears, and de Groot 2001; see for review Walther and Carr 2010).

The most basic research strategy that provided evidence for SIDE involved experiments manipulating the two factors, visual anonymity and type of identity (see, e.g., Postmes et al. 2001; Spears, Lea, and Lee 1990). In a prototypical experiment, half of CMC users communicate with one another in small groups using a text-only chat system, whereas the other half would use the chat system and be shown photos or personal profiles supposedly representing the members. The former condition provides visual anonymity, presumably instigating deindividuation, whereas the latter condition evokes visual identification and individuation. The second factor, group identification, is manipulated by prompting participants explicitly either to look for the unique and distinctive characteristics of the group in which they are involved, or to try to detect what made the individuals with whom they were conversing unique and different from one another. Such research has produced predicted interaction effects of visual anonymity/identifiability by group/interpersonal identity, with conditions involving both visual anonymity and group identity yielding the strongest group-oriented reactions (Lea et al. 2001; Postmes et al. 2001).

SIDE model advocates originally argued that the nature of group memberships to which CMC users identified were comprised of fairly general social categories (e.g., English vs. Dutch nationalities, psychology vs. business majors, men vs. women, etc.). Attempts to arouse these kinds of identifications in SIDE experiments, however, have not consistently produced predicted in-group/out-group effects. Some studies found that only when the individual differences in the extent to which one identified with the focal group was taken into account, the supposed SIDE effects emerged. For example, preference for the same- versus different-university partner was observed when no individuating cues were present, in the form of portrait picture and biographical information, but such effects occurred only among those who identified more strongly with the in-group (Tanis and Postmes 2003, Study 3). Similarly, those who did not exchange brief personal profiles with their partner were more likely to exhibit stereotype-consistent conformity behavior than those who did, but only when they were strongly gender-typed (E.-J. Lee 2007). When identification was targeted only with the local group, that is, the unique and specific small group involved in the interaction, SIDE’s predictions appear to be more robust. These results have led to revisions of the SIDE model, and recent versions focus on visually anonymous CMC leading to ingroup identification with the group of participants rather than via larger social categories.

5.2 SIDE today

Once a dominant theoretical framework for CMC research, the SIDE model appears to be past its prime due to the uncertainties about the components of the model itself, empirical “competitions” in which social and interpersonal components both appear, and new media forms which alternately extend or restrict the scope of its domain.

Although both deindividuation, which visual anonymity was said to produce, and a salient social identity are key predictors of the SIDE effects, empirical studies have questioned the potency and theoretical necessity of these constructs. In particular, although it was long presumed to be a catalyzing factor in SIDE theory, when examined specifically deindividuation has not proven to be a causal factor in producing SIDE-effects. Indeed, research has suggested that the more critical factor is not whether one is unaware of one’s self, but rather, whether one can distinguish among other online individuals or perceive others as an undifferentiated group (Douglas and McGarty 2001). As a result, the deindividuation term has been replaced with depersonalization, which refers to the “reduced salience of interpersonal differences due to the deficiency of individuating cues” (E.-J. Lee 2006: 424–425).

The theoretical necessity of a salient social identity has also been called into question in generating otherwise SIDE-theoretic effects. For example, even when no shared group identity was assigned or highlighted, those stripped of individuating cues have been more likely to engage in spontaneous category-based person perceptions and group-oriented behaviors, challenging the notion that a salient group identity is a prerequisite for SIDE effects to occur. Specifically, those provided with no information about their CMC partners reported stronger in-group feelings with the partners and conformed more to the majority opinion than those exposed to brief biographical profiles of their partners, in the absence of specific instructions to focus on their group identity (E.-J. Lee 2006). Likewise, albeit not primed with a group identity, the lack of individuating information led participants to utilize linguistic gender cues to infer the anonymous CMC partner’s gender (E.-J. Lee 2007).

In addition, researchers have directly compared SIDE-based versus interpersonally-based factors in the same study for their effects on the responses of CMC groups. One study examined virtual groups comprised of students in England and the Netherlands who worked over an extended period of time via asynchronous conferencing and real-time chat (Rogers and Lea 2004). Steps were employed to maximize the salience of each virtual group’s unique identity (e.g., researchers addressed groups by their collective name only rather than individually by member). Repeated measures indicated that group attraction did not maintain evenly or increase over time. To the contrary, interpersonal affiliation among members reflected marginal increases over the duration of the groups’ experience. More recently, an experiment with visually-anonymous online groups involved a SIDE-based assignment of four members into two distinct sub-groups (Wang, Walther, and Hancock 2009). The researchers further prompted one member of each group to enact interpersonally friendly (or unfriendly) behaviors toward the rest of the members. In general, other members evaluated the deviants on the basis of the individuals’ interpersonal behaviors and not on the basis of their ingroup or out-group status. These results suggest that SIDE is less robust than previously suggested when CMC users confront bona fide behavioral differences among members while remaining visually anonymous. A recent essay offers a more tempered view of when SIDE and other intergroup dynamics are likely to arise in CMC and when they give way to interpersonal dynamics (Walther and Carr 2010).

SIDE theorists themselves have suggested that perhaps CMC users get to know each other online over time, as individuals first – a process predicted by social information processing theory, described above – and then form group identifications (Postmes et al. 2005). Although social identity and personal identity have been treated as the opposite and mutually exclusive poles of a single continuum, a recent study indicated that personalized, rather than depersonalized, communication may facilitate group identification. In E.-J. Lee and Oh’s (2012) experiment, participants viewed a fictitious politician’s Twitter page, which contained either personalized or depersonalized messages. The message content was virtually identical, but the otherwise plain policy statements were couched in the candidate’s personal experience in the personalized condition. While those with stronger party identity rated the in-group (vs. out-group) candidate more positively and expressed higher levels of intimacy with him, those lower in party identification exhibited such in-group favoritism only in response to the personalized messages. Possibly, for those to whom group identity is not chronically salient, personalized messages might have temporarily heightened the salience of the source and associated characteristics, including the candidate’s party identity, thereby activating in-group favoritism. Albeit speculative, this seemingly ironic possibility that message personalization triggers group-based reactions from less group-oriented individuals deserves further investigation, as it prompts the question of how deindividuation (depersonalization) affects person perception and social judgments when group identity is not salient, the undertheorized half of the SIDE model.

Besides Twitter, other social media have also presented challenges for SIDE. Many of the most contemporary Web 2.0 applications involve both groups and visual cues about participants. In the most traditional SIDE approach, the presence of visual identification should terminate the applicability of the SIDE model. Yet there are open questions, as Walther and Carr (2010: 220) suggest:

Future research on the SIDE model of CMC, or reversion to its antecedent theory, social identity theory, offers interesting approaches to emerging questions such as these.