In order to ground this study of identity censorship an analytical framework that accounts for identity, anonymity, and self-presentation via computer mediated communication is required. An academic consideration of the field of identity is beyond the scope of this paper. Instead, I will cover a number of important themes from this area using two guiding questions: First, where does identity reside? and second, What are our identities composed of?

To answer these questions, I will take a historical perspective on identity by first accounting for a cultural understanding of the self as distinct from society. Next I will cover theories of personal identity that approach identity formation as result of individual psychological development. This will be contrasted with a more sociological approach to identity which focuses on the collection of group memberships by which an individual is defined. Next we will turn to Goffman’s ideas on self-presentation to consider how identity is performed. Following this, I will investigate the origins and operation of the attributes that compose our identities. Finally, I will examine the topic of anonymity, providing a working definition to guide this study.

Self-Presentation & Identity

Because digital identities seek to represent some aspect of the individuals they represent, it is important to consider some of the key academic contributions that have shaped our understanding of the often ambiguous term “identity”. Lay definitions generally consider identity as the means by which an individual distinguishes himself as unique from others. Asking someone to describe their identity, however, typically results in a list of personal attributes — specific characteristics used to both compose and distinguish one’s identity.

The Self and Society

Starting with Rousseau’s walk in the forest, our understanding of identity and subjectivity has relied on the distinction between “the self” and “society”. As Gutman described in his account of Rousseau’s Les Confessions:

In order for a man or a woman to be constituted as a subject, he or she must first be divided from the totality of the world, or the totality of the social body. For ‘me’ to emerge, a distinction must be made between the ‘me’ and the ‘not-me.’ The boundaries of the self are those lines that divide the self from all that which is not the self, which is beyond the self. (Gutman, 1988, p. 107)

In the field of Linguistics, Emile Benviniste asserted that the self could only emerge in the presence of another. For Benviniste, this distinction was a function of personal pronouns: “I posits another person… to whom I say you and who says you to me” (1986, p. 729).  The subjects I and you, once established are then available to house our contemporary notions of identity.

Personal Identity & Social Identity

In the 1900s, identity scholars were considering the development of this autonomous self, particularly in relationship to his surroundings. Mead (1925), for example, proposed three stages during which self-identity develops in response to social experience and environmental factors. He theorized that children develop an autonomous sense of self (what he calls the “I”) as well as an understanding of the self that is governed by social rules and external expectations (Mead’s “me”). He argued that identity was role-oriented, and emerged out of interpersonal interactions he called identity negotiation.

Psychologists following psychoanalytic and human development traditions believed in an environmental influence, but maintained that identity was a component of the psyche. In Erikson’s model of psychosocial development, for example, the development of personal identity and social identity are aspects of a lifelong process of identity formation (Levine, 2002). Erikson was the first psychologist to formally theorize human development as a process that continued through life, and not just through Freud’s genital stage. For Erikson, a healthy identity (he used the term ego identity) required a stable and contiguous self. Erikson’s work embodies many of the assumption to which our identities are held today. Namely, that an individual is responsible for their identity, a stable identity is ideal, and that while we play various social roles, and these roles are aspects of a single, and ideally integrated, self.

In contrast to theories of personal identity, sociologists have produced theories that approach identity based on group membership. Henri Tajfel and John Turner’s social identity theory (1979) is perhaps the best known, identifying four elements involved in our group identification: categorization, identification, comparison, and distinctiveness.  This theory does not reject a personal approach to identity, but it does consider the personal attributes which are shared across members of a social group. Tajfel & Turner argued that we categorize each other into these social groups, holding a positive bias for those groups to which we belong. Any behavior can then be plotted on a scale between interpersonal and intergroup. When individual characteristics are the most salient, we will interact with others on an interpersonal level. When our group identity is more prevalent, however, group values and rules govern our actions.


Social rules and identity as a performance were critical components in Goffman’s text The Presentation of Self in Everyday Life (1978). Introducing the now prolific ideas of a “front-stage” (public) and “back-stage” (private) self, Goffman examined the different ways in which individuals perform when interacting with a group, as compared to when they are alone. Goffman detailed a number of strategies by which people actively shape their front-stage selves, conforming to social expectations as they selectively self-represent for social advantage. Self presentation, however, is not fully in the control of the individual. Goffman distinguished between two types of information communicated during self-presentation: expressions intentionally “given” by the individual, and unintentional expressions unknowingly “given off”.  Given that individuals perform in a variety of social contexts, Goffman believed that an individual may develop a number of front-stage selves. These identities, however can also become institutionalized as others come know a particular presentation and begin to expect behavior in line with that identity.

Online, where the boundaries of public and private are easily traversed, we can select our environments and mediate contextual factors. Because our digital identities are influenced and limited by often static digital environments, our front-stage digital selves are always incomplete and selective representations of the individual sitting behind the computer. The digital identity I assume during a bank transaction, for example, is far different from one I may assume in a social chat room.  Given the specificity of purpose assigned to our digital identities, these two contexts produce more than just different versions of my identity. They may produce identities that are completely distinct.

Identity Categories

Up to this point we have consider the origins and location of our identities, but have yet to explicitly consider the elements that compose our identity. This may appear a rather mundane task. After all, we compose out identities online each day. When signing up for a user account or updating an online profile we must type our names, indicate our sex, location, and perhaps answer a security question like “What was the name of your first pet?” Prompts such as these span the spectrum of identity attributes, from the personal to the group. While my gender associates me with a large social group, for the sake of my online security I hope that very few pets shared my childhood cat’s name.

Some theorists believed that our identities are the byproduct of attributes such as these. In On the Genealogy of Morals, Nietzsche claimed subjectivity emerges out of a complex interplay between language and power, a fact that Foucault seizes to further his theory of disciplinary regulation (1995). Foucault claimed that practices of disciplinary regulation arose in the years following the middle-ages as the unified Church-regulation of a now mobile and educated population broke down. As result, new disciplines derived from prisons, hospitals, and schools emerged to create manageable categories for the population. Categories such as these, always codified in language, provided a set of terms relative to which we can understand ourselves, and by which we are regulated.

Ian Hacking, in his text Mad Travelers (1998), provides one example of how classifications systems produce identities. Hacking traces the history of a mental illness named “dissociative fugue”, an archaic disorder which the DSM-IV describes as the “sudden, unexpected travel away from home or one’s customary place of work, with inability to recall one’s past.” He explores the ways in which classifications can emerge from sets of behavior in relatively small geographical areas, gain wide-spread support, and eventually fall out of use. However, once established, as Hacking points out, even nonsensical classifications can be hard to eliminate: dissociate fugue remains in the DSM today.  We see another example of medical identities in a “table of casualties” included in Bowker and Stars’ Sorting things Out (1999, p. 22-23). The table dramatically illustrates the various causes of death in 17th century England.The list ranges from the predictable “Aged” to the confusing “Suddenly”, but also includes categories as incomprehensible as “King’s Evil”. As a list, these causes of mortality illustrate not only the expansion of medical knowledge during this time, but also the ways in which the very definition of “casualty” can change while antiquated classifications persist.

If the classification of dissociate fugue can persist simply due to its continued presence in diagnostic manuals, then we should also account for identity classifications at the personal and interpersonal levels. Work by narrative psychologist Dan McAdams (1996) does this well. Bridging the linguistic and categorical, he highlights the ways in which identity categories are not only socially constructed, but also socially granted. His model of coconstructed identity suggests that while an individual may be able to relate to a particular identity category, the social and personal impact of an identity attribute only occurs after the attribute is integrated into the individual’s coconstructed identity narrative.

Having reviewed the literature on identity and identity classifications, online profiles beg the consideration of identity classifications in digital environments. Online profiles, after all, exist to provide structure to identity. Name, age, and the name of a childhood pet are only the beginning of the classification work in which digital identities engage. SNSs are rampant with personality quizzes promising your Meyers-Brigs type as well as the Marvel superhero with whom you are most similar.


How does anonymity alter this discussion of identity? Anonymity is a primary concern on the internet (Rigby, 1995; Wallace, 1999; Negroponte, 1998; Christopherson, 2007). Still, providing a definition of anonymity for the Internet can be complicated. Much like its counter-party “identity”, anonymity as a term is a moving target, always asking “anonymous to who?” Anyone who publicly claims that we are anonymous on the internet is quickly reminded of that their computer is always identified by an IP address. This misses the point. The phenomenological reality is that when we are online, we often feel anonymous.

Hayne & Rice (1997) highlighted this important fact by producing two separate terms for anonymity: technical anonymity and social anonymity. Technical anonymity refers to cases in which there are not sufficient cues with which to identify an individual, for example when someone leaves a blog comment without using their name or a user account. Social anonymity, on the other hand, refers to an individual’s subjective sense of themselves as unidentified, typically in a particular social space (for example, when visiting a foreign city). Hayne & Rice contend that even though an individual may be technically anonymous, this does not guarantee that they are socially anonymous. Users make assumptions based on the content available, although, as Hayne & Rice point out, these assumptions can be wildly inaccurate.

For the sake of this paper, anonymity must be considered for two different actors: the author and the reader. For the author, even though the body is absent, some portion of their identity is revealed to observers by the digital artifacts they leave behind. On craigslist, the shortest submission identifies the user to the website’s audience, even if only as a user of craigslist. Extending Hayne & Rice’s social anonymity, anonymity and identity represent opposite, unachievable ends of a spectrum. To be anonymous is not to be unidentified, but rather less identified.

Our second user type, the reader, is allowed to retain their anonymity relative to the post’s author. Pfitzmann & Kohntopp (2001), taking a technical perspective, argue that anonymity can only exist inside of a “set of subject with the same attributes” (p. 6) This helps clarify the questions of “anonymous to who?” by providing the conditions for anonymity. While both types of craigslist users may feel anonymous, only the different attributes for these user types (that one posts a missed connection, where another does not), allows the reader to retain his technical anonymity relative to the user community on craigslist.

Computer Mediated Communication

Regardless the origins or make-up of our identity, we increasingly share these identities online. Here, both the identity attributes we can share, as well as the technical means by which they are transmitted, play crucial roles in the types of identities that can be effectively communicated online. The choice of communication technology can have dramatic impacts on the types on information we intentionally and unintentionally “give off”.

In the 1980s, as computer-based forms of communication proliferated, social science scholars began to consider the impact of interpersonal communication over technology that prohibited nonverbal and paraverbal communication cues. Early work by Sproull & Kiesler (1986) suggested that CMC was less personal than face-to-face (FtF) communication. Their study of CMC over corporate email found greater levels of self-absorption, and a reduced importance of social hierarchy between the sender and receiver (e.g., workers were less likely to use a formal tone when addressing a boss). However, their research also showed that CMC resulted in lower levels of personal inhibition, and greater levels of information sharing that otherwise would not have occurred.

Meanwhile, a number of non-academic pieces (see Van Gelder, 1985; Dibbell, 1993) documented cases of CMC that were anything but impersonal. Rheingold’s work in virtual communities (1993) showed that even via restrictive text-based communication, participants can produce relationships that are equally deep as those developed FtF. One explanation that has enjoyed a good deal of success can be found in Walther’s Hyperpersonal Model (1996). Walther argues that even though CMC may reduce interpersonal cues, the slow rate of communication in a medium such as chat affords users the ability to edit and revise their speech acts, selectively representing an optimized version of oneself. Overtime, communication partners will develop media-specific equivilents to paraverbal and nonverbal cues that will offset any negative impacts of having loss FtF-cues. This hyperpersonal communication, Walther claims, may be more socially desirable than FtF communication under certain circumstances. A Missed Connection is an excellent example. Authors are able to carefully craft and revise their digital identities to their advantage, a feature not afforded in FtF communication. In fact, it may be the case that some users elect to author a Missed Connection rather than face the potential negative outcomes of a FtF interaction.

These findings are partially in line with the Uncertainty Reduction Theory (URT; Berger & Calabrese, 1975), a theory that originally spoke to FtF communication, but has been successfully applied to CMC as well (Ellison et al., 2006; Gibbs, Ellison, & Heino, 2006; Nowak & Rauh, 2005; Parks & Floyd, 1996). URT suggests that communication in new relationships is motivated by the unpleasant nature of interpersonal uncertainty. During the entry phase of a relationship, URT predicts that conversation will be dominated by information seeking behavior structured by the shared norms and rules of appropriate interpersonal behavior. This type of probing communication can be seen throughout CMC, but the quick intimacy experienced by CMC users may suggest that digital environments have an alternative set of interpersonal rules by which normative behavior is judged.

Regardless of the particular rule set by which CMC is judged, both the Hyperpersonal Modal and URT are best suited for considering multiple CMC interactions across time. The Social Identity Model of Deindividuation (SIDE) theory, on the other hand, has been successful at accounting for the absence of verbal and nonverbal cues, particularly between new or anonymous communication partners (Lea & Spears, 1992; Postmes, et al., 1998; Postmes, et al., 2002). SIDE argues that because nonverbal cues are not possible during CMC, users base judgments on the discernable social categories of online communicators. This is particularly salient when considering digital identities. Profiles are a form of interpersonal communication (Boyd & Heer, 2006), but they are not a part of an ongoing synchronous conversation. Without the time required for Walther’s hyperpersonal model to take effect, SIDE theory provides an alternative that suggests a reader’s impression of a user will be based on the social categories included in, and implied by, the digital identity.