Introduction

The emergence of the information society has created some exciting theoretical and empirical challenges for social scientists. There is a need to understand the nature and form of the interaction patterns that are peculiar to its definition and to determine their consequences for the development of self conceptualization and social structure. It is also important to determine the ways in which informational modes of production emulate and contrast with industrial forms of social organization. A common feature of the information society landscape is the massive utilization of personal computers as a means for transacting personal and interpersonal communications (Computer Mediated Communications/CMC). The computer is quickly establishing itself as a primary conduit for human beings to carry out communications with others as well as with self. This digitized form of interaction is radically different than face-to-face encounters associated with previous historical epochs and may have deep ramifications for social and individual epistemology.

Human beings are fundamentally symbol makers. Symbols are the means by which phenomena are known and the mechanism for generating social and psychological realities. The ability to interpret the symbolic gestures of others is what allows humans to interact. Body gestures, vocal intonations, language, grunts and groans are the fabric of symbolic communication. Conventional modes of learning and transmitting these symbols have been through face-to-face (FTF) interaction. CMC, however, represents a significantly different communication architecture whose structure may ardently modify patterns of interaction that is common to industrial methods of social intercourse. As Couch (1992) has suggested, information technology creates the potential for new forms of social relationships that are not merely ephemeral social forms but have genuine consequences for describing a new social history that departs from the hegemony of industrial methods of social action. However, such sweeping propositions must be examined empirically. An initial step would be to determine if social patterns within new information technologies such as CMC are indeed different than traditional FTF communications common in industrial society. If differences are revealed, such differences could be seen as an indicator of historical transformation that is influenced by a new technology such as CMC.

The purpose of this paper is to assess some of the contours of this emerging interaction landscape utilizing the framework of symbolic interaction. Our research project focuses specifically on Internet relay chat room communication (IRC). This is a group, mass communication system in which users send and receive text-based messages. The time delay of these computer- mediated messages can be nearly instantaneous or "real time" (December, 1996). Because the process of symbol exchange in chat rooms differ substantially from the symbolically rich context of face-to-face communication from which the conceptual frameworks of symbolic interaction has been developed, we are concerned with the following broad questions:

1. To what extent are symbolic interaction frameworks conducive for analysing communication processes within a computer mediated environment?
2. In what ways does symbolic interaction as an analytical framework increase our understanding of emergent social processes and structures within electronic chat rooms?

Social Psychological Research of the Internet

The use of the Internet has increased exponentially over the past five years. By the end of 1997 more than 100 million people were using the Internet, and traffic on it is doubling every 100 days (The Emerging Digital Economy, 1998). Parallel to this expansion in utilization has been substantial changes in internet communication technology. Usenet newsgroup discussions and electronic mail were the predominant communication technologies during the earlier days. Today, the Internet offers a much wider scope of tools used for information transmission and retrieval, communication and interaction (December, 1995). These technologies have created the capacity to enhance both synchronicity, asynchronicity and interactivity in CMC (Newhagen and Rafaeli, 1996). In short, the extensive utilization of the Internet along with its hypertextual communication architecture has created a genuine need for conducting social science research and for assessing its relationship with human interaction patterns and psychological phenomena.

Indeed, recent research has shifted attention towards the social and psychological factors of CMC. Researchers have examined such things as how online use emerges into a community (Marcus, 1987, 1990) and media selectivity of organizational communication (Daft and Lengel, 1984, 1986; Rice and Shook, 1990; Daft, Lengel and Trevino, 1987, 1983). Research has also been conducted on the social context of CMC ( Feenberg, 1989, 1992; Fulk, Schmitz and Steinfield, 1990), social cues (DeSanctis and Gallupe, 1987; Kiesler, 1996; Culnan and Markus, 1987) and the formation of personal relationships (Parks and Floyd, 1996; Pool, 1983; Rheingold, 1993; Stoll, 1995).

The communication discipline has provided major players in the development of theories and research methodologies applicable to Internet social dynamics. Curiously absent are studies promoting the sociological imagination. We concur with Denzin's (1995) paraphrase of a statement made by Couch (1995) regarding sociological abstinence in this area – "The maturation of the electronic media is opening `new opportunities for humanity and social scientists are failing to note' this fact. They do so at great peril. This maturation of the media has `created a void in the academy that has been partially filled by the emergence of departments of communication.'" Our challenge is to speak to this void.

Research Issues within the Framework of Symbolic Interaction Theories

Based on the work of George Herbert Mead (1934), the basic tenets of the symbolic interactionist perspective, as it relates to human communication and self-conception, are:

1. Humans interact by perceiving and interpreting the symbolic gestures from others.
2. Through a process called "taking the role of the other," people are able to anticipate each other's responses based upon the richness of the symbolic information existing within a given social situation.
3. In defining the nature of a situation, individuals are capable of orienting themselves to a vast constellation of social and physical objects. They internally rehearse appropriate lines of action or interaction towards the other, while suppressing perceived inappropriate actions, a process Mead termed imaginative rehearsal.

Within this context, a very salient factor affecting communication efficacy is the richness of the information cues. The vast majority of the conceptual and methodological symbolic interaction schemes have emerged from face-to-face communications in which the informational landscape in comparison to computer mediated communication is much more concrete. CMC differs significantly from FTF in the following ways:

1. Nonverbal gestures such as facial, body, and body posture are missing as informational cues for defining the situation.
2. Verbal cues including voice tone, voice quality, voice modulation, and intonation are frequently absent from CMC but present within FTF.
3. Information related to the social and personal characteristics of the individual such as gender, physical appearance, and status are substantially suppressed within CMC.

Theories such as social presence theory (Rice, 1993; Rice and Love, 1987; Short, Williams, and Christie, 1976) and social context cues theory (Sproull and Kiesler, 1991) predict that the absence of informational cues within the CMC context would lead to more impersonal and nonconforming behaviours in comparison to FTF communications. They further predict that online communication would be less inhibited and that individuals would be more inclined to exhibit aggressive behaviours. These predictions have been somewhat supported empirically (Kiesler and Sproull, 1992; Dubrovsky, Kiesler, and Sethna, 1991; Siegel, Dubrovsky, Kiesler and McGuire, 1986; Parks and Floyd, 1996). Hostile and aggressive behaviours expressed in CMC, often referred to as "flaming," have been reported in studies by Hiltz, Turoff, and Johnson, 1989; Lea, O'Shea, Fung and Spears, 1992; Stoll, 1995; Beninger, 1987; and Berry, 1993; each make claims that only illusions of a stable social structure is possible within online "communities."

These claims, however, are being challenged by other studies that indicate that although the reduction in informational cues may slow communication down, it does not necessarily undermine interaction outcomes such as the formation of friendships. They further suggest that positive interaction outcomes do occur, but simply take more time to emerge within CMC as compared to FTF channels (Walther, et al.). A recent study by Parks and Floyd (1996) on the development of personal relationships through Internet discussion groups indicates that such relations were common and that the breadth and depth of these interactions were moderately high.

Missing from all of these studies is an attempt to systematically explore the factors that explain the differences in CMC interaction patterns such as flaming. For example, what explains the variance among hostile or non-hostile behaviours and to what extent do social psychological characteristics of the individual explain such variance? These are important empirical questions that must be systematically analysed for an enhanced understanding of CMC social and behavioural processes. We believe that symbolic interaction frameworks would be a heuristic guideline for delineating possible factors to be examined. We have chosen the sociology of emotions as our theoretical platform (Hochshild, 1979; Shott, 1979; Heise, 1977; Ridgeway, 1982; Stryker, 1987; Kemper, 1991; Turner, 1994). The basic tenet of the symbolic interaction approach to the study of emotions is that they shape the flow of interactions. Emotions are conceived as sociologically relevant phenomena because particular types of emotions, as expressed towards others, are moderated by situational and normative constraints. The absence of concrete symbols within a chat room represents a relatively unique situational context in which the impact of cultural or normative expectations on behaviour is relatively unknown. However, people do not enter chat rooms as a blank sheet of paper. Rather they take with them internalized social rules for interacting with others as well as experiences in defining face-to-face situations that are used to approximate the nature of chat room situations. Our task is to determine if emotions that are theorized to be socially relevant towards explicating behaviours in so-called normal face-to-face communication will help to explain behaviours within chat rooms. We will utilize the sociology of emotion theory by Shott (1979) as the conceptual platform for this study.

Shott's Social Control Theory

Susan Shott's role-taking theory is based upon the idea that emotive actions are influenced by situational definitions and social norms. The process of role-taking is the focus of her conceptual scheme. "Much role-taking is reflexive in that the individual has an internal conversation with self as an object, seen and evaluated from the perspective of specific and generalized others. In this evaluation process emotions are aroused and labeled; and if these emotions are negative, they mobilize the individual to adjust behaviour" (Turner, 1997).

Shott proposes two role-taking sentiments that serve as controllers of emotions during the role-taking process:

Reflexive role-taking emotions that are directed toward oneself. These consist of:

Guilt

Using Ausubel's (1955) definition, she states that guilt is "...the feeling that accompanies the negative self-evaluation which occurs when an individual acknowledges that his behaviour is at variance with a given moral value to which he feels obligated to conform."

Shame

Shame occurs when after taking the role of the other, one discovers that the other's perception of the behaviour is not congruent to her/his idealized image of self.

Embarrassment

Is a feeling that exists "when an individual's presentation of a situational identity is seen by the person and others as inept" (Turner, 1998).

Pride

A person experiences pride when through taking the role of others they obtain self-approval. Prideful persons attempt to present themselves to others in such a manner to maintain this self-approval.

Vanity

Shott's description of vanity is "Vanity, unlike pride, is an unstable and transient emotion; it is the form social approval may take when one is not sure of one's self image or the approval of others. Vain persons are therefore more immediately dependent on others for their self-conceptions" (1979; p1326).

Empathetic role-taking emotions

Shott conceives empathy as "The arousal in oneself of the emotion one would feel in another's situation."

Our objective in this paper will be to operationalize each of the concepts in Shott's theory summarized above and then apply them to assess the degree to which emotion management occurs when people are expressing their feelings and exhibiting various behaviours within a CMC chat room environment. We will also analyse the extent to which such emotion management is influenced by gender, locus of control, sociability, and perceptions of the normative context of chat room environments.

Research Questions

This paper will explore the following research questions:

1. To what extent does the normative orientation of the individual influence the sending of hostile and aggressive messages in Internet chat rooms? Normative orientation is conceptualized within this study as the user's perception of whether or not chat rooms have an unwritten standard of conduct. We expect to find an inverse relationship between our normative variable and hostile communications such as flaming.
2. To what extent does the locus of control (LOC) of the individual influence the sending of hostile communication messages in chat rooms? Locus of control (Rotter, 1966) is a personality orientation variable which delineates how individuals attribute outcomes related to their actions. People who see themselves as being able to "control" events of their actions are referred to as internals. Those who are more oriented towards believing that events are outside of their control are characterized as being externals. We expect to find a negative correlation between LOC and the hostility interaction variables.
3. To what extent does the sociability of the individual influence hostile type communication in chat rooms? We anticipate a negative correlation between sociability and the hostile communication variables.
4. To what extent does gender moderate the relationship between chat room users emotional orientation and their tendency towards sending hostile messages in chat rooms? To what extent does the emotional orientation of the individual influence his/her tendency towards exhibiting hostile and aggressive behaviours in Internet chat rooms.

The degree and manner in which emotive factors influence individual behaviour within a social situation is contingent upon how that situation is defined. In order to investigate the relationship between emotion magnet and chat room communication more systematically, we will also examine the moderator influence of cybernorm, sociability, and locus of control on this relationship. Similar to gender and ethnicity, each of these variables may influence how the individual defines social situations which would in turn influence the relationship between emotion work and communication. As such, we expect the relationship to vary in terms of direction and strength across categories of each of the moderator variables.

Methodology

Participants

Subjects (N=114) were selected from a large, Midwest university that has more than 25,000 students in approximately 180 fields of study. The diversified student body represents both full-time and part-time students, with women comprising almost 60 per cent of the student population.

The study was conducted during April and May 1998. Most respondents were undergraduates enrolled in the basic studies course, Understanding Technology, which is one of several courses that fulfill the university's science requirement. Although this is a freshman-level course, students of all levels take it. All respondents in this course who indicated that they participate in online chat rooms were selected for this sample. Students in this course are entering a wide variety of disciplines throughout all the colleges of the university. Others were graduate students in the Master of Liberal Studies in Technology program. bellamy_chart1.gif Chart 1presents the demographic structure of the study.

All subjects were given their questionnaires during class time. Questionnaires were completed by the students on a voluntary basis. All students who stipulated that they used relay chat rooms agreed to complete the questionnaire. Thus, our study consists of a 100 per cent sample of the chat room users from a selected subpopulation of the university.

To assess whether a subject's online behaviour is influenced by emotive factors, Schott's six emotion management factors were measured by the following six Likert-type questionnaire items consisting of five scale points with anchors ranging from Agree to Disagree:

Guilt

"I feel guilty if I say something to offend someone in a chat room."

Shame

"I feel a sense of shame when someone in a chat room points out to me that my messages are inappropriate."

Embarrassment

"There have been times that I have felt embarrassed in a chat room because of how I presented myself."

Pride

"It is important for me that people see the best side of myself."

Vanity

"I am the type of person that does not need to get approval from others in chat rooms."

Empathy

"When people discuss their problems with me in chat rooms, I am able to feel what that person is feeling."

Sociability of the respondents was measured by a seven-item Likert-type scale developed by Hanewicz and Bellamy (1998) consisting of five scale points ranging from Agree to Disagree. It revealed an alpha reliability of .81. (See Appendix A for the entire scale.) Scores ranged from 7 to 35 with a median of 27. For categorical analyses, scores of 27 and above were designated as high sociability (n=53); scores below 27 were delineated as low sociability (n=57).

Locus of control was measured by a ten-item scale developed by Burger (1986) which consisted of seven scale points ranging from strongly agree to strongly disagree. Scores ranged from 32 to 64 with a median of 48.79. The alpha reliability coefficient on this scale for this population was .61. (See Appendix A for the entire scale) When used as a categorical variable, scores higher than 49 were designated as high locus of control (internals) and scores lower than 50 were deemed to be low locus of control (externals).

User perception of standards of conduct within chat rooms was measured with the following item "I believe that there is an unwritten code of conduct that people must follow in chat rooms." We will refer to this variable as cybernorm. When used as a categorical variable, individuals who selected agree or slightly agree were delineated as being high cybernorm, while individuals who chose neither agree/nor disagree and below were delineated as low cybernorm.

The interaction variables and their measurement are as follows:

Flaming

"I send `flaming' (hostile) messages."

Hostility

"I am more likely to send hostile messages in chat rooms than in face-to-face communications."

Displaying anger

"It is more appropriate to display anger in chat rooms than in face-to- face communication."

The cybernorm and interaction variables utilized a five-point Likert-type scale with response possibilities ranging from Agree to Disagree.

Results

Our first task is to determine the zero order relationships between the interaction variables, emotion management, cybernorm, locus of control, and sociability. The findings presented in bellamy_table1.gif Table 1 reveal that the independent variables are more closely associated (in terms of the expected inverse relationship) with the flaming communication variable than those of hostility and anger. The largest statistically significant relationship is found between cybernorm and flaming. This is a very salient finding when considering the large attention currently given towards the flaming phenomena. The negative relationship indicates that when people define the chat room situation as having normative standards of conduct, flaming behaviour is reduced. This supports our initial expectation that this type of defining of the situation would circumscribe flaming behaviour.

A relatively weak (although not statistically significant) inverse relationship is shown between flaming and locus of control which indicates that internal oriented individuals have a tendency to engage less in flaming communications than persons with an external orientation. A weak inverse relationship between sociability and hostile communications is revealed. The relationship does not have statistical significance.

(A bellamy_matrix.gifcorrelation matrix showing the relationships between all of the variables within this study is presented in Appendix B). Our next task will consist of determining if different statistical patterns are revealed within the categories of gender, ethnicity, cybernorm, locus of control, and sociability.

The Moderator Influence of Gender

As revealed in bellamy_table2.gif Table 2, cybernorm once again has the strongest and most significant relationship with the flaming variable. However, the correlation is stronger for males than for females which implies that the male's normative orientation towards chat room communication is more circumventing of flaming messages as compared to females. Overall, even though most of the correlations revealed are not statistically significant, this pattern is repeated among most of the relationships. Thus, gender does indeed moderate the relationship between the independent and criterion variables within this study. However, the nature of this influence is somewhat unexpected, particularly in regards to the emotion management factors. These factors appear to constrain flaming communication more among males than females which is counter to culturally held beliefs that women are more emotive oriented than males. These findings suggest that women are defining and utilizing electronic communication platforms in different ways than men.

The Moderator Influence of Cybernorms

Does the manner in which individuals define the chat room situation in terms of its containing normative standards of conduct influence the correlation of the communication variables with locus of control, sociability, and emotion management? The correlations presented in bellamy_table3.gif Table 3illustrate that although there are few statistically significant relationships found among these variables, overall, cybernorm does affect these relationships. We anticipated that stronger inverse relationships between the communication variables and the emotion management factors would be found among individuals with high normative perceptions of chat rooms in comparison to those with low normative perceptions. For the flaming variable, this expectation is supported only for the shame and guilt factors and for the guilt factor in relation to the sending of hostile messages.

For individuals with a high normative orientation in comparison to those with a low normative orientation, there is a slight tendency to send fewer flaming messages and to express anger less frequently as the locus of control moves towards that of internal orientation.

Locus of Control

In analysing the results in bellamy_table4.gif Table 4 we see definite differences among the relationships between LOC categories. For both high and low LOC groups, shame and guilt are shown as factors that control flaming interactions in predicted ways. This is more pronounced for externals on the guilt factor. A very weak relationship between pride and flaming is shown among externals in comparison to a relatively strong correlation among externals. This seems to be indicating that pride is more operative as an emotion control strategy within chat room environments for individuals that may have less self-esteem than others. Although LOC does not directly measure self-esteem, it has been shown to be highly correlated with this construct esteem. In short, there appears to be a stronger need among externals who may have less confidence and self-esteem to project a positive image in chat rooms, and this need may explain why pride is more of an influencing factor for constraining flaming behaviour among externals as compared to internals. This statistical pattern, however, is not revealed for the hostility and anger variables. Among these variables, empathy is a much stronger constraint for hostile behaviour for internally oriented individuals than externals.

Sociability

The results presented in bellamy_table5.gif Table 5 once again illustrate a strong negative correlation between flaming and cybernorm. This is the case for both high sociable and low sociable individuals. Interesting enough, the highest correlation between these two variables is revealed among the low sociable people. Weak correlations are shown between cybernorm and the other communication variables. The same is true for locus of control and the communication variables. Among the emotion management factors, the highest negative correlation is found between the guilt factor and flaming among high sociables. A moderate negative relationship exists between pride and flaming among low sociables.

Discussion

Both Hochshild (1975) and Shott (1979) suggest that the patterning of affective experience is profoundly influenced by the individual's cultural experience. "Members of some segments of a society tend to feel certain emotions more often or more intensely than members of other segments because their position in the social structure subjects them more frequently to certain types of experiences" (Shott, 1979, p1318).

The differences found in the correlations between the interaction and emotion variables according to categories of gender strongly allude to the possibility that males and females are defining the CMC situation in different ways. This subsequently suggests differences in role-taking and cognitive rehearsal processes which further insinuates the possibility of different epistemological outcomes for each of these groups as they engage in virtually real social processes. Issues concerning the epistemological ramifications of cyberspace social interactions and its impact on such things as role identity and identity change, have been seriously considered within the framework of symbolic interaction theories (Kiecolt, 1994; Marcus and Wurf, 1987; Fein, 1990). Further research should be conducted to determine the different ways in which gender groups are utilizing cyberspace as platforms for creating and enacting new role identities (Rheingold, 1993). This study has illustrated that although behaviours such as flaming do indeed occur in CMC, its occurrence is modified by social psychological factors. This is a very meaningful finding because it provides additional information to the claims made by previous studies implying that the anonymous nature of CMC lifts normative constraints commonly associated with face-to-face communication gives impetus for sending hostile messages such as flaming. Our study implies that there is not a direct linear association between technology and behaviour in the manner in which it has been described by previous studies. This paper has shown that a very significant factor influencing the type of behaviour expressed in CMC consists of the extent in which chat rooms are defined as situations that have normative standards of conduct. Indeed, the negative and relatively strong correlation between cybernorm and flaming was consistent throughout all of the variable categories examined within this study. Although chat rooms can be correctly described as environments that contain very few concrete symbolic cues, human characteristics such as emotion, do indeed influence (similar to FTF situations) behaviour contingent upon the way in which individuals define the chat room situation.

It further implies that CMC technologies such as chat rooms should not be conceptualized or studied as causal factors but rather as a contextual variable that operates as a dynamic scaffold from which a complexity of social patterns may emerge. A crucial factor affecting such patterns consists of how individuals define the situation.

At this point in time, we are witnessing the emergence and contextual formation of chat rooms. As new technologies are added that will make the interaction more virtually real (Chayco, 1993) individuals will have frames of analysis (Goffman, 1974) that provide the ability to replicate traditional modes of interaction. Subsequently, it will be imperative to utilize more dynamic conceptual and methodological schemes for assessing the nature of the digital human condition.

As with any cross-sectional survey design, this study has definite internal validity weaknesses. Moreover, the generalizability of this study's findings is limited to a collegiate population of mostly undergraduate students. Nevertheless, we feel strongly that this investigation serves the highly useful exploratory purpose of mapping the landscape of human interaction within a digitized environment.