Isn't That Spatial?: Distance and Communication in a 2-D Virtual Environment
Dean H. Krikorian,
(Ph.D., University of California, Santa Barbara) is an Assistant Professor in the Department of Communication at Cornell University. His research examines organizational communication, small group decision-making processes, and the Internet. He is director of the Cornell Communication Network Laboratory, which examines network communication patterns, particularly in online environments. His recent work examines the role of designing online virtual communities (i.e., Internet clubs) using group and computer-mediated communication principles.
Doctoral student in Department of Communication at Cornell University. His research interests include human-computer interaction and electronic commerce. He is currently involved in a major initiative examining Web commerce and the use of the Web as an information source. He holds a BS and an MS in Chemical Technology and has worked as a software programmer in the computer industry before earning his master's degree in Telecommunications from Michigan State University.
Ph.D. Candidate in Communication at Cornell University. She received her B.A. in Political Science, M.L.I.S., and M.A. in Speech from the University of Hawaii at Manoa. Her research focuses on the ways that people's interpersonal and intercultural experiences shape the processing of media messages. In addition to her interest in interpersonal interactions in a computer-mediated environment, she has been involved in a series of studies examining the ways in which people use their knowledge of what is typical in making reality judgments about media messages.
(M.A., Michigan State University, B.A., Iowa State University) is a Ph.D. student in the Department of Communication at Michigan State University. His main area of interest involves applying interpersonal communication theory and methods in computer mediated contexts. Specifically, his research looks at virtual environments through a micro-organizational framework.
Department of Communication, Cornell University, 305 Kennedy Hall, Ithaca, NY 14853, USA.
336 Kennedy Hall, Cornell University, Ithaca, NY 14853, USA.
336 Kennedy Hall, Cornell University, Ithaca, NY 14853, USA.
480 CAS, Michigan State University, East Lansing, MI 48824, USA.
This article examines the spatial relationships between avatars (i.e., graphical identities or icons) over time in a 2-dimensional online chat environment. The Spatial Distance Analysis Program (SDAP) was developed to measure the distance between avatars in a specially designed Palace environment. Correlations between distance and interpersonal communication constructs of (1) conversational appropriateness, (2) social attraction, and (3) uncertainty reduction indicate that distance effects are significant in an online environment. Specifically, it was found that general conversational appropriateness mediated between uncertainty reduction and specific conversational appropriateness for individuals who moved closer together and farther apart over time, respectively. Furthermore, the relationship between social attraction and distance indicated a significant positive parabolic function; that social attraction (i.e., liking) decreased at middle distances and increased at low and high distances. This finding suggests that there are three interpersonal distance zones in online communication.
Introduction: The Internet as a Communication Medium
The Internet has altered communication as we know it, allowing for new forms of interaction that were previously impossible. Internet Clubs, RealAudio, desktop videoconferencing, streaming video, and multi-dimensional graphical interfaces provide sophisticated tools for communication as we proceed into the 21st Century. Soon, such tools will be replaced by even more sophisticated technologies. Computer-mediated communication, as we know it today (circa 2000), is indeed vastly different than what we saw in 1994, 1996, or even 1998. New technologies are being developed even before old ones have had time to stabilize (and hence be examined). What was new is rapidly becoming old and without looking ahead, one falls behind.
Internet technologies may come and go, but their common ground is that they are all means to communicate. Whether it be one-way, two-way, or interactive (Rafaeli & Sudweeks, 1997), communication remains constant in all technological advances on the Internet. The Internet is a feedback mechanism which allows people to interact and share information. Electronic commerce, for example, is a means for buyers to find sellers online. It enables the communication linkage between two people, allowing a transaction to occur. From a technical communications perspective, one computer has a “handshake” with another computer and a transaction occurs. From a human communication perspective, a buyer finds a seller who offers Pokemon cards for much less on ebay.com than the corner store and a transaction occurs. Communication is at the root of the transactional nature of the Internet.
Problem: Spatial Orientation as a New Form of Online Communication
Throughout its history, computer-mediated communication (CMC) has long relied on a traditional set of assumptions regarding the channel of communication. Initially, traditional text-based messaging systems such as email, newsgroups, or bulletin boards (see Garton & Wellman, 1995; Parks & Floyd, 1996; Walther, Anderson, & Park, 1994;) predominated in studies of CMC. Such environments relied on typed, asynchronous messages. Chat provided real-time text-based messaging systems in an on-line environment as an advance over text-based, asynchronous communication (see Liu, 1999; Reid, 1991). Web pages provided still graphics, movies, and sounds that made multimedia applications more of a reality (see Hoffman, Novak, & Chatterjee; 1998; Jackson, 1997). Despite this progression, however, we argue that recent technological advances that incorporate graphical interfaces transform CMC into another dimension–that of the spatial element. This example can be clarified using traditional MUD's (Multiple User Domains) and MOO's (Multiple Object Orientations) (see Curtis, 1996; Reid, 1996; Turkle, 1995) which incorporated a spatial element, but did so using a text interface (e.g., GO TO CORNER, PICK UP KEYS). In today's graphical environment (imagine a videogame), one simply goes to the corner and picks the keys up using a personified image (or avatar). Distance becomes a factor in such an environment, whereas previous computer-mediated interfaces would not allowed for this to occur. This may be because of the demand for more sophisticated videogames, but the results are improved graphical interfaces which are reified via virtual environments (see Biocca, 1997; Biocca & Levy, 1995; Mantovani, 1996). Too often in CMC literature of the past the graphical interface was neglected because it was not sophisticated enough. Today, however, spatial orientation can no longer be ignored; it has come of age. Virtual environments reflect a new medium of online communication because they offer a new channel of nonverbal communication, that of spatial distance.
Purpose: Examining Communication and Spatial Distance
Our purpose in this study is to examine communication and spatial distance in an online graphical environment. Our goal is to look at spatial distances among avatars (i.e., graphical images or icons) in a 2-dimensional space, examining how online spatial orientation affects the communication constructs of uncertainty reduction, conversational appropriateness, and social attraction. After examining past research and describing the research design of this study, a tool is developed to measure distances among objects in a graphical space. This tool is made available as a means for researchers to measure distances in 2-dimensional space. Results relating the longitudinal distance measures and aforementioned communication factors are presented, discussed, and summarized. Finally, the overall implications and limitations of this particular study and future research directions examining spatial proximity and communication in a virtual Internet environment will conclude our account.
Literature Review: Distance Communication Constructs and Research Questions
Researchers have long been interested in the role of relative distance in human interaction. Traditional studies have found that seating arrangements influence the interaction patterns of the group (Hare & Bales, 1963; Leavitt, 1951; Strodtbeck & Hook, 1961). Steinzor (1950) found that individuals in a circular seating arrangement interact more with individuals opposite rather than adjacent to them. Mehrabian and Diamond (1971) observed that in four-person groups, more conversation occurred among persons seated closer together and facing one another, but only for those sensitive to rejection. Gardin, Kaplan, Firestone, and Cowan (1973) found that when eye-contact was blocked, a side-by-side arrangement was preferable to an across-the-table arrangement. Sommer (1969) found that rearranging the location of chairs increased the incidence and duration of conversation among patients in a nursing home. Perhaps most famous was the study by Allen (1977), who demonstrated that the probability of two people communicating in an organization is a decreasing hyperbolic function of the distance separating them (i.e., it is rapidly decreasing past the first 30 meters of physical desk separation). These studies indicate the importance of spatial arrangements of information in a face-to-face context. While distance measures indicate an important facet of face-to-face communication, the effects of personal distance in a computer-mediated environment (especially in the absence of non-verbal cues) remain to be determined. Thus our first research question asks, “What are the spatial constraints on communication in a virtual (i.e., graphical) computer-mediated environment?”
Another approach to the effects of spatial distance in human communication derives from the work of Edward T. Hall (1959, 1963, 1966), who pioneered the study of spatial communication, or proxemics. Hall identified four ranges of distance based on the nature of the relationship between individuals: 0–18 inches is intimate distance; 18 inches-4 feet is personal distance; 4–12 feet is social distance; and 12–25 feet is public distance. Although these ranges appear to be supported for face-to-face interaction, virtual computer-mediated distances have yet to be examined. Our second research question asks, “What are the distance ranges in a virtual computer-mediated environment?”
Uncertainty Reduction Theory (URT) (Berger, 1973; Berger & Calabrese, 1975) argues that individuals are mainly concerned with predicting behavior and use communication behaviors to facilitate this process. URT is particularly adept at articulating how uncertainty is reduced between strangers in initial interactions (see Berger, 1973; Clatterbuck, 1979; Douglas, 1990). In a computer-mediated environment, communication among strangers readily occurs in chat rooms, bulletin boards, and via email. Uncertainty reduction in a graphical chat environment occurs in various forms, such as asking a person's name, age, and sex (Suler, 1996). Uncertainty reduction seemingly becomes more salient because personal characteristics apparent in a face-to-face context are often missing in a computer-mediated context. Our third question asks, “How do individuals reduce uncertainty in an virtual computer-mediated environment?”
Expectancy Violations Theory (Burgoon & Hale, 1988; Burgoon, Buller, & Woodall, 1989) applies to the increase or decrease of distance to another person in an interpersonal interaction. Different cultures and contexts elicit alternative expectations on distance of interaction. Feeley and de Turck (1995), for instance, found that persons who violate expected spatial relationships are judged to be less truthful than those who do not commit such violations. One way to determine expectancy violations in CMC is to measure the conversational appropriateness of another in a virtual environment (Canary & Spitzberg, 1987). Our fourth question asks, “How do individuals violate appropriate conversational distances in a virtual computer-mediated environment?”
Equilibrium Theory (Argyle & Dean, 1965), posits that intimacy and distance vary together. From this perspective, it is hypothesized that the closer the distance, the greater the intimacy; and, conversely, the greater the distance, the less the intimacy. Intimacy in an online context may be difficult to operationalize given the reduced non-verbal cues. For that reason, we propose that social attractiveness (McCroskey & McCain, 1974) provides an analogous measure of intimacy. In an online context, this could be measured in a relatively straightforward manner by correlating spatial distance to perceived social attractiveness of the other individual. Our fifth question asks, “What is the relationship between social attraction and spatial distance in a virtual CMC environment?”
The research discussed in this paper provides the basis for testing the above questions empirically. Specifically, the measure of spatial distance in an online environment will be examined in detail and compared to various social-psychological constructs based on previous research in face-to-face interpersonal communication. High and Low-Context Communication (level of uncertainty), Conversational Appropriateness, and Social Attraction will be used to examine the spatial constraints (i.e., distance ranges) in a virtual computer-mediated communication environment (i.e., 2D graphical space). The following table summarizes the basic research questions we examine in this paper.
Research Design: Uncertainty Reduction Theory, Conversational Appropriateness, and Social Attraction vs. Spatial Distance
The impetus for thus research was an observation that French-speaking avatars seemed spatially closer in online graphical conversations (a hypothesis yet to be verified). Building upon the idea that of real-life attributes are replicated in an online context (Reeves & Nass, 1996), the “French-closer-distance” hypothesis generated a series of conversations culminating in the study presented in this paper. Research is not performed in a vacuum, however, and should build upon extant theoretical constructs. The study presented in this paper utilizes interpersonal communication research because of its emphasis on interactions between people and frames this research in an interactive computer-mediated environment. In this section, we examine three constructs deemed applicable to a graphical environment predicated on spatial distance as a measure of online behavior. Uncertainty Reduction Theory, Social Appropriateness, and Social Attractiveness are described, justified, and operationalized in a virtual CMC environment.
Uncertainty Reduction Theory
Clatterbuck (1979) extended Uncertainty Reduction Theory by suggesting that the process of uncertainty reduction was not simply calculated by correctly choosing which alternative will occur or which explanation is correct. Rather, he explained it as a function of the quality of information gathered about the other individual. Clatterbuck operationalized uncertainty reduction through an attributional confidence scale (CLUES). A sixty-five-item retroactive scale (CL65) and a seven-item proactive instrument (CL7) were developed from the axioms and theorems originally proposed by Berger and Calabrese (1975). Building upon Clatterbuck's (1979) attributional confidence scale, Gudykunst, Yang, and Nishida (1985) incorporated aspects of high-and low-context communication (Hall, 1976) as a manifestation of culture (see also Gudykunst, 1983; Hofstede, 1983; Smith, 1981). High context communication is internalized in the person, with very little coded in the explicit part of the message. Low-context communication, alternatively, reflects a message in which verbal (or written) communication contains much of the meaning in the message. Gudykunst and Nishida (1986) developed and tested an attributional confidence scale which included 12 items representing low- and high-context communication. This scale combined Clatterbuck's (1979) seven-item (CL7) scale (low context communication) with five items measuring high-context communication as a means to examine cultural variation in uncertainty reduction (i.e., attributional confidence). The Attributional Confidence (0–100%, continuous) scale developed by Gudykunst and Nishida (1986) was amended to a five-item low-context communication and three-item high-context communication based on their factor analysis of scale items. The amended Attributional Confidence scale will be applied in the context of this paper.
Appropriate communication is defined as that which “avoids the violation of the situational and relational rules governing the communication context” (Canary & Spitzberg, 1987, p. 93–94). It can be judged in terms of verbal sensitivity, relational context, and environmental context (Weimann & Backlund, 1980). Studies of communication competence have examined both the appropriateness and the effectiveness (or goal accomplishment) of conversation (Canary & Spitzberg, 1987; Spitzberg & Cupach, 1984). Canary and Spitzberg's (1987) General and Specific Conversational Appropriateness Scales measure the perceived appropriateness of each person's conversational partner. These scales, and an accompanying Conversational Effectiveness Scale have been primarily used in studies of conflict. Although the items composing the Effectiveness Scale are particularly related to conflict situations, the Conversational Appropriateness measure is applicable to a wider range of different types of conversational interactions. The original Conversational Appropriateness Scale consisted of 20 items rated using a 7-point Likert type scale, with responses ranging from 1) strongly disagree to 7) strongly agree. A factor analysis with Kaiser VARIMAX rotation performed by Canary and Spitzberg (1987) identified two factors within the appropriateness scale. The first was “specific appropriateness” and was exhibited in 4 items relating to behavioral subsets of portions of the conversational episode. These included items such as “Some of the things he or she said were embarrassing to me” (reverse coded). The second factor, “general appropriateness,” was exhibited in 5 items relating to global impressions of the communicator's suitability. The 4 item Specific Conversational Appropriateness scale and 5 item General Conversational Appropriateness scale will be used in this study.
Interpersonal attraction between online communicators was assessed using the “social attraction” component of McCroskey and McCain's (1974) Interpersonal Attraction Scale (IAS). Research suggests that interpersonal attraction is characterized by at least three dimensions, a) a liking or social dimension, b) a task or respect dimension, and c) a physical dimension (McCroskey & McCain, 1974). Although all three dimensions are usually measured by the 15 item IAS scale, a number of researchers have reported high internal reliability coefficients for each dimension (Ayres, 1989; Brandt, 1979; McCroskey & McCain, 1974; Wheeless, Frymier, & Thompson, 1992). In the context of virtual CMC, communicators often do not have immediate access to information about physical characteristics. Many virtual environments also do not require participants to engage in any task. Using these arguments, Task Attraction and Physical Attraction are not assessed in the context of this study. The perceived social attraction (liking or social dimension) of each person's conversational partner was assessed using a 5 item 7-point Likert-type scale ranging from 1) Strongly disagree to 7) Strongly agree. The Social Attraction scale asks respondents to report their agreement with statements such as “I think he (she) could be a friend of mine.” The 5 item perceived Social Attraction Scale will be used in this study.
Methods: Study Description and Tools
Computer-Mediated Environment: The Palace
The computer-mediated environment in this study consisted of separate and dedicated rooms created in and sponsored by The Palace. The Palace software is a free service which enables users to communicate with others in real time, using text, two-dimensional graphics, and movement. A Palace site is an online community where users can come together and chat while being represented by a graphical image (or avatar). There are thousands of existing Palace sites which contains a “mansion” with often over 100 rooms. Rooms can be entered via virtual doorways or passages and users can roam through a palace by directly specifying a room or by navigating from room-to-room. The Palace also allows users to create their own Palace site, populated and controlled by means of a specific domain name. Since the service was first offered for free in July, 1998, tens of thousands of servers have been downloaded and deployed and the 25 most populated palaces are updated every fifteen minutes. The Palace provides many opportunities for business partnerships and contains many other services such as PalacePresents, an auditorium for live, moderated presentations integrating streaming audio, video, and synchronized delivery of HTML based slide shows.
Subjects were 160 university students. Half were from a large midwestern university (Mage=20.25) and half were from a large eastern university (Mage=19.83). Sixty-four subjects were male and ninety-six subjects were female. The students had not met before and were classified as strangers. The students were offered extra credit for their participation in the study.
Upon arrival, participants were greeted by the experimenter and were told that development of an online class was in progress and that communication options were being tested. Subjects were informed that they were to “get to know” another student. Subjects were asked to sign a consent form stating that they understood any conversation online would be recorded and would be kept confidential. After the student completed the consent form, s/he went though a basic tutorial on how to use the computer software utilized in the particular study. Subjects were also informed that if at any point in the study they had questions the experimenter would assist them and that they could discontinue their conversations at any time without ramification.
The basic tutorial described the Palace and subjects from the same location (i.e., midwestern or eastern) were assigned yellow “Smiley Face” avatars which depicted their university and number designator (e.g., Eastern University Student 1). Within-university students began interacting with each other in a Palace designed room as part of the tutorial. Students were encouraged to move around and chat to get the feel of the 2-dimensional virtual environment. After five minutes, the experimenters from both locations contacted each other via phone and synchronously gave each individual instructions how and when to enter their private room with a student (stranger) from the respective distant university. Separate Palace sites were created for this study and were not accessible to outside intervention (i.e., only the dyad was allowed access to their private room). The subjects were again reminded that the purpose of the interaction was to simply try to “get to know” their counterpart in their private room. The two subjects were given 7–10 minutes to interact (allowing for completion of flows of conversations). After this time, subjects were told to log off the system. During the interaction the log (script) of the conversation was recorded and the actual visual chat was recorded using Lotus' ScreenCapture program (see Video Capture Methods below). Subjects were then asked to complete a questionnaire after their interaction. The questionnaire consisted of Gudykunst and Nishida's (1986) Attributional Confidence scale which contained seven low-context items and five high-context questions. Analysis examined the five low-context items and the three high-context items which had high factor loadings per Gudykunst and Nishida's (1986) factor analysis of this scale. Social appropriateness was measured using the general and specific conversational appropriateness measures taken from Canary and Spitzberg's (1987) Conversational Appropriateness and Effectiveness Scale. Social attractiveness was measured using the Social Attractiveness Component of McCroskey and McCain's (1974) Interpersonal Attraction Scale. Computer experience questions were combined with demographic questions at the end of the questionnaire. At one of the locations, student participants wrote a two-page qualitative evaluation of their experiences in the online environment as part of a class assignment. After the study was completed, the subjects were fully debriefed.
Video Capture Methods
One of the salient features of the study is that it focuses on chats in graphical online environments rather than the usual text-based Internet chats (e.g., Internet Relay Chat, IRC). Analyzing text-based online conversation only requires text files that contain the script of conversations. The need to measure the spatial distance between avatars during the online conversation requires special tools that record all the information including graphics. Videotaping computer screens is an alternative, but it lacks a digitalization feature which is essential to computerized analysis. Also, videotaping requires extra hardware settings which makes data analysis more cumbersome.
To solve this problem, we decided to use computer software that captures all text and graphical information during the conversation. One such tool available on the market is the Lotus ScreenCam program. ScreenCam captures all activities on a computer screen onto a digital movie file. The movie file can later be converted into a Microsoft Windows wave file (sound only), audio video interlaced (AVI) file (video only or sound+video), or self-running executable file. The file conversion capability of the software gives it increased data-analytic flexibility. Once the ScreenCam movie file is converted to an AVI file, it is relatively easy to find or construct software for data analysis. Computerized data analysis is easier and more accurate than manual coding which tends to result in errors. Since it records all screen activities, other information such as users' typing speed, unsent messages, avatar alteration, etc., are also available when using screen capture software.
The movie file size depends on the number of screen colors and screen resolution. To keep the file size manageable, computer screens were set to 256 colors and 800x600 resolution. Like many other Internet services, online chat in the Palace is free of charge, but users are expected to see ad banners that pop up in a small window. Since the pop-up window can distract subjects during conversation, something was needed to block the banner window. We designed a program for that purpose which essentially was a blue rectangle which covered all advertisements on the Palace. Before the actual conversation, subjects were informed to press a key to initiate the screen capture program. Upon completion of a conversation, researchers stopped the program and saved the file with a predetermined name. Since the Palace program supports saving the script of conversations in text files, one text file and one movie file were acquired for each conversation and were compared to verify the accuracy of matched pairs and the contents of each conversation.
Spatial Distance Measurement
Once all ScreenCam files were converted to AVI file format, we developed the Spatial Distance Analysis Program that automatically calculated the pixel distance between avatars. The distance was measured at two points in time for each conversation: (1) When users start exchanging messages (Time 1); and (2) When users terminate their conversation (Time 2). The measured distance is the distance between the central point (centroid) of the round avatar and the unit of distance is the pixel (800×600 screen resolution = 800×600 pixels in the screen). A pixel is a unit of distance which depends on the resolution of the screen and the size of the window used in the Palace environment. Within the constraints of this study, a pixel is equivalent to .0179245288 inches. The software used for the measurement gives the total length of the conversation and requires a coder to simply click on each avatar at appropriate times. Then the program automatically calculates and saves the distance information, which prevents possible typing or measurement errors. With the Spatial Distance Analysis Program's play, pause, stop, rewind, and fast forward buttons, the coder can easily measure the distance at appropriate times. Appendix 1 provides the basic information contained in the Spatial Distance Analysis Program.
Results: Empirical Responses to Research Questions
The results of the study answer the research questions provided in Table 1 and includes a section on scale validation. Specifically, distance ranges and their delineations, uncertainty reduction, conversational appropriateness, and social attraction results are reported in this section.
Table 1. Research Questions
RQ1. What are the distance ranges in virtual computer-mediated communication?
RQ2. What are the delineations of distance ranges in virtual computer-mediated communication?
RQ3. What are the relationships between uncertainty reduction (high and low-context communication) and spatial distance?
RQ4. What are the relationships between conversational appropriateness and increases/decreases in spatial distance?
RQ5. What are the relationships between spatial distance and social attraction?
Table 2 provides the reliabilities (alpha coefficients) of each of the scales utilized in the present study and compares them to those of their sources in a face-to-face context.
Table 2. Alpha Coefficient Comparison
.85 (Gudykunst & Nishida, 1986)
.94 (Gudykunst & Nishida, 1986)
General Conversational Appropriateness
.80 (Canary & Spitzberg, 1987)
Specific Conversational Appropriateness
.85 (Canary & Spitzberg, 1987)
.84 (McCroskey & McCain, 1974)
The distances were found using the Spatial Distance Analysis Program. The mean distances are reported in Table 3 for each of the two time periods. Time 1 specifies the distance when conversations first commenced. Time 2 specifies the distance at the end of the conversation. Overall time refers to the mean and standard deviation values of combined Times 1 and 2.
Table 3. Distance Ranges (N=160 at each time period)
Mean Distance (in Pixels)
Next, we examined the relational movements for each time period and partitioned the distances by looking at whether individuals became closer, farther, or stable in their spatial distance from Time 1 to Time 2. The number of individuals in each of these categories, the mean distance (negative distance refers to individuals moving closer together), and standard deviations are reported in Table 3. The total changes in distance ranges of Time 2-Time 1 are also reported in Table 4.
Table 4. Changes in Dyad Distance (N=160
Number of Individuals
Mean Distance (in Pixels)
Total (Time 2-Time 1)
Means and Standard Deviations
Table 5 lists the study results in terms of means, standard deviations, and correlations in terms of Low-Context (LC) and High-Context (HC) Communication (Uncertainty Reduction), Conversational Appropriateness (General [GCA] and Specific [SCA]), and Social Attraction (SA). The ensuing sections examine these results in further detail.
Table 5. Results: Means, Standard Deviations, and Correlations
Correlation is significant at the 0.01 level (2-tailed);
Correlation is significant at the .05 level (2-tailed)
Low-Context Communication (0–100 scale)
High-Context Communication (0–100 scale)
General Conversational Appropriateness (1–7 scale)
Specific Conversational Appropriateness (1–7 scale)
Social Attraction (1–7 scale)
Uncertainty Reduction Results
The results in Table 5 reflect an individual's uncertainty reduction following a dyadic interaction with another person. Relating this information to Gudykunst and Nishida's (1986) low- and high-context communication (i.e., attributional confidence) results for communication with strangers yields some differences in the overall values. For low-context communication items, attributional confidence in the present study yielded MLC= 29.18 and SDLC=19.84 versus in MLC=40.70 and SDLC=38.56 in Gudykunst and Nishida's (1986) face-to-face US strangers. For high-context communication items, attributional confidence in the present study yielded MHC=39.10 and SDHC=24.39 versus MHC=42.59 and SDHC=58.23 in Gudykunst and Nishida's (1986) face-to-face US strangers. The correlation between low-context and high context communication in the present study was .64 compared to that of .68 for Gudykunst and Nishida's (US) face-to-face groups. A paired sample t-test indicated that there were significant differences between high and low-context communication [t=6.74; df=159; p<.001]. This result validates these items as separate constructs, while recognizing their correlational similarities (i.e., they are both measures of uncertainty reduction).
Significance was determined by regressing low- and high-context communication on change in spatial distance (i.e., delta distance) over time. The use of the over-time change in spatial movement was based on the argument that uncertainty reduction implies a change over time, thus necessitating the use of like measures of distance. For the overall group (N=160), the comparison of spatial distance movement of Time 2-Time 1 with low-and high context communication did not yield any significant findings. We found no significant relation between the overall group's movement and whether a person was a low- or high-context communicator. Similarly, the within-group comparisons of the spatial distance movement of Time 2-Time 1 for the farther group (N=60) and stable group (N=48) with low- and high-context communication did not yield any significant findings. We found no significant differences between the groups that did not move and those that moved farther apart with respect to low- or high-context communication. However, the within-group comparison of spatial distance movement of Time 2-Time 1 for the closer group (N=52) with low- and high-context communication did yield significant results. There was a significant relation between individuals who moved closer together and low- and high-context communication. For the low-context items, F(1,51)=5.50, p≤.05 (beta=.315). For the high-context items, F(1,51)=4.44, p≤.05 (beta=.286). For the group that moved closer together over time, greater spatial distance signified higher uncertainty reduction (i.e., attributional confidence). This result indicates that attributional confidence increased for the group which moved closer together, but only for individuals who were more distant. As a means to explore this relationship further for the closer group, a within-group paired sample t-test was conducted. It indicated that there were significant differences between high- and low-context communication [t=5.5; df=51; p≤.001]. Between group (i.e., closer, farther, and stable) comparisons were performed, but did not yield any significant results.
Conversational Appropriateness Results
The results for conversational appropriateness reflect the perceived conversational appropriateness of the other person based on their dyadic interaction. Relating this information to Canary and Spitzberg's (1987) results yields similarities. The correlation between general and specific appropriateness in the present study was 55. compared to .51 obtained by Canary and Spitzberg for face-to-face groups. The t-test results did yield significant differences between general conversational appropriateness versus specific conversational appropriateness [t=7.38; df=159; p≤.001]. This result validates these items as separate constructs, while recognizing their correlational similarities (i.e., they are both measures of appropriateness). As a result, specific and general conversational appropriateness will be included separately in the ensuing significance tests.
Significance testing was done regressing general and specific conversational appropriateness on change in spatial distance from Time 2 to Time 1. The use of the over-time change in spatial movement was based on the initial argument in this paper, which stated that expectancy violations (and hence social appropriateness) relate to the increases or decreases in “distance” in interpersonal interactions (Burgoon & Hale, 1988).
For the overall group (N=160), the comparison of spatial distance movement over time with general conversational appropriateness yielded significant results [F(1,159)=4.58; p≤.05 (beta=.168)]. There was an overall significant relationship between an individual's attribution of general conversational appropriateness to the other and spatial distance movement from Time 2-Time 1. For all groups, as the change in distance increases over time (i.e., they become farther apart), general conversational appropriateness is higher. Examining this finding further, the within-group comparisons of the spatial distance movement to general conversational appropriateness from Time 2 to Time 1 for the farther group (N=60), stable group (N=48), and closer group (N=52) did not yield any significant findings. We found no significant relationship between spatial distance movement and whether a person perceived his or her conversational partner as generally appropriate. There were no significant differences between the stable versus closer groups (N=100) and stable versus farther groups (N=108) with respect to spatial distance movement over time. However, significant differences were found between the closer versus farther groups (N=112) with respect to spatial distance movement over time [t=2.10; df=111; p≤.05]. This result indicates that there was a significant difference between individuals who moved (closer versus farther) with respect to their perception of the other's general conversational appropriateness. For all groups that moved over time (i.e., closer and farther), greater spatial distance signified higher general conversational appropriateness. This result suggests that the variance found in overall group general conversational appropriateness (i.e., the initial significant finding) may be accounted for by differences between closer and farther movers; that the overall variation in general conversational appropriateness stems from the “movers and shakers.”
For the overall group (N=160), the comparison of spatial distance movement over time with specific conversational appropriateness did not yield significant results. We found no significant relationship between the overall group's movement of Time 2-Time 1 and whether the other was perceived to have exhibited specific conversational appropriateness. Similarly, the within-group comparisons of spatial distance movement of Time 2-Time 1 for the closer group (N=52) and stable group (N=48) with specific conversational appropriateness did not yield any significant findings. We found no significant relationship within the groups that did not move and those that moved closer together and whether the perceived other exhibited specific conversational appropriateness. However, the within-group comparison of spatial distance movement of Time 2-Time 1 for the farther group (N=60) with specific conversational appropriateness did yield significant results. There was a significant relationship between moving farther apart and whether the other was perceived to exhibit specific conversational appropriateness [F(1,59)=4.30, p≤.05 (beta=.263)]. For the group that moved farther apart over time, greater spatial distance signified higher specific conversational appropriateness. As a means to explore this relationship further, between group (i.e., closer, farther, and stable) comparisons were performed, but did not yield any significant results.
Social Attraction Results
The results for social attraction (as with conversational appropriateness) reflect the perceived social attraction of the other person based on their dyadic interaction. As noted earlier, the perceived social attraction (liking or social dimension) of each person's conversational partner was assessed using the “social attraction” component of McCroskey and McCain's (1974) Interpersonal Attraction Scale (IAS). The results from the present study indicate that the mean value of social attraction (7-point scale) is 4.96 and that the standard deviation is 1.13.
Significance testing was done by regressing social attraction on the absolute spatial distance at Time 2. The use of the absolute distance at Time 2 varies from the previous analyses (which used the change in distance from Time 2 to Time 1). The use of the absolute distance is based on the initial argument in this paper: “the closer the distance, the greater the intimacy” (Argyle & Dean, 1965). The absolute distance value thus has relevance in determining the actual space between individuals in a dyadic conversation. Time 2 is used as the referent point because it occurs closest to the time when social attraction was measured (i.e., after the conversation concluded). The use of the absolute distance also allows insight into RQ2, which asks, “What are the delineations of distance ranges in online communication?” It is argued that the delineations in distance should be obtained using actual distances, allowing for applicability of the results in a real-life context. The absolute distance becomes meaningful because it measures the distance (in pixels) between avatars in a graphical environment. Recall that a pixel is a unit of distance which depends on the resolution of the screen and the size of the window used in the Palace environment and is equivalent to .0179245288 inches. Hence, the results in this section reflect the actual distance (in pixels) at the end of the conversation to the perceived social attraction of the other person in the dyadic interaction.
This spatial distance range was determined to be between 37 and 471 pixels (0.66 inches and 8.44 inches, given the conditions in this study). Examining the distribution (i.e., scatter plot) of the absolute distance versus the mean perceived social attraction produced insight into the type of regression analysis utilized. We noticed a pattern of distribution resembling a positive parabolic function. A positive parabolic function is indicative of a “U-shaped” distribution which centers around an absolute distance (i.e., has a vortex). The vortex of these results is centered was identified as around 190 pixels (eye-balled figure). At this point, a quadratic regression analysis was performed using the functional form:
Equation 1. Parabolic Function
Where: y= perceived social attraction; b is the (unstandardized) slope of the parabola; a is the vortex, and A is the unstandardized coefficient (constant). When using a parabolic function as a means for quadratic regression, the value of a must be entered to complete the analysis. The initial value used was the eye-balled absolute distance of 190 pixels which yielded a significant result (i.e., < .05 significance level). This process was continued until the value entered yielded a non-significant result (i.e., >.05 significance level). The results of the quadratic regression comparing absolute distance at Time 2 to perceived social attraction of the other were significant at the .05 level for distances ranging from a=175–274. This means that there was a range of 100 pixels (175–274, inclusive) which indicated significant relations between absolute distance at Time 2 with perceived social attraction for that entire range. The range of highest significance was found between 212 and 226 pixels [F(1,160)=6.25, p<.01; (b=.2.55 × 10−5; A=4.76)], which is indicative of the vortex (or lowest point) of the parabola. The results of this analysis yield the following parabolic equation reflecting the (unstandardized) relationship between social attraction (y) to that of pixel distance (Note: the pixel range from 212–226 was averaged):
Equation 2. Social Attraction (y) as a Function of Pixel Distance (Unstandardized)
This equation provides a means to predict social attraction given the absolute distance (in pixels) of avatars. Figure 1 denotes the parabolic relationship between social attraction (y-axis) and spatial distance at Time 2 (x-axis).
In Figure 1, the close-range distance (37–174 pixels) and far-range distance (275–471 pixels) zones reflect an increase in social attraction as one moves away from the middle-distance (or Danger) zone; as the absolute distance between avatars approaches the value of 219 pixels (approximately 4 inches, in this case), perceived social attraction is decreased. These results imply that there is an avatar distance that you DO NOT WANT TO BE if you want another person to perceive you as socially attractive. According to these results, the Danger Zone distance for interpersonal attraction is from 175–274 pixels or 3.12–4.88 inches, given the conditions in this study. Further t-tests of mean values of absolute spatial distances indicated significant differences between the close range (37–174 pixels) and middle range (175–274 pixels) [t=2.27; df=140; p≤.01 (one-tailed)], the far range (275–471 pixels) and middle range (175–274 pixels) [t=1.65; df=66; p≤.05 (one-tailed)], and the far range and the close range [t=0.35; df=104; p=ns (two-tailed)]. These results indicate that there were significant differences between low versus middle and high versus middle ranges. There were no significant differences between low and high distance ranges. This last result is important because it provides evidence that low distance and high distance ranges were similar in terms of social attraction (and different from those in the “danger zone”). Close-Range and Far-Range spatially distant individuals were more socially attracted to their conversational partner. In summary, our results indicate that there is a “danger zone” in sustained dyadic conversation distance in relation to social attraction (liking) of 175–274 pixels. Results also indicate that avatars that were closer in spatial distance (37–174 pixels) and farther in spatial distance (275–471 pixels) were more socially attracted to their dyadic partner.
Summarizing the results in this section, Table 6 provides a list of the key findings in this section.
1a. For the group that moved closer together over time, greater spatial distance signified higher uncertainty reduction (i.e., attributional confidence)
1b. For the group that moved farther apart over time, greater spatial distance signified higher specific conversational appropriateness
1c. For all groups that moved over time (i.e., closer and farther groups), greater spatial distance signified higher general conversational appropriateness
2. The relationship between spatial distance and social attraction (i.e., liking) is a positive parabolic function
2a. Close-Range and Far-Range spatially distant individuals were more socially attracted to their conversational partner
2b. Middle-Range spatially distant individuals were less socially attracted to their conversational partner
General Results Discussion
This section discusses the results reported in Tables 2–4. Specifically, scale reliability comparisons, mean distance and standard deviation ranges, and changes in dyad distance are covered in this section.
The alpha coefficient comparisons in Table 2 are indicative of the applicability of face-to-face interpersonal scales to a virtual CMC environment. The alpha coefficient comparisons are as follows: low-context communication, alpha =.86 (CMC) compared to .85 (FTF); high-context communication, alpha =.77 (CMC) compared to .94 (FTF); general conversational appropriateness, alpha =.75 (CMC) compared to .80 (FTF); specific conversational appropriateness, alpha =.86 (CMC) compared to .85 (FTF); social attraction, alpha =.79 (CMC) compared to .84 (FTF). In general, the alpha coefficients were all relatively high and were similar to those obtained in face-to-face environments. Interestingly, with the exception of high-context communication, social attraction, and general conversational appropriateness, the reliabilities were all slightly higher in a CMC environment. In fact, only the high context communication alpha was more than .05 different in CMC versus face-to-face environment. The high-context communication scale may be the most suspect in terms of application in a CMC environment. Nevertheless, the reliability of this scale was still acceptable (.77). While these results do not revalidate the scales used (a factor analysis needs to be performed to determine to see if the factor loadings change in a CMC environment), they do indicate that the instruments are reliable in a CMC environment.
From Table 3, the mean distance range was 159.34 pixels at Time 1 and 171.69 pixels at Time 2. Overall mean distance between these two ranges was 165.52 pixels. Given the constraints of this study (i.e., resolution of screen and size of palace window), 1 pixel is equivalent to .0179245288 inches. The mean distances(in pixels) translate to 2.85 inches at Time 1 and 3.07 inches at Time 2. This indicates that there was a slight increase in distance from Time 1 (the distance at the beginning of the conversation) to Time 2 (the distance at the end of the conversation) of 0.22 inches. Overall this is a slight increase in distance. The standard deviations were 58.81 pixels (1.05 inches) at Time 1 and 73.43 pixels (1.31 inches). This indicates that there was variation around the mean distances at each time period and that they moved slightly more (.26 inches) at Time 2. There was movement over time in the dyadic conversations and the changes in dyad distance examines these movements.
From Table 4, the movement of dyads from Time 1 to Time 2 were nearly equally distributed. Fifty-two individuals moved closer together, sixty individuals moved farther apart, and forty-eight individuals did not move during the 7–10 minutes conversation. This indicates that there was movement both closer together and farther apart over time. The mean distance range was thus a combination of individuals who moved closer and farther apart. The mean distances were -55.00 pixels (negative values signify closer together; Time 2-Time 1) for individuals who moved closer together and 80.78 pixels for individuals who moved farther apart. The mean distances (in pixels) translate to .99 inches for the individuals who moved closer together and 1.45 inches for individuals who moved farther apart. Standard deviations of these changes indicates similar results. The standard deviations were 42.47 pixels (.76 inches) for the closer group and 83.03 pixels (1.48 inches) for the farther group. These results indicate, in general, that individuals moved .46 inches farther apart and had a higher range of movement when they moved apart. This finding may be a result of the fact that individuals had more room to move farther apart than closer together, given that the maximum distance for all conversations was 8.44 inches and the fact that they initially started 2.85 inches apart. This could also be explained by possibility of “invasion of personal space.” This explanation would suggest that there is a “intimate” personal zone which restricts movement to close distance ranges. Given that the minimum distance for all conversations was 37 pixels (.67 inches), this indicates that there is a minimum distance that normal conversation can occur (i.e., otherwise avatars would be on top of each other and, hence, undifferentiable). There is some evidence of what happens when avatars get too close together. Figure 2 provides an example of this occurrence.
In Figure 2, an avatar is partially blocking the top of another avatar. The blocked avatar's response is “get off my forehead.” This indicates that there is some form of minimal distance range and that this may be indicative of a intimate zone infraction analogous to the intimate proxemic range (Hall, 1966). It also indicates a translation of virtual representation of everyday phenomena, as the top of an avatar is translated into “my forehead.”Suler (1996) describes this phenomenon in detail:
Blocking: Members consider it a social faux pas to place your avatar on top of or too close to another person's prop. Unless the person is a friend who's in a mood to be close, it's an invasion of personal space. “Please get off me!” and “You're sitting on me!” are two common complaints. Again, some naive users do this without knowing it is inappropriate, or the person may be lagging and unable to move. But some hostile people deliberately accost others by blocking them.“
Specific Results Discussion
This section examines the specific results for each theoretical construct tested by elaborating on the means and standard deviations for each measure (i.e., Table 5) and the empirical results for each measure (i.e., Table 6). Each of the measures will be discussed according to the framework provided in Table 6, particularly in response to the research questions posed at the outset of this paper. The basic format used will be to state the overall finding (i.e., Finding 1 and Finding 2 in Table 6), and then elaborate on the specific findings which further describe components of the overall finding.
Finding 1 indicates that individuals find spatial distance as a regulator of whether the other person is, in general, conversationally appropriate. The farther individuals position their avatars apart, the more conversationally appropriate they are perceived. This finding is not surprising and mirrors what we experience in face-to-face conversations. Personal space is considered as territorial and when this personal space is invaded, one's behavior could be considered inappropriate. Sommer (1969) posits that a person who moves closer to another in conversation is perceived as more approachable up to a certain point. It is argued that there is a certain point where it is considered inappropriate (i.e., obnoxious, rude, overbearing, etc.) and moving farther from a person does not encroach upon personal space and would therefore not be considered inappropriate. This finding indicates that it may be considered inappropriate for strangers to move too close in initial conversations, but that moving farther apart would not be considered inappropriate. Interestingly, none of the within-group analyses (i.e., closer, farther, or stable groups) replicated this finding; that greater spatial distance and higher general conversational appropriateness was not due to any one type of movement (i.e., closer or farther). This suggests that other factors play a role in this process.
Finding 1a–for groups that moved closer together, greater distance signified higher uncertainty reduction (i.e., attributional confidence)–provides insight into the Finding 1 (which found no significant results for this group). The closer groups seemed to use greater spatial distance as a form of uncertainty reduction. It should be noted that no other significant results were found relating uncertainty reduction to spatial distance movement. This indicates that uncertainty reduction (i.e., attributional confidence) was found to characterize those who moved closer together, but not those who moved farther apart or remained stable. For those who moved closer, there seemed to be a “too close” range and by moving into this range, an individual's uncertainty may have increased; that greater distance reduced uncertainty for the group that moved closer together. Relating this to Finding 1, it is posited that uncertainty reduction (i.e., attributional confidence for both high- and low-context communicators) can explain higher general appropriateness for the group which moves closer together. Recall that the uncertainty reduction measures (both low- and high-context communication) were based on how people perceives themselves in their interactions with others. This could be attributed to the fact that uncertainty reduction measures one's own perceptions of uncertainty versus the role of the perceived other (as in social attraction or conversational appropriateness).
It should be noted that Finding 1a was observed for those who are both low- and high-context communicators. Further tests indicated that high-and low-communication were indeed different constructs, although they were highly correlated. The correlation between low- and high-context communication in CMC (.64) is similar to that of face-to-face interactions (.68) (Gudykunst & Nishida, 1986). The mean and standard deviations, however, reveal differences from the face-to-face environment. For low-context communication items, attributional confidence in the present study yielded MLC= 29.18 and SDLC=19.84 versus MLC=40.70 and SDLC=38.56 in Gudykunst and Nishida's (1986) face-to-face US strangers. For high-context communication items, attributional confidence yielded MHC=39.10 and SDHC=24.39 in the present study versus MHC=42.59 and SDHC=58.23 in Gudykunst and Nishida's (1986) face-to-face US strangers. The low-context communication mean was much lower than in the face-to-face context. This could be a result of the lack of social cues in a CMC environment or the relatively short time duration (7–10 minutes) spent interacting. Another explanation could be that there may have been no anticipation of future interaction and information exchange between strangers (see Kellermann, 1986). As a result, the amount of uncertainty reduced would be smaller in the CMC environment given the reduced motivation in reducing this uncertainty (Kellermann & Reynolds, 1990). The standard deviations in the CMC environment were also much lower. This could be a result of the relatively low means (i.e., the range of movement would be limited), but also may be indicative of the reduced social cues and limited discussion time. The relatively high standard deviations reported by Gudykunst and Nishida (1986) may also be a product of higher cultural variability among their sample of US students.
Finding 1b, that for the group that moved farther apart over time, greater spatial distance signified higher specific conversational appropriateness, provides another piece of information regarding Finding 1. Given that specific and general conversational appropriateness were found to be different measures, this finding indicates that the farther-apart groups used greater spatial distance as a form of specific conversational appropriateness. It should be noted that no other significant results were found relating specific conversational appropriateness to spatial distance movement. This indicates that specific conversational appropriateness was found to be related to those who moved farther apart, but not for those who moved closer together or remained stable. Recalling that specific conversational appropriateness dealt with items such as whether one said anything embarrassing (reverse coded), this measure would seemingly follow the idea that one cannot be considered inappropriate if they move farther apart, but could be deemed inappropriate if they invaded someone's territory. The mean value for specific conversational appropriateness was rather high (M=6.33 on a 7-point scale), indicating that most individuals were appropriate in their specific conversations. Given that the subjects knew that the conversations were being monitored, it is not surprising that there was very little specifically inappropriate behavior. In a setting where anonymity is more prevalent (i.e., a public Palace), it would seem that specifically inappropriate behavior such as flaming, SHOUTING, or using foul language would occur more frequently.
Finding 1a provided insight into the group which moved closer together and Finding 1b provided insight into the group which moved farther apart. These specific within-group comparisons are evidence toward the argument that closer groups had higher reduced uncertainty at greater distance intervals and that farther groups had higher perceived specific conversational appropriateness at greater distance intervals. Finding 1c–for the groups that moved over time (i.e., closer and farther groups), greater spatial distance signified higher general conversational appropriateness–lends insight into the relationship between Finding 1a and Finding 1b. General conversational appropriateness can be seen as a mediating variable between closer and farther groups; it cannot explain within-group differences, but can explain between-group differences. Figure 3 depicts a model incorporating the findings in this section establishes the relationship between variable leading to greater spatial distance over time:
Figure 3 provides a model which allows for testing the findings discussed in this section. The model depicts the relationship of greater distance movement as a combination of higher Uncertainty Reduction for those that move closer together and higher Specific Conversational Appropriateness for those that move farther apart. The resultant interaction is that of higher General Conversational Appropriateness. If one wants to reduce uncertainty in a virtual environment, the model suggests moving closer to another avatar over time, but not too close. If one wants to be specifically appropriate, then the model suggests moving farther from an avatar over time. The result could incorporate different phases in an online relationship and also suggests that spatial movement can be utilized in different ways depending on the purpose of the relationship.
Utilizing the model in a practical setting provides some interesting applications. One example of applying the model would be to the case where another person misinterprets a message to be rude or offensive (a common misunderstanding because non-verbal cues are missing). In this case, the model suggests moving farther away from this person while continuing the conversation. At this point one could “back off,” using the distance between avatars, and try to clarify your meaning. Moving closer to the other person would not be suggested in this case. As another example, suppose you would like to get to know a person better or would like to learn more about them (i.e., reduce uncertainty). The model suggests that moving closer would be an acceptable spatial solution, but that one should be careful about moving too close–that there should still be some distance between yourself and the other person (the actual distances of these movements will be discussed in the last part of this section). Incorporating both of these examples, the model predicts that one would be considered as generally appropriate by using spatial distance as a means to establish relationships in a virtual CMC environment. In the case where one wants to be perceived as generally appropriate while conversing in a graphical chat environment with a boss or respected figure (i.e., someone by whom one would want to be perceived as generally appropriate), one could adopt the above examples as a strategic form of spatial communication. By utilizing Uncertainty Reduction Theory and Specific Conversational Appropriateness, one adds a new non-verbal communication channel, that of spatial distance, to their online repertoire.
Finding 2. The relation between spatial distance and social attraction (i.e., liking) is a positive parabolic function
Finding 2 relates the absolute distance to perceived social attraction. This finding should be differentiated from Finding 1 which linked the change in distance over time to general conversational appropriateness. This finding further specifies the distance ranges in virtual computer-mediated communication. This spatial distance range was determined to be between 37 and 471 pixels (0.66 inches and 8.44 inches, given the conditions in this study). This indicates that there is a minimum and maximum distance by which communication occurred in this study.
The mean value of social attraction (7-point scale) was found to be 4.96 and the standard deviation was 1.13. This result shows that individuals perceived the other person as somewhat attractive (i.e., likeable). There was some variation around this value, but individuals did not generally perceive the other as socially unattractive. This changed when looking at the distance at Time 2 compared to social attraction. The quadratic regression indicated that there was a positive parabolic function which described the relation between social attraction and absolute spatial distance at Time 2. The equation relating this relationship was:
Equation 2. Social Attraction (y) as a Function of Pixel Distance (Unstandardized)
where: y= perceived social attraction; 2.55 × 10−5 is the (unstandardized) slope of the parabola; 219 is the vortex (in pixels), and 4.76 is the unstandardized coefficient (constant). It should be noted that this equation is unstandardized and should be used for comparison purposes only (i.e., the values have no absolute meaning). Equation 2 provided a means to predict social attraction given the absolute distance (in pixels) of avatars and was illustrated in Figure 1. This equation states that as the absolute distance between avatars approaches the value of 219 pixels (approximately 4 inches, in this case), the perceived social attraction is decreased.
The results indicate that there are three distance ranges which occur in 2-D virtual communication. The first range is termed as the “Danger Zone,” or middle-range distance. This is the range you do not want to be in if you want to be perceived as socially attractive: The Danger Zone is 175–274 pixels or 3.12–4.88 inches in the context of this study. The second range is termed the “Close-Range Zone.” This distance yields higher social attraction: The Close-Range Zone is 37–174 pixels or 0.67–3.10 inches in the context of this study. The third range is termed “Far-Range Zone.” This distance also yields higher social attraction. The Far-Range Zone is 275–471 pixels or 4.90–8.44 inches in the context of this study.
These results are interesting because they provide a response to RQ2, which asked about the demarcations in range. The significant results indicate that there are three separate distance ranges in relation to perceived social attraction. Further tests indicated that the Danger Zone was significantly different than both the Close-Range and Far-Range Zone. Interestingly, the Close-Range and Far-Range Zone were not significantly different in relation to social attraction. These results suggests that the Close-Range and Far-Range Zones reflect higher social attraction, whereas the Danger Zone (Middle-Range) reflects lower social attraction (Results 2a and 2b, respectively).
In light of past research on proxemics, Hall (1966) specifies four distance ranges of Americans: (1) intimate (0–18 inches); (2) personal (18 inches-4 feet); (3) social (4 feet-12 feet); and (4) public (12 feet-25 feet). These four categories are based on observations that both animals and man exhibit territorial behavior. The intimate range from 0–18 inches is largely characterized by non-verbal communication cues which may be absent in a Palace-type computer-mediated environment. Although it cannot be determined given the results in this study, perhaps the three ranges found in this represent personal, social, and public distances.
Equilibrium Theory (Argyle & Dean, 1965), posits that intimacy and distance vary together–the closer the distance, the greater the intimacy; and, conversely, the greater the distance, the lower the intimacy. The findings in this study indicate that spatial distance versus attraction (i.e., liking) reflects a parabolic function; that higher perceived social attraction occurs in Low-Range and High-Range individuals and that lower perceived social attraction occurs in middle-distant individuals. This is puzzling. It differs from what we experience in face-to-face contexts and (at least initially) is difficult to believe. What is happening here?
First of all, it should be noted that this relation is in terms of perceived social attraction. This measure was found by examining questions such as “I think he (she) could be a friend of mine.” It has been found that social attraction is highly correlated to general conversational appropriateness (r=.57) and specific conversational appropriateness (r=.52) (Canary & Spitzberg, 1987). This study found social attraction to be significantly correlated with general conversational appropriateness, r=.48 (p≤.01) and specific general conversational appropriateness, r=.47 (p≤.01) (see Table 5). Table 5 also provides the significant correlations between social attraction and low-context communication (r=.25; p≤.01) and high-context communication (r=.32; p≤.01). These results suggests that social attraction has some of the same underlying causal mechanisms as that of general conversational appropriateness. Recall that we found that greater spatial distance over time results in higher specific conversational appropriateness. Correlating specific conversational appropriateness with the social attraction range (i.e., parabola) reflects the “A” relationships depicted in Figure 4.
Logically, there are three ways to move farther apart (left to right in Figure 4) and end up out of the Danger Zone: (1) Staying within the Close-Range Zone (1a); (2) Moving into the Far-Range Zone (1b); or (3) Staying within the Far-Range Zone (1c). However, the parabola indicates that (1a), staying in the Close-Range Zone, results in a decrease in social attraction. Because a significant decrease in specific conversational appropriateness was not found Figure 4 does not include (1a). Correlating the uncertainty reduction results with the social attraction range (i.e., parabola) reflects the “1” relationships in Figure 4. Similarly, there are three ways to move closer together (right to left in Figure 4) and end up out of the Danger Zone: (1) Staying in the Close-Range Zone (2a); (2) Moving into the Close-Range Zone (2b); or (3) Staying in the Far-Range Zone (2c). However, the parabola indicates that (2c), staying in the Far-Range Zone, results in a decrease in social attraction. Because a significant decrease in uncertainty reduction was not found, Figure 4 does not include (2c). In Figure 4, the distance ranges for this study are provided (37–471 pixels). The three zones are also provided: 37–274 pixels is the Close-Range Zone; 174 to 275 pixels is the Danger Zone; and 275–471 pixels is the Far-Range Zone.
As an example, consider first the case where people move farther apart (left to right) from the Danger Zone into the Far-Range Zone (arrow 2a). In this case, the distance indicates that an individual is moving into a more socially attractive zone via conversational appropriateness. This can be explained as a strategy of increased specific conversational appropriateness. As another example, consider the case where people move closer together (right to left) into the Close-Range Zone (arrow 2b). In this case, the distance indicates that an individual is moving into a more socially attractive zone through conversational appropriateness. This can be explained as a strategy of increased uncertainty reduction (i.e., increased attributional confidence). Figure 4 thus represents a culmination of the study as it correlates the Social Attraction parabolic function illustrated in Figure 1 with the elements of the model specified in Figure 3–the general conversational appropriateness results specify the direction and change in distance, while the social attraction results specify the actual range (or zones) to which these individuals moved.
In summary, we will relate the results to the research questions posed at the outset of this paper.
RQ1. What are the distance ranges in virtual computer-mediated communication?
The results show that the distance ranges are between 37 and 471 pixels. In the context of this paper (factoring in the screen resolution and size of the window of The Palace conversation), this equates to a range from 0.76 to 8.44 inches.
RQ2. What are the delineation of distance ranges in virtual computer-mediated communication?
The results indicate that there are three ranges (1) the Close-Range Zone (37–274 pixels = 0.76–3.10 inches), (2) the Mid-Range (i.e., Danger) Zone (175–274 pixels = 3.12–4.88 inches), and (3) the Far-Range Zone (275–471 pixels = 4.90–8.44 inches).
RQ3. What are the relationships between uncertainty reduction (high-and low-context communication) and spatial distance?
Uncertainty reduction is a component of general conversational appropriateness. For the group that moved closer together over time, greater spatial distance signified higher uncertainty reduction (i.e., attributional confidence).
RQ4. What are the relationships between conversational appropriateness and increases/decreases in spatial distance?
Specific conversational appropriateness is a component of general conversational appropriateness. For the group that moved farther apart over time, greater spatial distance signified higher specific conversational appropriateness. General conversational appropriateness was found to mediate between specific conversational appropriateness and uncertainty reduction. For all groups that moved over time (i.e., closer and farther groups), greater spatial distance signified higher general conversational appropriateness. A model (see Figure 3) was developed depicting the relationship between general conversational appropriateness to specific conversational appropriateness and uncertainty reduction.
RQ5. What are the relationships between spatial distance and social attraction?
Social attraction was found to vary according to the (unstandardized) parabolic equation y = 2.55 × 10−5*(Distance-219)2+ 4.76; where y is the variable for social attraction (i.e., liking). The three zones specified in RQ2 have significant meaning when compared to liking. Close-Range Zone and Far-Range Zone spatially distant individuals were more socially attracted to their conversational partner. Mid-Range Zone (i.e., Danger Zone) spatially distant individuals were less socially attracted to their conversational partners. These results imply that there is a non-linear (i.e., parabolic) relationship between absolute distance and social attraction (see Figure 1). Finally, Figure 4 illustrated the relationship between social attraction and general conversational appropriateness.
The implications of this study bear upon four main themes. First, distance matters in a virtual computer-mediated environment. Using a tool developed to measure spatial distance in a 2-Dimensional virtual space (the Spatial Distance Analysis Program), we determined the distances between avatars in a virtual environment in The Palace. While this tool measures 2-D spatial distances, it can be easily amended to measure distances in a 3-Dimensional virtual space (e.g., Virtual Worlds). Second, we found that measuring distance over time provided insight into the non-verbal aspects of distance communication. Communication necessitates interaction over time and by examining distance changes over time, we were able to infer causal relationships between distance and communication constructs utilized in this study. Third, we recognized the importance of previously validated interpersonal communication scales in examining our theoretical constructs. The alpha coefficients for our scales indicated highly reliable measures comparable, in the most part, to reliabilities in face-to-face interpersonal contexts. Interpersonal communication theory as a measure of interaction on the Web also has great potential in examining the interactive communication network known as the Internet. Finally, we pose a footnote to our spatial distance ranges. The actual distances may vary depending on the screen resolution and size of the window for online communication. For this reason, we suggest that the actual pixel distance be applied with respect to the monitor resolution and that the size of the monitor (in inches) be factored in to determining the actual length (in inches) of spatial distances.
There are several potential limitations to this study. First, the time for the subject interaction was variable (from 7–10 minutes). This was used to ensure that a train of thought was completed and that questionnaire information was not based on unfinished conversations. Also, this time may not have been long enough to ensure that uncertainty reduction occurred between pairs of strangers. While there was evidence that distance effects stabilized over time, it would be prudent to examine these conversations in longer time frames (10–15 minutes). Next, the interface that we tested was new and had not been studied previouly. We learned from our study and encourage others to examine the effects of distances in other experimental contexts and virtual environments. As an example of these effects, we used generic “smiley face” avatars with no names, only location identifiers. Such an artifact decreases the personability and social presence in a CMC environment and may limit the amount of interaction that will occur online. The actual distances should also vary depending on the level of communication. For instance, communication among a crowded virtual group would enforce constraints on spatial movement. Nevertheless, when speaking directly to another person, dyadic distance norms would seemingly still play a significant role.
There are, of course, many directions to go in examining new communication media. One immediate next step is to examine the qualitative information gathered (i.e., experiential reports) for this study in greater detail. There is a unique richness to qualitative data and in examining discourse processes, further insight can be gained from this data. Performing a semantic network analysis on the word co-occurrences, for example, would help identify the flow of conversations and the themes in initial interactions in a virtual environment. Interaction Process Analysis (Bales, 1950) and SYMLOG (Bales & Cohen, 1979) could be used as categorization schema for classifying the types of statements in virtual interactions among groups. Also, research on the nature of uncertainty reduction among strangers (Rubin, 1979), indicates that demographic questions dominate the first third of face-to-face conversations. Such common CMC practices as asking name, age, and sex would lead to a greater understanding of the content of initial conversations. An analysis of the results using sex as a discriminating factor could also lend insight into the nature of these demographic questions.
It is necessary to develop and “refurbish” measurement scales for virtual CMC environments. Factor analyzing the scales used could help further specify the factor loadings and further refine and define what issues are most salient in a computer-mediated, virtual environment. Another approach would be to vary the experimental conditions used in this study. For instance, specifying the sex of the avatar may lend further insight into FTF communication differences and also examine the perceptions of gender swapping and gender transformation (not uncommon practices) in a virtual CMC environment (see Suler, 1996). Also, exploring 3-Dimensional spaces would also be a worthwhile endeavor. The effects of that third dimension (e.g., whether one is above or below their conversational partner) may have interesting ramifications in terms of one-up and one-down relationships (Rogers & Millar, 1979; Watzlawick, Bavelas, & Jackson, 1967). Finally, the spatial relationships between avatars in larger groups would be an interesting study, particularly in light of past research examining the amount of communication and perceptions of other with respect to distance (Allen, 1977; Mehrabian & Diamond, 1971).
Appendix 1. Spatial Distance Analysis Program
The authors would like to thank Frank Biocca and Alice Chan for the use of facilities to conduct this investigation. Special thanks to edupalace.com for providing the Palace server used in this study free of charge. Laura Weisbein and Eun Jung Lee assisted in the data collection phase of this study. All comments regarding this manuscript should be directed to email@example.com.