avatar gender and personal space invasion anxiety level in desktop collaborative virtual...

11
ORIGINAL ARTICLE Nasser Nassiri Norman Powell David Moore Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments Received: 26 February 2004 / Accepted: 9 November 2004 / Published online: 15 January 2005 Ó Springer-Verlag London Limited 2005 Abstract We report an investigation exploring the effect of avatar gender on the anxiety level caused by personal space (PS) invasion in desktop collaborative virtual environments (DCVE). We outline an experiment in which participants, of both genders, whose avatars’ PS were ‘‘invaded’’ by other avatars of either gender, re- ported their anxiety levels through the use of a post- experiment questionnaire. The data from the question- naire are analysed and discussed. The results suggest that the combination of the gender of the invading avatar and the avatar being invaded has an influence on the PS invasion anxiety level and that the ranking of gender combination groups has a striking difference from those observed for PS invasion in physical envi- ronments. Results also show that the participants in general did not register high anxiety, contrary to what one might expect from personal space invasion in the physical world. Keywords Personal space Collaborative virtual environment Avatars 1 Introduction Many researchers have defined personal space (PS) in the physical world as an area with invisible boundaries surrounding individuals, which functions as a comfort zone during interpersonal communication [13]. Per- sonal space is often referred to as ‘‘interpersonal dis- tance’’—the distance apart from each other that conversational partners adopt. Personal space invasion occurs when an individual enters another’s PS and produces different signs of discomfort and anxiety [2, 4, 5]. The concept of personal space invasion anxiety level (PSIAL) refers to the degree of anxiety generated from an invasion of someone’s PS Hall [2] identified four regions of PS: intimate (0– 45 cm), personal (45 cm–1.20 m), social (1.20–3.60 m), and public (3.60 m onward). This grouping gives a useful image of PS, but has some problems if interpreted literally because it may lead to the perception that PS is of fixed size. This is not true as individuals’ PSs tend to change based on many factors such age [6], culture [3], gender [7], relationship [8] and similarity [7]. For these reasons Aiello [3] has suggested we use the term inter- personal distance instead of PS. There is some evidence that the concept of interper- sonal distance and PSIAL exist in collaborative virtual environments. For example, Bailenson et al. [9] con- ducted an experimental study in immersive virtual environments (IVEs) and found that individuals did not violate others’ PSs. Sommer [10] and Krikorian et al. [11] also researched PS in IVEs with similar results to Bailenson et al. [9]. Becker and Mark [12] noticed that people in ActiveWorlds (AW) 1 , a desktop collaborative virtual environment, get annoyed if another avatar comes too close. Similarly, Jeffrey [13] observed AW for several weeks and noticed that individuals tend to maintain a distance between their avatars when they are interacting. Jeffery [13] also noticed that individuals tend N. Nassiri (&) BIT Department, Higher Colleges of Technology, Dubai Women’s College, Dubai, 16062, United Arab Emirates E-mail: [email protected] Tel.: +971-4-2089572 Fax: +971-4-2673939 N. Powell D. Moore ISLE Research Group, School of Computing, Leeds Metropolitan University, Leeds, UK E-mail: [email protected] Tel.: +44-113-2832600 Fax: +44-113-2833182 E-mail: [email protected] Tel.: +44-113-2832600 Fax: +44-113-2833182 1 Copyright ActiveWorlds, Inc. Virtual Reality (2005) 8: 107–117 DOI 10.1007/s10055-004-0142-0

Upload: nasser-nassiri

Post on 14-Jul-2016

216 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

ORIGINAL ARTICLE

Nasser Nassiri Æ Norman Powell Æ David Moore

Avatar gender and personal space invasion anxiety levelin desktop collaborative virtual environments

Received: 26 February 2004 / Accepted: 9 November 2004 / Published online: 15 January 2005� Springer-Verlag London Limited 2005

Abstract We report an investigation exploring the effectof avatar gender on the anxiety level caused by personalspace (PS) invasion in desktop collaborative virtualenvironments (DCVE). We outline an experiment inwhich participants, of both genders, whose avatars’ PSwere ‘‘invaded’’ by other avatars of either gender, re-ported their anxiety levels through the use of a post-experiment questionnaire. The data from the question-naire are analysed and discussed. The results suggestthat the combination of the gender of the invadingavatar and the avatar being invaded has an influence onthe PS invasion anxiety level and that the ranking ofgender combination groups has a striking differencefrom those observed for PS invasion in physical envi-ronments. Results also show that the participants ingeneral did not register high anxiety, contrary to whatone might expect from personal space invasion in thephysical world.

Keywords Personal space Æ Collaborative virtualenvironment Æ Avatars

1 Introduction

Many researchers have defined personal space (PS) inthe physical world as an area with invisible boundariessurrounding individuals, which functions as a comfortzone during interpersonal communication [1–3]. Per-sonal space is often referred to as ‘‘interpersonal dis-tance’’—the distance apart from each other thatconversational partners adopt. Personal space invasionoccurs when an individual enters another’s PS andproduces different signs of discomfort and anxiety [2, 4,5]. The concept of personal space invasion anxiety level(PSIAL) refers to the degree of anxiety generated froman invasion of someone’s PS

Hall [2] identified four regions of PS: intimate (0–45 cm), personal (45 cm–1.20 m), social (1.20–3.60 m),and public (3.60 m onward). This grouping gives auseful image of PS, but has some problems if interpretedliterally because it may lead to the perception that PS isof fixed size. This is not true as individuals’ PSs tend tochange based on many factors such age [6], culture [3],gender [7], relationship [8] and similarity [7]. For thesereasons Aiello [3] has suggested we use the term inter-personal distance instead of PS.

There is some evidence that the concept of interper-sonal distance and PSIAL exist in collaborative virtualenvironments. For example, Bailenson et al. [9] con-ducted an experimental study in immersive virtualenvironments (IVEs) and found that individuals did notviolate others’ PSs. Sommer [10] and Krikorian et al.[11] also researched PS in IVEs with similar results toBailenson et al. [9]. Becker and Mark [12] noticed thatpeople in ActiveWorlds (AW) 1, a desktop collaborativevirtual environment, get annoyed if another avatarcomes too close. Similarly, Jeffrey [13] observed AW forseveral weeks and noticed that individuals tend tomaintain a distance between their avatars when they areinteracting. Jeffery [13] also noticed that individuals tend

N. Nassiri (&)BIT Department, Higher Colleges of Technology,Dubai Women’s College,Dubai, 16062, United Arab EmiratesE-mail: [email protected].: +971-4-2089572Fax: +971-4-2673939

N. Powell Æ D. MooreISLE Research Group, School of Computing,Leeds Metropolitan University, Leeds, UKE-mail: [email protected].: +44-113-2832600Fax: +44-113-2833182E-mail: [email protected].: +44-113-2832600Fax: +44-113-2833182 1Copyright ActiveWorlds, Inc.

Virtual Reality (2005) 8: 107–117DOI 10.1007/s10055-004-0142-0

Page 2: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

to show some discomfort and anxiety feeling in theirreactions when other avatars invade their avatar PSs.

One significant influence on PS (or interpersonaldistance) in the physical world is gender, an influencewhich may or may not operate in the virtual world.Several researchers have studied extensively the effect ofgender differences on PS in the physical world [6, 14].Such studies have suggested that (a) females interact atcloser distances than male only groups [15, 16], (b)mixed-gender groups interact at closer distances thanmale only genders [17, 18], (c) women allow closer ap-proaches from others than men allow [1, 19, 20] and (d)female and mixed-gender interactants use touch morethan male only interactants [21–24].

Similarly, Hewitt and Henly [25] identified an orderfor the four gender combinations of PS invasion in thephysical world: men allow women to invade their PS tothe highest degree, followed by women allowing otherwomen to invade their PS, then men allowing men toinvade PS, and finally women allowed men to invadetheir space the least.

An important issue for collaborative virtual envi-ronments (CVE) is the extent to which these influenceson PSIAL in the physical world operate in the virtualworld. An understanding of this will have an importantbearing on how people should interact with each other inCVE. What, for example, are the appropriate distancesthat should be observed between the avatars of differentusers if an interaction free of PSIA is to take place?Despite the importance of this issue, there has been verylittle previous research addressing it. Given this, theaims of the research discussed in this paper are (a) toinvestigate the effect of avatar gender on the anxietylevel caused by PS invasion in desktop collaborativevirtual environments (DCVE) and (b) to investigate theorder of such anxiety levels of the four possible gendercombinations. In order to investigate this, an experimenthas been conducted which involves interactions betweenthe four possible combinations of genders (man–man,woman–woman, woman–man, and man–woman) in aDCVE designed for the experiment.

2 The gender experiment hypotheses

The experiment reported in this paper aims, then, toprovide evidence concerning the following issues:

– Does avatar gender affect the anxiety level caused byPS invasion in the DCVEs?

– Does the level of anxiety for the various gendercombinations have the same order as that in thephysical world?

Thus, the hypotheses of the experiment are:

1. Gender is a significant factor in determining theextent to which people using a DCVE feel anxiousbecause of PS invasion in that DCVE.

2. As in the physical world, there is a specific rank orderof PS invasion anxiety level among gender combi-

nation groups in the DCVEs, namely: male invadingfemale followed by male invading male then femaleinvading female, and finally female invading male.

3 The experiment’s variables

Since the influence of avatar’s gender on PS invasionanxiety is the subject of investigation then the indepen-dent variable for the experiment is the combination ofavatar genders. There are consequently, four values ofthis independent variable as follows: MinvM, MinvW,WinvM, and WinvW, where MinvW represents an eventwhere a man-like avatar is invading the PS of a woman-like avatar in the DCVE. The dependent variable of thisexperiment is the anxiety level that a DCVE user expe-riences when his or her avatar’s PS is invaded. Theprediction of the order of the anxiety level variable,based on physical world studies by Hewitt and Henley[25], is shown below (Table 1) where 4 is the highestanxiety level and 1 the lowest:

In the physical world, there are several factors thatinfluence PS when people are interacting. Some of thesefactors are age [6], culture [3], gender [7], relationship [8],and similarity [7]. Further, it is generally found thatwhere attraction between users is strong, the PS isdecreased [26]. Conversely, where people are unfriendlyand dislike each other, their PSs are increased i.e. thepeople move further apart. Assuming that these physicalworld phenomena apply also in the virtual world, thevariables that might potentially affect the outcome of theexperiment are the user relationship or friendship, cul-ture, age, similarity and attraction. These variablestherefore need to be controlled, given that our interest inthis experiment is gender influence.

This control was effected via careful selection of theexperimental participants. Participants were asked toplay one of two roles. Some participants (the ‘‘subjects’’)were asked to carry out a simple task in the VE (cf. Sect.7 below). Other participants (‘‘confederates’’) were givenspecific instructions by the experimenter to ‘‘invade’’ thesubjects as they (the subjects) were carrying out theirassigned task. The approach that was taken to controlthe relationship or the pre-knowledge variable betweenparticipants in the experiment was to ensure that theyhad not met prior to their meeting in the collaborativevirtual environment. This was accomplished by selectingthe confederates and the subjects from two different

Table 1 Expected anxiety level

Avatar gender combination Expected orderof anxiety level

Anxiety level of MinvW 4Anxiety level of MinvM 3Anxiety level of WinvW 2Anxiety level of WinvM 1

108

Page 3: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

departments at the American University of Beirut(AUB). Therefore, the expectation is that the experi-mental participants will have no prior relationships witheach other. Further, the identity of confederate is neverdisclosed to the participant and vice versa, so if bychance there is a prior relationship, neither the subjectnor the confederate will be aware of this during theexperiment. This can also be expected to cater for theconfounding variable of attraction or friendship, since,given that they have never met or at least are unaware ofany previous meeting, no attraction or friendship basedon previous meetings exists between the participants. Tocontrol the age and the culture variable, the subjectswere chosen to be from the same culture and with thesame range of age (university age, between 20 years and23 years). This can also be expected to cater for theconfounding variable similarity, since all the subjects arestudents, their situations are likely to be similar in manyaspects. Regarding the culture, even though there areinevitable cultural differences, the prior knowledge ofthe group by the experimenter suggests that such dif-ferences will be minimal.

4 The experiment’s virtual environment

The experiment’s environment was built using AW, adesktop collaborative virtual environment with 3D vir-tual representations of the participants in the environ-ment, called ‘‘avatars’’. AW uses text entered viakeyboard for communication. All public messages ap-pear in a scrollable window and may also appear abovethe avatar’s head for 30 s or until the next message ap-pears. Although there have been some critiques of AWin the literature [e.g. 27], this environment was chosenfor the experiment because it is widely used and itsexisting functionality and representation would providethe opportunity to observe differences in humanbehaviours when their avatars’ PSs are invaded. In AW,users are represented by full-bodied avatars that canwalk, wave, dance, and fight. They can also move in 3Dusing the arrow keys or mouse.

The experiment used a virtual environment of acompound consisting of five identical virtual fully fur-nished houses, see Fig. 1. Each house consists of a largeliving room, a fully furnished kitchen, a family room,and a bedroom. The participants can interact with theenvironment by navigating within the houses and mov-ing around the compound. They can communicate byusing a text-based chat interface, where they can typemessages and send them to their paired confederatesthrough the whispering facility provided by AW.

5 The experiment participants

The experiment involved 50 participants—40 subjectsand 10 confederates; the former were students from thesecond year psychology courses and the latter were

students from computer courses in AUB in Lebanon,but from a different department. The subjects were di-vided into eight groups of five subjects with the samegender within each group. Each group participated inonly one session. Each member of the group wasembodied into the virtual environment by using either amale avatar or female avatar depending on the subject’sactual gender. Figures 2 and Fig. 3 show the two built-in avatars in AW that were used for the subjects’ ava-tars.

The confederates were divided into two groups offive with the same gender within each group. Eachmember of the confederate group was assigned one ofthe avatars depicted in Figs. 4 and 5, based on the

Fig. 1 Interior layout of the house

Fig. 2 Male subject avatar identity

Fig. 3 Female subject avatar identity

109

Page 4: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

confederate’s actual gender. The confederates andsubjects avatars were designed to be wearing typicalAUB student dress.

6 Procedure

The experiment ran over two consecutive days andconsisted of a series of eight sessions that ran sequen-tially. Each day hosted four sessions. The duration ofeach session was around 20 min. Each session used thefive constructed virtual and identical houses discussedearlier. In each session, a specific house was assigned toeach pair of participants (a subject and a confederate),and each pair remained within their assigned house. Thegender of the groups was alternated throughout the dayto avoid any temporal effects. Each group of subjectswas called into the lab where each member was shownthe exact location of the virtual house and the name ofthe confederate avatar that was assigned to him/her. Thegroup members were shown a small demo of the virtualenvironment and how to navigate in AW, in general.Also, each group member was given his/her usernameand the name of the avatar he/she was to use and shownhow to select it from the available avatar list. Eachsession required five confederates; each confederate’savatar waited for the subject’s avatar in one of thehouses in the environment. The experiment started whenthe subjects logged in to the environment and trans-ferred the avatars in front of all the houses. Each subjectmoved his/her avatar in to the assigned house, where the

confederate avatar was waiting for him/her. The exper-imental task for the subject was to enter the house andread and count the number of occurrences of eachcharacter in three notices displayed on three signs lo-cated in the virtual house, see Figs. 6 and 7. When thesubject entered the virtual house, the confederate avatarmet the subject avatar and introduced the virtual housewith its rooms and garden to him/her. Then the con-federate avatar explained the required task to the sub-ject.

To standardize the behaviour between the confeder-ates, a script and action plan was provided that all theconfederates followed when they interacted with thesubject’s avatars. While the confederate avatar andthe subject avatar were touring the house, the confed-erate’s avatar maintained a ‘‘social distance’’ from thesubject avatar.

When the subject avatar was reading the sign, theconfederate moved his/her avatar to the ‘‘personal’’ andthen to the ‘‘intimate zone’’ of the subject from the front;see Figs. 8, 9, and 10 for subjects views of confederatesat these distances. See also Figs. 11, 12, and 13 for thecorresponding confederate views of the subjects.

Fig. 6 The signs to be read by the subjects (subject‘s view)

Fig. 5 Female confederate avatar identity

Fig. 4 Male confederate avatar identity

Fig. 7 The signs to be read by the subjects (confederate‘s view)

110

Page 5: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

7 The questionnaire

There are a number of psychological consequences thatmay arise as a result of PS invasion in the physicalworld. Some of these psychological consequences arefeeling of threat, a feeling of discomfort [28] or flirtationor attraction [29]. In light of this, a post-experimentquestionnaire was created to measure the PSI anxietylevel. It does so by addressing issues corresponding tothe concepts concerned with anxiety as identified in thephysical world: (1) threat, (2) discomfort feeling, and (3)flirtation and attraction. Each factor (i.e. threat, dis-comfort, and flirtation) was allocated three direct andthree reverse questions; reverse questions were princi-pally used to ensure that the effect of subjects responding

predominantly positively (or negatively) does not affectthe overall response to the questionnaire.

Subjects were asked to respond to the questionnaireby indicating to what extent each item described theirexperience, using a five-point Likert scale. The ques-tionnaire used is shown in Table 2. Note that the ques-tionnaire as seen by the subjects did not include thequestion type and moreover the questions were scram-bled to rule out any leading information that could af-fect the subjects’ answers.

8 Results

The data obtained from the questionnaire were recodedso that they are all pointing in the same direction. For

Fig. 8 Subject’s view of the confederate avatar at ‘‘social spacedistance’’

Fig. 9 Subject’s view of the confederate avatar at ‘‘personal spacedistance’’

Fig. 10 Subject’s view of the confederate avatar at ‘‘intimate spacedistance’’

Fig. 11 Confederate’s view of the subject avatar at ‘‘social spacedistance’’

Fig. 12 Confederate’s view of the subject avatar at ‘‘personal spacedistance’’

Fig. 13 Confederate’s view of the subject avatar at ‘‘intimate spacedistance’’

111

Page 6: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

non-threatening, comfort, and aversion type questionsthe following conversion formula were used: new va-lue=3�old value. For the other questions, the conver-sion formula used was: new value=old value�3.

Depending on the subscale factor of the question-naire, the score of 2 is interpreted as a high degree offeeling, the score of 0 is interpreted as neutral, and thescore of �2 is interpreted as a low degree of feeling. Thescore of each subsection of the questionnaire (threat,discomfort and attraction) is found by adding the re-coded scores of all the component questions. Theresulting scores range from �12 to +12, where the scoreof �12 represents extreme lack of threat, extreme com-fort and extreme aversion. The score of 0 represents‘‘neutral’’, i.e. neither threatened nor unthreatened,neither comfort nor discomfort, neither aversion norattraction. The score of +12 represents extreme threat,extreme discomfort and extreme attraction. The overallanxiety level is calculated as follows: Anxiety =threat + discomfort � attraction.

The anxiety level range, in principle, is from �36 to+36, where �36 represents an extreme non-anxiousstate, 0 represents neither anxious nor calm state, and+36 represents an extreme anxious state.

Statistical analysis was carried out in statisticalpackage for social scientists (SPSS). The descriptive

statistics of all the dependent variables are shown inTable 3. Since both the Skewness and the Kurtosis arewithin one standard error of zero, the distributions of allthe variables appear normally distributed. Histogramsdrawn with the scale of �12, +12 for threat, discomfort,and attraction shown below with the normal distributioncurve (Fig. 14, 15, 16) illustrate this. The histogram ofthe overall anxiety (Fig. 17 drawn in a scale of �36,+36) shows that anxiety is also normally distributed.The descriptive statistics of anxiety with the means and95% confidence intervals of each gender invasion groupare shown in Table 4.

The multiple comparisons based on the anxiety var-iable are shown in Table 5. Two post hoc tests are ap-plied to the differences between the anxiety levels of thegender groups; least significant difference (LSD) andBonferroni methods. The LSD method takes no accountof multiple comparisons whereas the Bonferroni methoddoes. The least significant difference criterion shows thatadjacent groups are not significantly different. However,the groups that are more than one position away in theorder are significantly different. In the Bonferronimethod, only the highest and lowest groups have sig-nificantly different means.

The one-way ANOVA procedure (Table 6) showsthat there is a statistically significant difference in anxi-

Table 2 Post-experiment questionnaire

Question Type

1 Seeing the guide close by gave me a sense of assurance Reverse threat (no threat)2 I felt self-confident when interacting with the avatar guide at a close distance Reverse threat (no threat)3 The avatar guide helped my self-confidence whenever they approached me Reverse threat (no threat)4 I dreaded it whenever the guide avatar moved towards me Direct threat (threat)5 It made me feel apprehensive when I saw the guide avatar close by Direct threat (threat)6 I felt scared when the guides’ avatar was in the same hall Direct threat (threat)7 I was comforted when the avatar guide spoke to me Reverse discomfort (comfort)8 It was a relief to find the avatar guide next to me Reverse discomfort (comfort)9 The avatar guides provided me with support when they approached me Reverse discomfort (comfort)10 I was anxious when the avatar guide was close to me Direct discomfort (discomfort)11 The avatar guides made me feel nervous when they approached me Direct discomfort (discomfort)12 I felt uneasy being next to the guide avatar Direct discomfort (discomfort)13 I felt excited when the guide avatar came towards me Direct attraction (attraction)14 It was exciting to see the guide avatar Direct attraction (attraction)15 The guide avatar was an interesting person to talk to Direct attraction (attraction)16 It was irritating when the guide avatar was close to me Reverse attraction (aversion)17 I felt annoyed when the guide avatar stood next to me Reverse attraction (aversion)18 The guide avatar’s behaviour is frustrating because it seemed that I am attractive to them Reverse attraction (aversion)19 Do you have any comments on any aspect of the work you have just been doing Qualitative

Strongly disagree=1, slightly disagree=2, do not know=3, slightly agree=4, strongly agree=5

Table 3 Descriptive statistics

Number Minimum Maximum Mean SD Skewness Kurtosis

Statistic SE Statistic SE

Anxiety 40 �25 17 �3.95 9.39 �0.077 0.374 �0.251 0.733Discomfort 40 �10 7 �1.88 3.94 0.388 0.374 0.141 0.733Threat 40 �9 5 �1.20 3.51 �0.065 0.374 �0.474 0.733Attraction 40 �8 9 0.88 4.55 �0.302 0.374 �0.643 0.733

112

Page 7: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

ety level among the gender combinations (F=4.599,significance=0.008). This supports the hypothesis thatgender affects the anxiety level caused by PS invasion inDCVEs. The graph for the anxiety level split by gendercombination (Fig. 18) shows that the anxiety level of thepair MinvM is the highest followed by WinvW, thenWinvM, and finally MinvW.

9 Discussion

Themethod we used tomeasure the PSIAL in theDCVEsin this experiment was a subjective method through theuse of a bespoke questionnaire, partly on the grounds that

there is good precedent for the use of an experiment fol-lowed by questionnaire when measuring a psychologicalfeeling in virtualworlds [30–32].Althoughno reliability orvalidity checks were carried out on the items in the scalebefore the experiment was carried out, a post hoc reli-ability measurement was made of the items, using Cron-bach’s Alpha; this yielded a result of 0.71, suggesting anacceptable level of reliability (>0.70) [33].

However, it might be argued that a more appropriateapproach would be physiological measurements (heartrate, blood pressure level, voice tone), on the groundsthat medical measurement devices give accurate andobjective results. However, this approach was not con-sidered appropriate because it is hard to interpret thedata from such measurement, since all that is shown isthat there have been certain physical changes, but whatprecisely these changes indicate (e.g. excitement or fear)is debatable.

Fig. 14 Threat histogram

Fig. 15 Discomfort histogram

Fig. 16 Attraction histogram

Fig. 17 Anxiety histogram

113

Page 8: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

The statistical analysis reported in the previous sec-tion (cf. Sect. 8) supports hypothesis 1—gender doesindeed appear to be a significant factor in determiningthe extent to which people using DCVE feel anxious inthe event of an invasion of their PS. This is establishedby the ANOVA results (Table 6) and in particular fromthe LSD figures in Table 5, in which, for example, it canbe seen that the MinvM response is significantly differ-ent from the WinvM and MinvW responses. Thus re-sults suggest that avatar gender is an important factor in

PS in DCVEs, in that it has an effect on the level ofanxiety caused by PS invasion.

Interestingly, however, the results suggest a differentorder of gender combinations from those found in thephysical world by [25]. Our results do not, therefore,support hypothesis 2. Figure 18 shows that the rankorder of degrees of anxiety amongst participants in ourDCVE experiment from highest degree of anxiety tolowest degree of anxiety was MinvM, followed byWinvW, then WinvM, and finally MinvW. In particular,unlike the physical world, in which a male invasion of afemale’s PS is highest in terms of anxiety level, in ourDCVE environment it has turned out to be the lowest.Indeed, the anxiety level value for MinvW is negativesuggesting that the participants being invaded are feelingthe opposite of anxiety.

Another unexpected and important finding from ourstudy is that statistics for three of the four gendercombination groups show average anxiety levels that are

Table 4 Descriptive statistics ofanxiety and gendercombination

Number Mean SD SE 95% Confidence interval formean

Minimum Maximum

Lower bound Upper bound

M–M 10 2.10 8.36 2.64 �3.88 8.08 �11 17W–W 10 �0.90 7.81 2.47 �6.49 4.69 �14 13W–M 10 �6.50 10.80 3.42 �14.23 1.23 �25 8M–W 10 �10.50 5.36 1.69 �14.33 �6.67 �21 �4Total 40 �3.95 9.39 1.48 �6.95 �0.95 �25 17

Table 5 Multiple comparisons

*The mean difference is signifi-cant at the 0.05 level

(I) Group (J) Group Mean difference (I–J) SE Sig. 95% Confidence interval

Lower bound Upper bound

LSDMinvM WinvW 3.00 3.72 0.425 �4.54 10.54

WinvM 8.60* 3.72 0.027 1.06 16.14MinvW 12.60* 3.72 0.002 5.06 20.14

WinvW MinvM �3.00 3.72 0.425 �10.54 4.54WinvM 5.60 3.72 0.141 �1.94 13.14MinvW 9.60* 3.72 0.014 2.06 17.14

W-M MinvM �8.60* 3.72 0.027 �16.14 �1.06WinvW �5.60 3.72 0.141 �13.14 1.94MinvW 4.00 3.72 0.289 �3.54 11.54

MinvW MinvM �12.60* 3.72 0.002 �20.14 �5.06WinvW �9.60* 3.72 0.014 �17.14 �2.06WinvM �4.00 3.72 0.289 �11.54 3.54

BonferroniMinvM WinvW 3.00 3.72 1.000 �7.38 13.38

WinvM 8.60 3.72 0.159 �1.78 18.98MinvW 12.60* 3.72 0.010 2.22 22.98

WinvW MinvM �3.00 3.72 1.000 �13.38 7.38WinvM 5.60 3.72 0.844 �4.78 15.98MinvW 9.60 3.72 0.084 �0.78 19.98

WinvM MinvM �8.60 3.72 0.159 �18.98 1.78WinvW �5.60 3.72 0.844 �15.98 4.78MinvW 4.00 3.72 1.000 �6.38 14.38

MinvW MinvM �12.60* 3.72 0.010 �22.98 �2.22WinvW �9.60 3.72 0.084 �19.98 0.78WinvM �4.00 3.72 1.000 �14.38 6.38

Table 6 The ANOVA for the anxiety level

Sum ofsquares

df Meansquare

F Significance

Between Groups 953.100 3 317.700 4.599 0.008Within Groups 2486.800 36 69.078Total 3439.900 39

114

Page 9: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

not significantly different from neutral see Table 4 andFig. 18. The only group that does show a significantdifference from a neutral response is the group MinvWwith a mean of �10.50, which is an indication that thisgroup’s reaction is the opposite to anxiety. All thissuggests that the participants are tending to be neutralabout their space being invaded, or even positive aboutit. It should be noted, however, that these overall figuresmay be masking large variations within the participants.In fact, there is a standard deviation of 9.4 for the wholegroup and a range of �25 to 17, and this indicates thepresence within the group of individuals feeling stronglyanxious and individuals feeling greatly relaxed or non-anxious. These strong feelings of anxiety are also dem-onstrated by some of the verbal comments made to theexperimenter. For example, one of the female partici-pants was annoyed and complained to the experimenterbecause a male confederate avatar was showing hersome sort of admiration and she said ‘‘he liked mydress’’. Similarly, a male subject said to the experimenter‘‘I did not understand what that lady wanted from me, shewas approaching me very closely, I really was scared’’.

Caution is needed when considering the above re-sults, for at least three reasons. First, because our datacollection method involved a post-experiment question-naire, participants were not able to express their feelingat the time of the invasion, rather they reported it afterthe experiment, and as such they may have forgotten theexact nature of the feeling when it occurred. Neitherwere they able to write how they felt at different dis-tances (social, personal, and intimate) from the confed-erate avatar. Second, it might be argued that using oneconfederate instead of ten might provide stronger re-sults, on the grounds that using more than one confed-erate brings the danger that the confederates are likely toexhibit different behaviours. However, the use of oneconfederate would itself bring concerns of different

behaviour, since the putative confederate may forexample become fatigued or bored, and therefore behavedifferently. Further, all confederates were given a scriptto follow, with a view to securing uniform behaviourduring the interaction.

A third reason for caution concerns the features ofthe specific DCVE that we used in our study. The AW,unlike the real world, has no collision detection facilityand thus allows an avatar to travel unhindered throughthe space apparently being occupied by another avatar.Indeed, ‘‘collisions’’ did occur during the experiment,when the confederate’s avatar invaded the intimate PSof the subject’s avatars, as it is impossible to control theavatar steps when the keyboard forward/backwardarrows are used. Figure 19 shows a situation where anavatar is standing in the PS of another avatar. Figure 20shows part of avatar eye and part of the scene behind theavatar after only one single hit of the forward arrow keyfrom the position depicted in Fig. 19. When this situa-tion occurred, the confederate avatar went back andforth until the intimate view was achieved, and thismight require several moves. It could be that this situ-ation reduced the degree of immersion felt by the par-ticipants in the DCVE and therefore lead to an absenceof anxiety.

Further, AW users can communicate only throughtext messages —voice messages are not supported. Thismight also affect the results as the subject might be busywith typing at the keyboard when the confederate’savatar was invading his/her avatar PS and thus may notnotice such an invasion, especially if the subject’s touchtyping skill is poor. The switching of attention betweenthe screen and the keyboard may also reduce the par-ticipants’ degree of immersion in the environment.Again, the interaction model of AW does not supportfacial expression nor voice messages, which are alsoessential factors that transmit many signals in the

Fig. 18 Anxiety means and95% confidence intervals forgender combinations

115

Page 10: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

physical world [34]. This absence of representing suchimportant factors (facial expression and voice messages)in the interface of AW might also affect the results of theexperiment.

Our results may to some extent, therefore, be peculiarto AW in particular, rather than DCVE in general.However, the results remain relevant to any DCVE thatadopts similar interface policies to AW. Further, evi-dence from the qualitative questionnaire data suggeststhat the environment was perceived as realistic, forexample ‘‘I felt myself like in a real environment’’, ‘‘lookslike real’’, and ‘‘impressive environment’’.

As well as being of intrinsic interest, our results maysuggest implications for the design of future collabora-tive virtual systems. In the physical world people aregenerally well-rehearsed at avoiding invasions of thespace of others, and at taking evasive action if their ownspace is invaded, by for example turning their orienta-tion away from people who approach too close from thefront, or moving physically away. Various analogieshave been used to describe PS, such as a bubble [35] oran electrical field [34]. Argyle [36] identified the shape ofthe PS as almost circular but with more space in front.

In addition to people’s individual ability to managePS issues, certain aspects of design in the physical world

take account of PS issues. Wiles [37], for example, founda relationship between the PS allocated and the antici-pated time of the event; thus, telephone booths areallocated small space because the time we spend in themis little, whereas homes for the elderly and prisons areallocated much larger spaces. Further, interpersonaldistance can influence the level of understandingbetween interactants. Latane et al. [38] found that thefurther individuals are away from the source of com-munication the less influence it has over them. In med-ical clinics, people are more forthcoming when there isabout 5 ft space between themselves and the therapist.Moreover, Sommer and Ross [39] found that altering thelayout of chairs from being in rows in a clinic to being incircles increases the level of interactions between thepatients.

If collaborative virtual environments are to be suc-cessfully used, we may need to reflect these twoapproaches to PS management—reliance on the indi-vidual and supportive environmental design—in thevirtual environments. Our results suggest that this willnot be easy, partly because PSI in a virtual world maynot straightforwardly parallel PSI in the physical world,and partly because there appear to be large variations inindividual responses to invasions of PS. Our results alsosuggest that means of fine grained and easy avatarmovement should be provided to users, so that (a) theycan easily, e.g. via one mouse click, move their avataraway to adopt a new comfortable interpersonal distance,(b) they do not inadvertently invade the PS of otherusers, and (c) they can not pass through the avatarsof other users. The collision detection feature isimplemented already in some DCVEs such as OnLiveTraveler (http://www.digitalspace.com/traveler/) andMASSIVE from Nottingham University.

10 Conclusions and further work

The findings from the experiment outlined in this papersuggest that avatar gender combination had an influenceon the PS invasion anxiety level in the desktop collab-orative virtual environments, the ranking of avatargender combination groups had a striking differencefrom those observed for PS invasion in physical envi-ronments, and the participants in general did not registerhigh anxiety as might be expected from PS invasion inphysical environments. Future work will involve similarexperiments with the addition of open-ended interviews,with a view to getting deeper insights into the views ofthe participants with regards to personal space invasionanxiety.

The absence of facial expression, voice messages, andcollision detection facilities from the interface of AWmight affect the results of the experiment. Therefore, ourimmediate future work is to conduct similar experi-mental work in a different DCVE, with a view toestablishing whether the presence of facial expressionsand voice messaging in the DCVE interaction model

Fig. 19 View at the personal space

Fig. 20 The next view after one click

116

Page 11: Avatar gender and personal space invasion anxiety level in desktop collaborative virtual environments

affects levels of PSIA. Another aspect of future work isto investigate how issues of PS are dealt with in naturallyoccurring interactions in the virtual worlds, over a per-iod of time.

Further investigations of means of designing DCVEto minimise invasions, and of enabling users to handlesuch invasions as do occur, are also planned.

Acknowledgements The authors would like to thank Dr. AhmadNasri and his students at the American University of Beirut inpreparing the participants and the computer labs required for theexperiment. Also, the authors are grateful to Dr. Janet Finlay andMr. John Gray for stimulating and helpful discussions of earlierdrafts.

References

1. Dosey M, Meisels M (1969) Personal space and self-protection.J Pers Soc Psychol 11:93–97

2. Hall E (1959) The silent language. Doubleday, NY3. Aiello J (1987) Human spatial behaviour handbook of envi-

ronmental psychology. Wiley, New York4. Albert S, Dabbs J (1970) Physical distance and persuasion. J

Pers Soc Psychol 15:265–2705. Aiello J, Thompson D (1980) Personal space, crowding and

spatial behaviour in a cultural context in Altman. Rapoportand Wohlwill 107–178

6. Hayduk L (1983) Personal space: where we now stand. PsycholBull 94:293–335

7. Gifford R (1996) Environmental Psychology, 2 edn. Allyn andBacon, Boston

8. Allegier A, Byrne D (1973) Attraction towards the opposite sexas a determinant of physical proximity. J Soc Psychol 90:213–219

9. Bailenson J, Blascovich J, Loomis J (2001) Equilibrium theoryrevisited: mutual gaze and personal space in virtual environ-ments. Presence 10:583–597

10. Sommer R (2002) Personal space in a digital age. In: BechtelRB, Churchman A (ed) Handbook of environmental psychol-ogy. Wiley, New York

11. Krikorian D, Lee J, Chock T, Harms C (2000) Isn’t that spa-tial? Distance and communication in a 2-D virtual environ-ment. J Comput Med Commun [online] 2000:4

12. Becker B, Mark G (1998) Social conventions in collaborativevirtual environments. In: CVE 98 Manchester

13. Jeffrey P (1998) Personal space in a virtual community. In:Human factors in computing systems (CHI ’98 Summary), LosAngeles

14. Burgoon J, Buller D, Woodall G (1998) Nonverbal communi-cation: the unspoken dialogue. Harper and Row, New York

15. Aiello J, Jones E (1971) Field study of the proxemic behaviourof young school children in three subculture groups. J Pers SocPsychol 19:351–356

16. Mehrabian A, Diamond S (1971) Seating arrangement andconversation. Sociometry 34:281–289

17. Baxter J (1970) Interpersonal spacing in natural settings.Sociometry 33:444–456

18. Cook M (1970) Experiments on orientation and proxemic.Hum Relat 23:61–76

19. Patterson M, Edinger J (1987) A functional analysis of space insocial interaction. In: Siegman AW, Feldstein S (eds) Nonver-bal behaviour and communication. Erlbaum, Hillsdale, pp 523–561

20. Willis E (1966) Initial speaking distance as a function of thespeaker’s relationship. Psychon Sci 5:221–222

21. Elzinga R (1975) Nonverbal communication: body accessibilityamong the Japanese. Psychologia 18:205–211

22. Hall J, Veccia E (1990) More touching observations: new in-sights on men, women, and interpersonal touch. J Pers SocPsychol 59:1155–1162

23. Henley N (1973) Status and sex: some touching observations.Bull Psychon Soc 2:91–93

24. Jones S (1971) A comparative proxemic analysis of dyadicinteraction in selected subcultures of New York City. J SocPsychol 84:35–44

25. Hewitt J, Henly R (1987) Sex differences in reaction to spatialinvasion. Percept Motor Skill 64:809–810

26. Little K (1965) Personal space. J Exp Soc Psychol 1:237–24727. Schroeder R (2002) The social life of avatars. Springer, London28. Patterson M, Mullens S, Romanao J (1971) Compensatory

reactions to spatial intrusion. Sociometry 34:114–12129. Abbey A, Harnish R (1995) Perception of sexual intent: the role

of gender, alcohol consumption, and rape supportive attitudes.Sex Roles 32:297–313

30. Gerhard M (2003) A hybrid avatar/agent model for educa-tional collaborative virtual environments. PhD Thesis, LeedsMetropolitan University

31. Fabri M, Hobbs D, Moore D (2002) Emotive signals for virtualworlds. In: HCI2002, conference proceedings, Sept 2002,London

32. Slater M, Pertaub D, Steed A (1999) Public speaking in virtualreality: facing an audience of avatars. IEEE Comput GraphicAppl 19:6–9

33. Nunnally JC (1967) Psychometric theory. McGraw-Hill, NewYork

34. Knapp M (1978) Nonverbal communication in human inter-action, 2nd edn. Holt Rinehart and Winston Inc, New York

35. Knowles E, Kreuser B, Haas S, Hyde M, Schuchart G (1976)Group size and the extension of social space boundaries. J PersSoc Psychol 33:647–654

36. Argyle M (1988) Bodily communication, 2nd edn. Methuen,New York

37. Wiles J (1978) Reassessing personal space in classroom.Southern J Educ Res 12:111–114

38. Latane B, Liu J, Nowak A, Bonevento M, Zheng L (1995)Distance matters: physical space and social impact. Pers SocPsychol Bull 21:795–805

39. Sommer R, Ross H (1958) Social interaction in geriatrics ward.Int J Soc Psychiatry 4:128–133

117