Computers in Human Behavior 27 (2011) 2380–2385

Contents lists available at SciVerse ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier .com/locate /comphumbeh

Social evaluations of embodied agents and avatars

Rosanna E. Guadagno a,⇑, Kimberly R. Swinth b, Jim Blascovich b

a Department of Psychology, University of Alabama, United Statesb Department of Psychology, University of California, Santa Barbara, United States

a r t i c l e i n f o

Article history:Available online 19 August 2011

Keywords:Social interactionEmbodied agentsAvatarsFacial expressionsCollaborative virtual environmentsNonverbal behavior

0747-5632/$ - see front matter � 2011 Elsevier Ltd. Adoi:10.1016/j.chb.2011.07.017

⇑ Corresponding author. Address: Department oAlabama, P.O. Box 870348, Tuscaloosa, AL 35487-034348 7803.

E-mail address: Rosanna@ua.edu (R.E. Guadagno).

a b s t r a c t

The purpose of this study was to examine social evaluations (i.e., perceptions of empathy and positivity)following peoples’ interactions with digital human representations. Female research participantsengaged in a 3-min interaction while immersed in a 3-D immersive virtual environment with a ‘‘peercounselor.’’ Participants were led to believe that the peer counselor was either an embodied agent (i.e.,computer algorithm) or an avatar (i.e., another person). During the interaction, the peer counselor eithersmiled or not. As predicted, a digitally-rendered smile was found to affect participants’ social evaluations.However, these effects were moderated by participants’ beliefs about their interaction partner. Specifi-cally, smiles enhanced social evaluations of embodied agents but degraded them for avatars. Althoughthese results are consistent with other findings concerning the communicative realism of embodiedagents and avatars they uniquely demonstrate that people’s beliefs alone, rather than actual differencesin virtual representations, can impact social evaluations.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction In the current study, female research participants engaged in a

Facial expressions play an important role in communication andsocial interaction. Facial displays express participants’ emotionalstates, involvement, responsiveness, understanding, and validatetheir thoughts and feelings (Bavelas, Black, Lemery, & Mullett,1986; Fridlund, Ekman, & Oster, 1987). As a pervasive facial expres-sion, smiling has received considerable attention in the literature.The ability to smile and to recognize a smile is well-developed earlyin life (Srofe & Waters, 1976). Moreover, a smile is a readily under-stood gesture of friendliness (Thompson & Meltzer, 1964) and is animportant antecedent to interpersonal attraction (Byrne, 1971).Smiling affects the ways individuals are perceived and evaluatedby others (Lau, 1982; Reis et al., 1990). There are many types ofsmiles and subtly different ones are associated with different feel-ings and functions. Ekman and Friesen (1982), for example, distin-guished between Duchenne (i.e. genuine) and non-Duchenne (i.e.false) smiles and noted that Duchenne and non-Duchenne smilesinvolve different facial muscles. Although research examining theeffect of different kinds of smiles on social evaluations is limited,Frank, Ekman, and Friesen (1993) found that individuals can reli-ably distinguish between Duchenne and non-Duchenne smilesand that Duchenne smiles are associated with more positiveimpressions and evaluations of others.

ll rights reserved.

f Psychology, University of8, United States. Tel.: +1 205

short session with a peer counselor within an immersive virtualenvironment. Participants’ beliefs about their interaction partnerwere manipulated. Half were told that they were interacting inreal-time with a digital representation of one of our undergraduateresearch assistants. In this condition, both the participant and re-search assistant were represented by a 3-dimensional virtual hu-man or avatar, which is a digital representation of another actualperson in real time. The other half were led to believe that their peercounselor was a computer-generated ‘‘embodied agent’’ (a digitalrepresentation of a computer algorithm designed to look like a per-son; Blascovich et al., 2002). During the interaction, the digital peercounselor either smiled when it was socially appropriate to do so, ornever smiled. The digitally rendered smiles were animated so as toapproximate a Duchenne smile. Following the interaction, researchparticipants evaluated their peer counselor on a number of dimen-sions including perceived empathy and positivity.

1.1. Study goals

Although there is much research examining how communica-tive realism – the extent to which a virtual human acts like a realperson – affects the way people evaluate and respond to virtual hu-mans (Bailenson, Blascovich, Beall, & Loomis, 2003; Bailenson et al.,2005; Baylor & Soyoung, 2009; Blascovich, 2002; Blascovich &Bailenson, 2011; Fridlund, 1991; Garau, 2003; Guadagno, Blasco-vich, Bailenson, & McCall, 2007; Von der Pütten, Krämer, Gratch,& Kang, 2010), most previous research examining communicativerealism has focused on the role of eye gaze or head movements.Much less is known about how the inclusion of facial non-verbal

R.E. Guadagno et al. / Computers in Human Behavior 27 (2011) 2380–2385 2381

behaviors affect evaluations of embodied agents compared to ava-tars. Thus, this study adds to the existing literature by examining asocially-important non-verbal behavior, namely smiling, and com-pares people’s reactions to identical digitally-rendered smiles thatare exhibited by embodied agents and avatars.

1.2. Predictions

Based on previous research examining the effect of smiling on so-cial evaluations (Lau, 1982; Reis et al., 1990; Srofe & Waters, 1976;Thompson & Meltzer, 1964), we expected participants to evaluatevirtual humans more positively when the latter smiled than whenthey did not. We also expected an interaction between the presenceof smiles on the part of virtual humans and participants’ beliefsregarding the virtual humans with whom they interacted.

Previous research has shown that communicative realismmanipulations such as facial expressions and other non-verbalbehaviors such as mimicry (Bailenson, Yee, Patel, & Beall, 2008;Stel et al., 2010) sometimes affect people’s perceptions and evalu-ations of embodied agents and avatars differently (Bailenson et al.,2003; Guadagno et al., 2007). As a result of these prior researchfindings, we were interested in examining the effect of smilingand agency beliefs (i.e., beliefs about whether one’s virtual partnerwas a human or computer) on social evaluations such as perceivedempathy and positivity. Given its social nature (Fridlund, 1991),smiling may have a greater impact on people when they believethey are interacting with another person, an avatar, relative to acomputer, an agent. Thus, we might expect smiling to result inmore positive evaluations for avatars. However, research has alsoshown that the fidelity of representations of actual others entersthe mix (Slater & Steed, 2001). Because people may be more sensi-tive to the slightest imprecision in animations of avatars comparedto agents, we might expect even slightly imperfect renderings ofavatars’ facial expressions to elicit less positive or even negativefeelings. Given these competing hypotheses that we examinedthe interaction between smiling and agency beliefs without mak-ing an a priori directional hypothesis in terms of smiling effectson interactants by avatars vs. agents.

2. Method

2.1. Participants

Participants were 38 undergraduate women whose average agewas M = 20.2 (SD = 1.55). They received psychology course credit orwere paid for their participation. Owing to women’s greater sensi-tivity to non-verbal behavior (Hall, 1978), we only used femaleparticipants. This provided us with a methodologically cleanesttest of our hypotheses.

2.2. Design

A 2 (Type of Interaction Partner: agent vs. avatar) X 2 (SmileCondition: present vs. absent) between-subjects factorial designwas employed. Type of interaction partner referred to participants’beliefs about whether their interaction partner was a computeralgorithm (embodied agent) or an actual person in real time (ava-tar). The smile condition, reflected whether the virtual peer coun-selor smiled (smiles present) or not (smiles absent).

2.3. Procedure

Participants were greeted by a female experimenter, blind tocondition, and escorted to a private room. After providing consent,participants completed the first part of the study on a desktop

computer. Participants read information on the computer researchinvolving digital virtual humans and the difference between anavatar and an embodied agent was explained. They were also in-formed that immersive virtual environment technology (IVET)had advanced such that it was difficult for people to tell whetherthey were interacting with a computer or another person.

Participants were told that they would be discussing a personaltopic with a digital representation of a peer counselor in an immer-sive virtual world and that following their interaction they wouldprovide feedback about their partner and the interaction. Theywere presented with three discussion topics (aspects of yourselfthat make feel you uncomfortable or embarrassed; an event thatdamaged your sense of self-worth; or problems with a past or cur-rent relationship) and were asked to rank them in the order of dis-cussion preference. Everyone discussed her second-choice topic.Next, participants were given 3 min to prepare by identifying theirdiscussion points.

Participants assigned to the avatar condition were told thattheir peer counselor, ‘‘Beth,’’ was an undergraduate research assis-tant who would join them in a virtual world and that both of themwould be represented by a 3-dimensional virtual human duringtheir real-time interaction. Participants in the embodied agent con-dition, were told that they would be interacting with a peer coun-selor named ‘‘Beth’’ and they were asked to imagine what it wouldbe like visit a counseling-oriented website to discuss theirthoughts and feelings with a computer-generated counselor wholooked and acted like Beth. Regardless of condition, all participantsactually interacted with a human research assistant, blind to con-dition, who was trained to respond in a consistent manner duringthe interactions.

After the partner type manipulation, participants were broughtinto another room for the peer counseling session. To provide theparticipant with privacy, the experimenter left the room after help-ing her don a head-mounted display. During the session, the re-search assistant ‘‘smiled’’ via the digital representation when itwas socially appropriate to do so by pushing gamepad button. Inthe smiles enabled condition, the button triggered a smile responseon the virtual human. In the smile disabled condition, nothing hap-pened when she pushed the button. After the interaction, partici-pants filled out post-interaction measures using RiddleMeThis(Loewald, 2009), were debriefed, and released.

2.4. The virtual environment

All participants were immersed in the same 3-dimensional vir-tual world for their interaction and they all interacted with thesame peer counselor. The virtual environment and virtual humanbodies were created using 3D Studio Max and Blaxxun Avatar Stu-dio, respectively (see Fig. 1). The heads and faces of the virtual hu-mans were created using 3D Me Now Software (see Fig. 2). Theselected face was rated as above average on a variety of interper-sonal traits (i.e., warm, kind, supportive, open-minded, attractive,intelligent, etc.) during pre-testing.

2.5. Equipment

Participants and the research assistant donned Virtual ResearchV8 stereoscopic head mounted displays (HMDs). The HMDs fea-tured dual 680 horizontal by 480 vertical pixel resolution LCD pan-els that refreshed at 60 Hz. The display optics presented a visualfield subtending approximately 50 degrees horizontally by 38 de-grees vertically. Perspectively correct stereoscopic images wererendered by a 450 MHz Pentium III dual processor computer withan Evans & Sutherland Tornado 3000 dual pipe graphics card andwere updated at an average frame rate of 36 Hz. The simulatedviewpoint was continually updated as a function of the

Fig. 1. The virtual environment.

Fig. 2. The peer counselor without (left panel) and with (right panel) a smile.

Fig. 3. Mean counselor empathy ratings as a function of interaction partner type(avatar vs. embodied agent) and smile condition.

2382 R.E. Guadagno et al. / Computers in Human Behavior 27 (2011) 2380–2385

participants’ head movements. The orientation of the participant’shead was tracked by a three axis orientation sensing system (Inter-sense IS300, update rate of 150 Hz). The system latency was 65 ms.The software used to render and track was Vizard 2.0.

2.6. The interaction

Once immersed in the virtual environment, the participant wasintroduced to her peer counselor who was seated in the virtualworld facing her. The peer counselor greeted her by name and in-formed her that she could begin talking about her topic when shewas ready. The participant then talked about her thoughts andfeelings about her topic for 3 min. During the interaction, the lipsof the virtual humans moved in sync with the participant’s andpeer counselor’s speech and the eyes blinked in a realistic manner.Their head movements were veridically rendered based on theiractual head movements. Other facial expressions (i.e., mouth andeye area movements) were explicitly manipulated. The peer coun-selor maintained eye contact throughout the interaction, noddedencouragingly, and provided verbal feedback (e.g., ‘‘I see’’; ‘‘goon’’). If necessary, minimal additional verbal prompts facilitatedfurther self-disclosure (e.g., ‘‘how did that make you feel?’’).

2.7. The smile manipulation

Smiles were animated so that they had a gradual onset and off-set and lasted for 1 s. On average, the research assistants smiledthree times per interaction. The smiles were programmed to be

as realistic and natural as possible since the goal was to create aDuchenne smile. Thus, the corners of the mouth were pulled up-wards, mimicking the movement of the zygomatic major muscle,and the eyes narrowed slightly to approximate the movement ofthe orbicularis oculi muscle (see Fig. 3).

2.8. Dependent measures

2.8.1. Social evaluationsParticipants evaluated the peer counselor on 19 interpersonal

traits using a 7-point scale where �3 = strongly disagree and+3 = strongly agree. An exploratory factor analysis revealed thatthe 19 traits loaded onto two factors, thus two composite scaleswere created which we labeled: counselor empathy and generalpositivity. Counselor empathy was comprised of 12 items: sup-portive, responsive, warm, sympathetic, caring, open, accepting,genuine, distant, uninvolved, judgmental, and cold, a = .90. GeneralPositivity was comprised of 7 items: friendly, attractive, intelligent,professional, polite, likeable, and open-minded, a = .82).

2.8.2. Other dependent variablesParticipants also rated the peer counselor along a number of

other dimensions on a 7-point scale where �3 = strongly disagreeand +3 = strongly agree. Four items (a = .86) assessed the peercounselor supportiveness. Two items (a = .83) assessed peer coun-selor likeability. Three items (a = .88) assessed comfort with theinteraction. Interaction satisfaction and interaction enjoyablenesswere both assessed by 1 item. Six items (a = .86) assessed socialpresence. Nine items (a = .90) peer counselor trustworthiness. Fouritems (a = .86) assessed participants’ satisfaction interaction mode.

3. Results

All of the dependent variables showed strong, positive correla-tions with one another (See Table 1). We first examined our depen-dent measures to determine whether topic discussed or researchassistant acting as the peer counselor differed by session andexperimental condition. There were no significant differences bycondition or session for the topic discussed or the research assis-tant portraying the peer counselor. Thus, we were assured thatthere were no experimenter effects or differential discussion topicsacross the experimental design.

3.1. Social evaluations

In order to examine the impact agency and smiling behavior onsocial evaluations, we conducted a multivariate analysis of variance(MANOVA) using interaction partner type and smile condition as theindependent variables and counselor empathy and general

Table 1Means, standard deviations, and inter-correlations for dependent variables.

1 2 3 4 5 6 7 8 9 10

1. Counselor empathy – .75*** .73*** .68*** .76*** .79*** .62*** .71*** .53*** .69***

2. General positivity – .52*** .73*** .56*** .54*** .47** .62*** .31+ .49 **

3. Supportiveness – .72*** .63*** .73*** .56*** .55*** .59*** .47**

4. Liking – .64*** .72*** .53*** .56*** .34* .58***

5. Comfort – .76*** .70*** .79*** .71*** .80***

6. Copresence – .69*** .68*** .56*** .79***

7. Interaction satisfaction – .67*** .56*** .71***

8. Interaction enjoyment – .55*** .68***

9. Trust & openness – .51*

10. Satisfaction w/medium –

M .15 1.00 .82 .24 .26 �.38 �.21 .32 .14 �.29SD 1.18 .79 1.38 1.36 1.74 1.41 1.77 1.40 1.36 1.60

All variables were measured on 7-point scales ranging from �3 (strongly disagree) to +3 (strongly agree). Thus, positive scores indicate agreement and negative scoresindicate disagreement.

+ p < .10.* p < .05.

** p < .01.*** p < .001.

R.E. Guadagno et al. / Computers in Human Behavior 27 (2011) 2380–2385 2383

positivity as the dependent variables. Results of this analysis re-vealed no significant main effects for either interaction partner typeor smile condition. The interaction of partner type by smile conditionwas significant, Wilks’ K = .78, F (2, 33) = 4.55, p < .05, g2

p = .22. Inaddition, there was a significant univariate interaction for coun-selor empathy, F (1, 34) = 6.28, p < .05, g2

p = .16, but not positivity.Because counselor empathy and general positivity were

strongly correlated with one another, follow-up tests to comparethe effect of interaction partner type and smile condition on coun-selor empathy were conducted controlling for general positivity. Aswith the MANOVA results, a univariate analysis of covariance (AN-COVA), with general positivity as the covariate, revealed a signifi-cant interaction partner type by smile condition interaction forcounselor empathy, F (1, 33) = 8.36, p < .01, g2

p = .21. As seen inFig. 3, simple effects comparing the effect of smiling on counselorempathy separately for avatars and embodied agents revealed thatwhen participants thought that they were interacting with anotheractual person, the peer counselor was viewed as more empathicwhen she did not smile (M = .72, SD = .86) compared to when shedid smile (M = �.13, SD = .85), F (1, 33) = 5.88, p < .05, g2

p = .15. Forparticipants who thought they were interacting with a computer,the peer counselor was seen as somewhat more empathic whenshe smiled (M = .30, SD = .83) than when she did not (M = �.25,SD = .85), F (1, 33) = 2.73, p = .11, g2

p = .08. Thus, the same digitallyrendered smile enhanced social evaluations of empathy for embod-ied agents, but inhibited them for avatars.

Similarly, simple effects comparing avatars and embodiedagents separately within each smile condition revealed that whensmiling was disabled, avatars were viewed as significantly moreempathic relative to embodied agents, F (1, 33) = 7.36, p < .01,g2

p = .18. When smiling was enabled, however, embodied agentswere viewed as somewhat more empathic than avatars, F(1, 33) = 1.76, p = .19, g2

p = .05.

3.2. Other dependent measures

Results of a MANOVA using interaction partner type and smilecondition as the independent variables and the remaining eightinteraction outcome measures as the dependent variables revealedno significant main effects or interactions.

3.3. Counselor empathy

Next, a series of hierarchical linear regressions examined the ef-fect of counselor empathy and general positivity on the remaining

eight dependent variables by entering general positivity in Step 1followed by empathy in Step 2. As seen in Table 2, general positiv-ity towards the counselor was significantly related to the eightdependent variables. Except for liking, adding counselor empathyat Step 2 resulted in a significant increase in the R2 and reducedthe effect of general positivity non-significant. Thus, perceptionsof counselor empathy explained a significant proportion of the var-iance for seven of the eight dependent variables, even after control-ling for general positive feelings towards the counselor. Moreover,the only unique effect for these seven variables was for counselorempathy. Thus, greater perceptions of empathy led participantsto view their counselor as more supportive, they were more com-fortable and satisfied with their interaction, they enjoyed theirinteraction with her more, they experienced greater copresencewith her, they enjoyed interacting using immersive virtual envi-ronment technology more, and they felt more trusting and openin their interaction. Interestingly, although participants’ beliefsabout their partner and whether she smiled or not did not haveany direct effect on any of the interaction outcome variables, theamount of perceived empathy did vary as a function of bothwhether participants thought they were interacting with anotherperson or not and the nonverbal behavior of their partner. Empathythen predicted participants’ evaluation of their partner and theirfeelings about their interaction (see Fig. 4).

On the other hand, general positivity explained the majority ofthe variance in liking and adding counselor empathy in Step 2 re-sulted in only a marginal DR2 while general positivity remaineduniquely related to liking. The interpersonal traits comprising thegeneral positivity scale (i.e., friendly, attractive, intelligent, profes-sional, polite, likeable, and open-minded) might all be viewed astapping general liking for the counselor, thus their strong associa-tion with participants’ liking ratings is not surprising. Moreover,the pattern of results for general positivity with the remainingdependent variables mirrored those for liking.

4. Discussion

Overall, the results of this study indicate that choosing to rendera basic nonverbal behavior such as a smile can have a profound im-pact on how people experience their virtual interactions with oth-ers. Smiling increased perceptions of counselor empathy andempathy was positively associated with people’s evaluation of theirpeer counselor and their interaction. Specifically, higher empathywas associated with more positive counselor impressions, a greater

Table 2Beta predicting interaction outcome variables from counselor empathy and general positivity ratings.

Dependent variable b R2 DR2 Dependent variable b R2 DR2

Supportiveness Interaction satisfactionStep 1: .27*** Step 1: .23**

General positivity .52*** General positivity .47**

Step 2: .54*** .27*** Step 2: .38** .16**

General positivity �.06 General positivity .03Counselor empathy .78*** Counselor empathy .60**

Liking Interaction enjoymentStep 1: .53*** Step 1: .39***

General positivity .73*** General positivity .62***

Step 2: .57*** .04+ Step 2: .53*** .14**

General positivity .49** General positivity .21Counselor empathy .31+ Counselor empathy .56**

Comfort Trust & OpennessStep 1: .32*** Step 1: .10+

General positivity .56*** General positivity .31+

Step 2: .57*** .26*** Step 2: .30** .20**

General positivity �.01 General positivity �.20Counselor empathy .76*** Counselor empathy .68**

Copresence Satisfaction w/MediumStep 1: .29*** Step 1: .24**

General positivity .54*** General positivity .49**

Step 2: .62*** .33*** Step 2: .48*** .24***

General positivity �.11 General positivity �.07Counselor empathy .86*** Counselor empathy .75***

⁄p < .05.+ p < .10.

** p < .01.*** p < .001.

Smiling Counselor Empathy

Interaction Partner Type

Supportiveness

Comfort

Enjoyment

Satisfaction

Satisfaction w/ Medium

Copresence

Trust & Openness

Fig. 4. Model showing the relationship between smiling, empathy, and other variables.

2384 R.E. Guadagno et al. / Computers in Human Behavior 27 (2011) 2380–2385

sense of connection with and trust for the counselor, and morecomfort, enjoyment, and satisfaction with the interaction.

Importantly, however, the effect of rendering a smile on coun-selor empathy was impacted by whether people thought they wereinteracting with another person (i.e., an avatar) or a computer (i.e.,an embodied agent). Rendering a digital smile enhanced perceivedempathy for embodied agents, but doing so decreased perceivedempathy for avatars. In fact, an avatar who did not smile was per-ceived as significantly more empathic than one who did.

Previous research has shown that the effect of the communica-tive realism of virtual humans on a variety of social outcomes is com-plex and not necessarily intuitive (Bailenson et al., 2005; Fox &Bailenson, 2009; Garau, 2003; Garau et al., 2003; Guadagno et al.,2007; Yee & Bailenson, 2007; Yee, Bailenson, & Ducheneaut, 2009).While it might seem that increasing the extent to which an embod-ied agent or avatar possesses the verbal and non-verbal capacities ofhumans should increase its social influence, this does not seem to

always be the case. For example, Bailenson et al. (2005) demon-strated that although communicative realism affects perceptionsof copresence and various social responses to embodied agents,the effect of communicative realism varies as a function of the extentto which the embodied agent looks like a real human being.

Others have been found similar effects such that the effect of anavatar’s realism depends on the extent to which the avatar lookslike a real person (Garau, 2003; Garau et al., 2003). For example,Garau found that interactions with realistic-looking avatars wereviewed as more natural (i.e., more like a face-to-face interaction),they elicited greater copresence, and interaction partners wereevaluated more positively when avatars exhibited inferred eyegaze rather than random eye gaze. In contrast, interactions withless realistic-looking avatars were viewed more positively whenthe avatar exhibited random eye gaze (Ekman & Friesen, 1982).

Taken together, these studies on the effect of embodied agent oravatar appearance and behavior suggest that there needs to be a

R.E. Guadagno et al. / Computers in Human Behavior 27 (2011) 2380–2385 2385

match between the level of both appearance and behavior. When amismatch occurs, interactions are inhibited and social influence isreduced. According to Slater and Steed (2001), this is because peo-ple tend to hold higher expectations for more human-like repre-sentations than they do for less realistic-looking ones. Thus, themore realistic-looking an embodied agent or avatar appears, themore realistic its communicative behaviors such as facial expres-sions need to be in order for the digital representation to elicit so-cial responses. In contrast, less communicative realism is needed ininteractions with less realistic-looking embodied agents or avatars.

In this study, we attempted to increase the communicative real-ism of our peer counselor by adding a digitally-rendered smile. Be-cause the same digital peer counselor representation and samesmiling animation were used for all interactions, one might expectthat a smile would affect the positivity of social evaluations equallyregardless of whether participants thought they were interactingwith another person or a computer. However, this was not thecase. Instead, a digitally-rendered smile enhanced social evalua-tions of embodied agents while it decreased them for avatars.

Unfortunately, one limitation here is that the design did not al-low us to determine the reason for people’s negative reaction to asmiling avatar. Recall that the animations attempted to mimic themovement of the zygomaticus major and orbicularis oculi musclesas occurs with Duchenne smiling. Our animations, however, wereno doubt imperfect. Thus, something about the animation itselfmay have affected people differently depending on whether theythought their partner was a human or a computer. Alternatively,participants in the avatar condition may have been more sensitiveto the timing of the onset of the smile or its duration. Future re-search is needed to address these important issues.

References

Bailenson, J. N., Blascovich, J., Beall, A. C., & Loomis, J. M. (2003). Interpersonaldistance in immersive virtual environments. Personality and Social PsychologyBulletin, 29, 1–15.

Bailenson, J. N., Swinth, K. R., Hoyt, C. L., Persky, S., Dimov, A. E., & Blascovich, J.(2005). The independent and interactive effects of embodied agent appearanceand behavior on self-report, cognitive, and behavioral markers of copresence inimmersive virtual environments. Presence. Teleoperators and VirtualEnvironments, 14, 379–393.

Bailenson, J. N., Yee, N., Patel, K., & Beall, A. C. (2008). Detecting digital chameleons.Computers in Human Behavior, 24(2008), 66–87.

Bavelas, J. B., Black, A., Lemery, C. R., & Mullett, J. (1986). ‘‘I show you how I feel’’:Motor mimicry as a communicative act. Journal of Personality and SocialPsychology, 50, 322–329.

Baylor, A. L., & Soyoung, K. (2009). Designing nonverbal communication forpedagogical agents: When less is more. Computers in Human Behavior, 25,450–457.

Blascovich, J. (2002). A theoretical model of social influence for increasing the utilityof collaborative virtual environments. In Proceedings of the 4th InternationalConference on Collaborative Virtual Environments, Bonn, Germany.

Blascovich, J., & Bailenson, J. B. (2011). Infinite reality avatars, eternal life, new Worlds,and the Dawn of the virtual revolution. New York: HarperCollins.

Blascovich, J., Loomis, J., Beall, A., Swinth, K., Hoyt, C., & Bailenson, J. N. (2002).Immersive virtual environment technology as a methodological tool for socialpsychology. Psychological Inquiry, 13, 103–124.

Byrne, D. (1971). The Attraction Paradigm. New York: Academic Press.Ekman, P., & Friesen, W. V. (1982). Felt, false, and miserable smiles. Journal of

Nonverbal Behavior, 6, 238–252.Fox, J. A., & Bailenson, J. N. (2009). Virtual self-modeling: The effects of vicarious

reinforcement and identification on exercise behaviors. Media Psychology, 12(1),1–25.

Frank, M. G., Ekman, P., & Friesen, W. V. (1993). Behavioral markers andrecognizability of the smile of enjoyment. Journal of Personality and SocialPsychology, 64, 83–93.

Fridlund, A. J. (1991). Sociality of solitary smiling: Potentiation by an implicitaudience. Journal of Personality and Social Psychology, 60, 229–240.

Fridlund, A. J., Ekman, P., & Oster, H. (1987). Facial expressions of emotion: Reviewof the literature, 1970–1983. In A. W. Siegman & S. Feldsteing (Eds.), Nonverbalbehaviour and communication (pp. 143–224). Hillsdale, NJ: Lawrence ErlbaumAssociates, Inc.

Garau, M. (2003). The impact of avatar fidelity on social interaction in virtualenvironments. Unpublished Doctoral Dissertation, University College, London.

Garau, M., Slater, M., Vinayagamoorthy, V., Brogni, A., Steed, A., & Sasse, M. A.(2003). The impact of avatar realism and eye gaze control on perceived qualityof communication in a shared immersive virtual environment. In Proceedings ofthe SIG-CHI Conference on Human Factors in Computing Systems, April 5–10,2003, Fort Lauderdale, FL, USA, pp. 529–536.

Guadagno, R. E., Blascovich, J., Bailenson, J. N., & McCall, C. (2007). Virtual humansand persuasion: The effects of agency and behavioral realism. Media Psychology,10, 1–22.

Hall, J. A. (1978). Gender effects in decoding non-verbal cues. Psychological Bulletin,85, 845–857.

Lau, S. (1982). The effect of smiling on person perception. The Journal of SocialPsychology, 117, 63–67.

Loewald, T. A. (2009). RiddleMeThis. Tuscaloosa, AL: Loewald New Media. [Computersoftware available online at Riddlemethis.net.] .

Reis, H. T., Wilson, I. M., Monestere, C., Bernstein, S., Clark, K., & Seidl, E. (1990).What is smiling is beautiful and good. European Journal of Social Psychology, 20,259–267.

Slater, M., & Steed, A. (2001). Meeting people virtually: Experiments in virtualenvironments. In R. Schroeder (Ed.), The social life of avatars: Presence andinteraction in shared virtual environments. Berlin: Springer Verlag.

Srofe, L. A., & Waters, E. (1976). The ontogenesis of smiling and laughter: Aperspective on the organization of development in infancy. PsychologicalReview, 83, 173–189.

Stel, M., Blascovich, J., McCall, C., Mastop, J., Van Baaren, R. B., & Vonk, R. (2010).Mimicking disliked others: Effects of a priori liking on the mimicry-liking link.European Journal of Social Psychology, 40, 876–880.

Thompson, D. A., & Meltzer, L. (1964). Communication of emotional intent by facialexpression. Journal of Abnormal and Social Psychology, 68, 129–135.

Von der Pütten, A. M., Krämer, N. C., Gratch, J., & Kang, S. (2010). ‘‘It doesn’t matterwhat you are!’’ Explaining social effects of agents and avatars. Computers inHuman Behavior, 26, 1641–1650.

Yee, N., & Bailenson, J. (2007). The Proteus effect: The effect of transformed self-representation on behavior. Human Communication Research, 33, 271–290.

Yee, N., Bailenson, J. N., & Ducheneaut, N. (2009). The Proteus Effect: Implications oftransformed digital self-representation on online and offline behavior.Communication Research, 36(2), 285–312.