When Mii Is Me

Abstract

The present research examines the processes involved in the psychological and behavioral effects of avatar use. We offer and explore a new concept, avatar self-relevance, which potentially moderates avatar use effects and thus may help explain why such effects are augmented by using (compared with viewing) an avatar. Results from an experimental study suggest that avatar self-relevance after avatar use, as reflected by physiological responses to observing (without controlling) the avatar get beaten up, is higher for people who maintain a psychological connection to the avatar, while lower for people for whom the disconnection from the avatar is highly salient, with avatar gender consistency and an avatar-emotion connection contributing to the former and an avatar-body connection contributing to the latter. This research offers a middle ground between self-perception and priming-oriented explanations of the theoretical mechanisms involved in avatar use effects.

Keywords

avatars avatar effects Proteus effect psychophysiology virtual worlds heart rate skin conductance electromyography

Humans are adept at extending the self through the tools they use (Maravita & Iriki, 2004) and have been doing so for millennia (e.g., with sticks and rocks). Today, computing technologies have engendered many new forms of such self-extension. In particular, virtual worlds and video games often facilitate self-extension through one type of medium that is uniquely analogous to the self itself: avatars.

Avatars are mediated representations of an individual that appear and/or behave like the user (Bailenson & Blascovich, 2004). They are the medium of self in the medium, the virtual self. But they are also the tools that can extend the unmediated self into the virtual environment. In doing so, the avatar can be integrated into the avatar user’s conception of the self, just as physical tools can be integrated into the neurological gestalt of the self (Ratan, 2012). For example, regular tool use can lead to the perception that the tool is integrated into body schema (Maravita & Iriki, 2004). The present work builds off the notion that each element of the neurological framework of the self—comprised of body schema, emotions, and the collective memory of past emotions and anticipated future interactions (Damasio, 1999)—can be extended into an avatar (Ratan, 2012). Through such integration between the avatar and each element of the self, the avatar becomes part of the holistic self, potentially influencing the avatar user both inside and outside of the virtual environment.

This represents a new potential in human existence and is notable for many reasons, for example, the millions of people who use avatars every day and the research showing that avatar identity influences the user’s behavior, even after avatar use (cf. Fox, Bailenson, & Binney, 2009; Lee & Nass, 2012; Yee, Bailenson, & Ducheneaut, 2009). One specific type of post-use avatar effect, the Proteus effect, occurs when an avatar user conforms to the behavioral expectations associated with the avatar’s characteristics and was initially proposed to result from a change in the user’s self-perception (Yee et al., 2009). An alternate explanation of the Proteus effect is that the avatar’s identity characteristics simply prime related schema and thus influence behavior (Peña, 2011; Peña & Blackburn, 2013; Peña, Hancock, & Merola, 2009). In a study designed to tease apart these explanations, choices on a dating website (i.e., partner selection and height reporting) were influenced by avatar characteristics (i.e., attractiveness) for participants who controlled an avatar but not for participants who viewed without controlling the avatar (Yee & Bailenson, 2009). This result, which is consistent with other similar tests (cf. Yoon & Vargas, 2014), suggests that controlling an avatar augments avatar use effects, but it does not specify the causal mechanism, or more specifically, “how priming might interplay with self-perception” (Yee & Bailenson, 2009, p. 206).

The present research suggests a mechanism that is situated between the self-perception and priming-oriented explanations for the Proteus effect, potentially serving to reconcile the differences between these two approaches. Specifically, we note that previous research on avatar use (but not effects) has found that avatar use contributes to an association between the avatar’s characteristics and the user’s perception of self (Chandler, Konrath, & Schwarz, 2009; Klimmt, Hefner, Vorderer, Roth, & Blake, 2010). In other words, through avatar use, the schema of self-related concepts becomes associated with the schema of avatar-related concepts, and thus the priming of one schema potentially activates the other. This does not necessarily mean that the user’s perception of self is altered to include the avatar’s characteristics, though this may occur if the association between schemata is especially strong or long-lasting. Regardless of whether we consider such a mechanism to be a self-perception or a priming effect, this would explain why closer associations between the avatar’s characteristics and the self—as engendered through avatar use—lead to stronger avatar use effects.

In order to examine this concept of the association between the avatar’s characteristics and the self, the present research proposes the term avatar self-relevance, that is, the extent to which the avatar user perceives the avatar as relevant to the self. In the present study, we focus on avatar self-relevance after avatar use, which, we argue, influences physiological responses that can be measured while the user observes (without controlling) the avatar after use. We draw upon an experimental study to examine how such post-use avatar self-relevance is influenced by the preceding psychological experiences of avatar customization and avatar control (i.e., modification of the avatar itself and interaction with virtual objects through the avatar, respectively). Overall, our research suggests that the type of psychological connection to the avatar during use influences the extent of post-use avatar self-relevance, as reflected by physiological responses to observing the avatar, in theoretically consistent ways. This study is relevant to research on avatar use effects as well as to the application of avatars toward meaningful outcomes, such as education and training.

Avatar Self-Relevance, Defined

Avatar self-relevance can be defined as the extent to which an avatar user perceives an avatar as relevant to the user’s self, but such a definition requires refinement. As noted above, an avatar is a mediated representation of an individual that appears and/or behaves like the avatar user (Bailenson & Blascovich, 2004). This could be as complex as a three-dimensional anthropomorphic virtual character controlled by natural body movements or as simple as a customized pointer controlled by a mouse. Avatar self-relevance could be strong in either case, though it would be more likely in the former. Relevance is the noun form of relevant, defined in the Oxford English Dictionary as “closely connected or appropriate to the matter at hand” (“Relevance,” n.d.). Thus far, we can say that avatar self-relevance is the extent to which the user of a mediated representation that appears and/or behaves like the user is closely connected to the self.

This definition is simple up until the concept of the self. We adopt a neuroscientific perspective on what people generally experience as the self: body schema (the proto self), emotions experienced in response to interactions between body schema and objects in the environment (the core self), and the collective memory of previous emotions and future plans, which is akin to personal identity (the autobiographical self; Damasio, 1999). This three-level framework, which has been used in previous research about the psychological connections to avatars (Ratan, 2012), is an essential foundation of the present research, with each level related to a hypothesis (autobiographical self to Hypothesis 1, core self to Hypothesis 2, proto self to Hypothesis 3). However, we do not want the present work to preclude the potential for other perspectives on the self to fall within the purview of avatar self-relevance. For example, there is a wealth of psychological research on “possible selves,” including the notions of an ideal self, actual self, and ought self (Higgins, 1987; Markus & Nurius, 1986), and such research has been used fruitfully in the study of media effects (Bessière, Seay, & Kiesler, 2007; Messinger et al., 2008). Thus, in order to provide a fully expanded definition for our use, we turn to Oxford English Dictionary again, which defines the self as “a person’s essential being that distinguishes them from others” (“Self,” n.d.). Thus, avatar self-relevance is the extent to which the user of a mediated representation that appears and/or behaves like the user is closely connected to the user’s essential being that distinguishes them from others.

Physiological Indications of Avatar Self-Relevance

This research uses a psychophysiological approach to examine the primary outcome of interest, avatar self-relevance. This approach has become increasingly common in the study of communication and media in the past quarter century (Lang, Potter, & Bolls, 2009; Ravaja, 2004) because compared with more traditional methods (e.g., self-report survey), physiological signals are relatively impervious to demand characteristics and provide real-time information on unconscious processes. Furthermore, we focus mostly on phasic emotional responses, which have advantages over tonic physiological measures, such as their temporal precision, which facilitates the examination of responses to specific events during media use (Ravaja, Saari, Salminen, Laarni, & Kallinen, 2006).

Researchers in this area have used numerous psychophysiological methods to examine many types of media effects. Examples of psychophysiological research on traditional mass media effects include studies of differences when listening to audio messages of varied complexity (Potter & Choi, 2006), emotional valence and arousal differences related to modality type (Ravaja, Saari, Kallinen, & Laarni, 2006), differences in cardiac orienting related to the onset of emotional images (Wise & Reeves, 2007), and differences in arousal and cognitive effort related to the relative order of advertisements and programs (Wang & Lang, 2012). There are also several studies on psychophysiological effects of using interactive media, such as arousal differences related to violent content (Bartlett, Harris, & Bruey, 2008; Ivory & Kalyanaraman, 2007; Weber, Ritterfeld, & Mathiak, 2006), and immersion (Drachen, Nacke, Yannakakis, & Pedersen, 2010; Persky & Blascovich, 2007; Ravaja, Saari, Kallinen, & Laarni, 2006), as well as differences in emotional valence (Hazlett, 2006; Ravaja, Saari, Salminen, et al., 2006; Ravaja, Turpeinen, Saari, Puttonen, & Keltikangas-Järvinen, 2008) and brain activity during game play (Klasen, Weber, Kircher, Mathiak, & Mathiak, 2012; Mathiak & Weber, 2006; Weber et al., 2006).

In the present research, we expect that avatar self-relevance is associated with physiological changes during avatar use, which is consistent with the argument that “[identification with avatars] would be expected to result in emotional responses similar to those prompted by unmediated events” (Ravaja, Saari, Salminen et al., 2006, p. 348). Specifically, we expect that stronger avatar self-relevance is associated with more attention to the avatar and stronger emotionally valenced responses that are consistent with the valence of the avatar’s situation (i.e., positive responses to the avatar in positive situations, negative responses to the avatar in negative situations). People pay more attention to—or orient toward—stimuli that are significant and/or novel (Graham, 1979; Sokolov, 1963). When an avatar is relevant to the self, it is significant to the user. Thus, the greater the avatar self-relevance, the more attention the user will pay to the avatar (especially in significant or novel circumstances). Regarding emotional valence, people experience positive emotions in response to pleasant stimuli and negative emotions in response unpleasant stimuli, essentially by definition (i.e., pleasant experiences yield pleasure, unpleasant experiences yield displeasure). When an avatar is relevant to the self, observing the avatar in positive or negative situations would be perceived as more pleasant or unpleasant, respectively, than if the avatar is not relevant to the self. Thus, the greater the avatar self-relevance, the more the avatar user will experience positive or negative emotions in response to observing the avatar in positive or negative situations, respectively.

Avatar Self-Relevance After Avatar Use

Because previous research on avatar use effects has found that avatar characteristics can influence users beyond the period of use, we are particularly interested in avatar self-relevance after avatar use. Previous research has found that avatar characteristics are indeed associated with perceptions of the self after avatar use. For example, one study found that identification with a video game character during play can alter self-concept and “leave cognitive traces that can be found after game play” (Klimmt et al., 2010, p. 333), as measured by a task in which associations between me-related and avatar-related words were measured implicitly. Another study examined the relevance of the avatar to the self outside of the mediated context by administering a survey-based test of whether avatar users’ (who regularly played a game with the avatar) own body image was related to their avatars’ body sizes (Chandler et al., 2009). The participants for whom the avatar was “chronically accessible” (i.e., strong avatar self-relevance) reported body image perceptions that were positively related to avatar body size. Furthermore, they found that the participants for whom the avatar was not chronically accessible also showed a positive relationship between body image and avatar body size but only when the avatar had been brought to mind by previous questions about the avatar in the survey.

The present research examines physiological signals in response to viewing the avatar after use. These responses should reflect differences in avatar self-relevance that are independent of the avatar’s “chronic accessibility” and instead relate to differences in the psychological experience of previously using the avatar. In other words, the user’s physiological responses to such a presentation of the avatar should reflect attention and emotional connection to the avatar, thereby signaling the extent of avatar self-relevance beyond avatar use.

Antecedents of Avatar Self-Relevance

Having described our psychophysiological approach to examining avatar self-relevance after avatar use, we now turn to the characteristics of avatar use that we believe should influence the extent of such post-use avatar self-relevance. For the purpose of this research, we segment the process of avatar use into two distinct phases: customization and control. Avatar customization involves selecting an avatar from a pre-existing set of options or customizing multiple characteristics of an avatar’s appearance. During the avatar control phase, the avatar user manipulates the avatar’s behavior through some user interface (e.g., tangible or gestural). These two phases likely influence the relevance of the avatar in distinct ways.

Customizing an avatar should lead to increases in avatar self-relevance. Previous research supports this claim, finding that when using an avatar they have customized (compared with one that was simply assigned), users exhibit more arousal (Bailey, Wise, & Bolls, 2009), even when such customization involves simply selecting an avatar from a limited pre-existing set (Lim & Reeves, 2009). This suggests that through avatar customization, the user develops a feeling of ownership or affinity for the avatar that enhances the experience of using it. Previous research also suggests that avatar customization is an important element of avatar use, especially in virtual contexts that emphasize social interaction (Ducheneaut, Wen, Yee, & Wadley, 2009). Furthermore, people who choose to customize an avatar, compared with those who do not, report feeling that their identity is more reflected in the avatar (Ratan & Hasler, 2010), which is consistent with the notion that avatar customizers tend to imbue their avatar with identity elements of the self (Vasalou & Joinson, 2009). In other words, from a neuroscientific perspective, by customizing identity characteristics into the avatar, the avatar becomes relevant to personal identity (i.e., the autobiographical self; Damasio, 1999). Even after avatar use, the customized avatar continues to be relevant to the self, given that personal identity remains constant (which is likely if the period after use is short). Thus, we expect the following:

Hypothesis 1a: Avatar customization increases avatar self-relevance shortly after avatar use.

Avatar gender is one of the most important elements of avatar customization. While people sometimes “gender swap,” or use avatars of a different gender, avatar users—especially women—tend to customize avatars to reflect their own gender (Ducheneaut et al., 2009; Huh & Williams, 2010; Yee, Ducheneaut, Yao, & Nelson, 2011). This suggests that in general, people who are unable to imbue gender identity into an avatar would be hindered in their experience of avatar self-relevance compared with people who customize their gender into an avatar, all other factors of customization being equal. In other words, avatar self-relevance should be higher for people who are asked to use a same-gender avatar compared with people who are asked to use an other-gender avatar. Thus, we expect the following:

Hypothesis 1b: Using a same-gender avatar leads to more avatar self-relevance shortly after avatar use compared with using an other-gendered avatar.

Beyond customizing, having control over an avatar provides the user with agency in the virtual environment, facilitating interactions with objects in the virtual environment that allow the user to accomplish some sort of goal. Given that the user is motivated to accomplish this goal, these interactions with the virtual environment likely cause emotional responses in the user. Thus, from a neuroscientific perspective, by using the avatar to interact with a virtual environment, the avatar may become relevant to the core self, that is, the perception of self that arises through emotional responses to interactions between the body and objects in the environment (Damasio, 1999). This is consistent with research showing that after controlling an avatar, compared with only observing an avatar, people are more likely to be influenced by the avatar’s identity characteristics (Yee & Bailenson, 2009; Yoon & Vargas, 2014). Thus, an avatar-emotion connection likely increases avatar self-relevance during avatar use and persists beyond use for as long as the user remembers the experience of interacting with virtual objects through the avatar. If the period after use is short, then it is likely that the user would remember these interactions, and thus we expect the following:

Hypothesis 2: The more avatar-emotion connection during avatar use, the more avatar self-relevance shortly after avatar use.

Avatar control requires the user to engage with the avatar through some medium to which the user’s body is connected (e.g., tangibly or gesturally). As the user becomes increasingly adept at controlling the avatar, this interaction between the user’s body and the avatar may become increasingly natural for the user to the extent that the user perceives the avatar as an extension of the user’s own body (Ratan, 2012). For example, people who play games that require continuous, first-person control of the avatar (i.e., first-person shooters), compared with games in which the user is not required to continuously control all of the avatar’s movements (i.e., Massively Multiplayer-Online Role Playing Games), report feeling a stronger connection between their bodies and their avatars (Ratan, 2011).

This notion of an avatar-body connection is similar to the type of physical connections that people can experience with artificial objects (e.g., Armel & Ramachandran, 2003; Berti & Frassinetti, 2000). For example, people can be induced, through tactile stimulation, to perceive that their actual limb has been replaced by a rubber limb (Botvinick & Cohen, 1998) or even a fake limb in a virtual environment (Kilteni, Normand, Sanchez-Vives, & Slater, 2012; Slater, Perez-Marcos, Ehrsson, & Sanchez-Vives, 2009). Similarly, regular tool use can lead to the perception that the tool is integrated into body schema (Maravita & Iriki, 2004). This suggests that by using an avatar, some users may experience an avatar as an extension of the user’s body. In other words, from a neuroscientific perspective, during avatar use, the avatar becomes relevant to the proto self, that is, the neural mapping of what constitutes the body (Damasio, 1999).

However, after avatar use, when the user cannot control the avatar, such an avatar-body connection should dissipate immediately because the incongruity between the avatar’s behavior and the user’s intentions is an obvious cue that the avatar is no longer integrated with the user’s body schema (Ratan, 2012). This leads to a seemingly counterintuitive expectation based on the extent of the perceived contrast between the period during avatar use and the period after disconnection from the avatar. Namely, for someone who experiences a strong avatar-body connection during avatar use, the disconnection from the avatar after use represents a significant contrast because the user’s inability to control the avatar presents a large change from the previous state. Conversely, for someone who experiences a weak avatar-body connection during avatar use, the disconnection from the avatar after use represents a minor contrast because the change from the previous state is not large. Thus, the counterintuitive expectation is that people in the former category (stronger avatar-body connection during use) should feel more disconnected from the avatar after use than people in the latter category (weaker avatar-body connection), which would lead to less avatar self-relevance for the former group:

Hypothesis 3: The more avatar-body connection during avatar use, the less avatar self-relevance shortly after avatar use.

Method

Population and Design

The participant pool for this study consisted of 76 right-handed female undergraduate students from a large Western university between the ages of 18 and 28 (M = 19.83, SD = 2.07). The sample was restricted to right-handed participants because the physiological recording equipment could only be attached to the left arm due to constraints in the lab. The female-only sample was useful because of the interest in the effects of avatar gender consistency (Hypothesis 1b) and also because of an interest in the effects of avatar-induced gender stereotypes, a topic which is not the focus of the present article.

Participants were randomly assigned to one of four groups using a 2 (customization) × 2 (gender) between-subjects design. For the avatar customization variable, half of the participants used an avatar that they had customized to look like themselves, while the other half were assigned to use a generic avatar. For the avatar gender variable, half of the participants used a same-gender avatar, and the other half used an other-gendered avatar.

Media Materials

The avatar use context for this study was the game Wii Sports Resort, a suite of sports games for the Nintendo Wii, and participants played it on a Sony XBR 52″ television. The specific game used within Wii Sports Resort was Swordplay, a cartoon sword dueling game in which the player swings the controller (“Wii Remote”) to control a blunt sword and to knock an opponent off a platform. Although the premise of the game may seem violent, the avatars are depicted wearing safety gear, and the game is rated by the Entertainment Software Rating Board as appropriate for everyone. This game was chosen because it allows the player to contol the avatar from a hybrid first-person/third-person point-of-view, with the camera in the game placed slightly behind the avatar’s semi-transparent head (see Figure 1). From this perspective, the player is able to control the avatar while seeing cues of what the avatar looks like. The player is also reminded of the avatar’s identity when watching replays at the end of the rounds (from a third-person perspective) and when viewing avatar icons during menu navigation. In this way, the game potentially facilitates avatar self-relevance through both avatar appearance and control.

Figure 1.

Hybrid first-person/third-person point-of-view, Swordplay game, © Nintendo, Inc.

Procedure

Participants entered the lab and were instructed to read the institutional review board (IRB)-approved study information sheet. After indicating to the researcher that they had finished reading, they were given an iPad to complete a short demographic survey. The researcher then attached physiological sensors on the participants’ arms, finger, foreheads, and ankles (details below).

Participants were then instructed to spend five minutes creating an avatar, or “Mii,” as it is called on the Nintendo Wii. If the participant was in the same-gender condition, she was told to make this Mii as similar to herself as possible. If the participant was in the other-gender condition, she was told to build the Mii as a male version of herself. For additional clarification, she was told that this Mii should appear to be how she would if she were male. After creating the Mii, participants were instructed to relax for one minute of baseline physiological recording. Then the researcher set up the Swordplay game and told the participants what avatar they would be using. If the participant was in the customized avatar condition, the researcher simply selected the aforementioned avatar for her. If the participant was in the generic (non-customized) avatar condition, as she was selecting the avatar for play, the researcher said, “I know you made another Mii, but you’re actually going to be playing with this one.” The gender of this avatar was assigned to match the player’s gender condition.

The researcher then trained the participants on how to play and told the participants to watch the replays after each round and to keep playing matches until instructed to stop. He then left the play area and went to a sectioned off area of the room. After participants played for approximately 10 minutes, they were allowed to finish the current match and then told to stop playing. The researcher then turned off the television and gave the participants the iPad for a 15-minute survey.

After this survey, participants were instructed to use the Wii controller to click through the menu options and prepare for a final session of play to be followed by the final questionnaire. Participants chose the same avatar they had used in the first play session and then selected the “Showdown” sub-game of Swordplay. This sub-game is made for single-players to battle a series of computer-controlled opponents. However, before the game began, the researcher instructed the participants to put the controller down and just watch the game. The participants then watched as their avatars were defeated after being hit three times by a computer-controlled opponent. We chose this scenario so that the participant would no longer be engaged in avatar use but would still be forced to observe the avatar receive negative treatment.

The researcher marked the moment of each of the opponent’s three hits so that event-related physiological responses could be analyzed. The amount of time between hits varied randomly, thus creating variance in the total duration of this segment of the experiment (in seconds, M = 41.18, SD = 14.31, minimum = 14.96, maximum = 77.31). Furthermore, the pre-hit duration (amount of time before a given hit since the previous hit) was not normally distributed (M = 11.00, SD = 8.47, minimum = 2.31, maximum = 53.69, skewness = 1.93, SE = 0.161, kurtosis = 4.90, SE = 0.32). This is potentially problematic because during abnormally long pre-hit durations, the participants’ attention to the game and their avatars may have waned, making their responses to the hits unrepresentative of the whole sample. Thus, for all analyses, we restricted the sample to include only those with a pre-hit duration of less than 30 seconds, which was the largest pre-hit duration for which the sample’s kurtosis (0.60) divided by the standard error of kurtosis (0.33) was less than 2.0, indicating a low likelihood that the sample had excess kurtosis (Cramer, 1997). Thus, out of the 228 hits received by the 76 participants (three hits each), we analyzed responses to a maximum of 219 hits.

Survey Measures

The avatar-body connection and avatar-emotion connection measures were based on existing scales that measure the same constructs but with different names, that is, proto self-presence and core self-presence (Ratan, 2012). Previous research has utilized factor analyses to support these scales’ reliability, construct validity (Ratan, 2011; Ratan & Hasler, 2010), and discriminant validity from the superordinate construct of presence (Ratan & Sah, 2014). Cronbach’s alpha for the avatar-body connection measure was .89 and for the avatar-emotion connection measure was .77. The items used for each are listed in the appendix.

Because individual differences in experience with computing technology and video games potentially influence the psychological experience of using an avatar, we measured the number of hours participants spend using a computer each week as well as the frequency with which they play video games and use virtual worlds (five-point Likert-type scales ranging from never to every day). We include these measures in our analysis as control variables.

As a manipulation check of whether participants in the avatar customization condition actually imbued characteristics of their identity into the avatar, we used an avatar-identity connection measure based on the same questionnaire from which the other avatar connection measures were drawn (extended self-presence in Ratan, 2012). This measure has been used as an independent variable in previous research (e.g., Ratan, 2011; Ratan & Hasler, 2010), but it is more appropriate as a manipulation check in the present study given that customization is manipulated experimentally. Cronbach’s alpha for the measure was .89 and the items can be found in the appendix.

Participants also responded to a number of other survey questions that are not relevant to the current inquiry, such as personality inventory measures and a series of math questions.

Physiological Materials and Raw Data

All physiological data were acquired using the 4-channel MP36 system and Biopac Student Lab software from Biopac Systems, Inc. (www.biopac.com) at a 1000-Hz sampling rate, to which the participants were attached early in the experiment procedure (as described under the Procedure section). To record electrocardiogram (ECG) activity, the signal from which heart rate (HR) was calculated, disposable adhesive pre-gelled (electrolyte hydrogel) silver-silver chloride electrodes (one cm in diameter) were placed on the center of the left forearm and the medial side of the left ankle just behind the fibula, and with a ground electrode placed on the left wrist. The same electrode type (but with isotonic gel) was used to record skin conductance responses (SCRs), and these electrodes were placed on the distal phalanges of the index and middle fingers of the non-dominant (left) hand. To record corrugator electromyogram (EMG) activity, reusable 8-mm silver-silver chloride electrodes, filled with hypertonic gel to promote conductivity, were placed on the left side of the forehead, with a ground electrode on the temple. The skin locations for the EMG electrodes were prepped with an abrasive skin lotion (Lemon Prep) and an alcohol swab, while the skin locations for the ECG electrodes were prepped with an alcohol swab alone. An athletic headband was used to stabilize the EMG electrodes. Images of the electrode placements on the arm and face can be found in Figure 2.

Figure 2.

Electrode placement on the arm and face.

Data transformations

After recording, the raw physiological data were transformed using the AcqKnowledge 4.1 software, also from Biopac Systems, Inc. The ECG output was converted to HR, expressed as beats per minute (BPM), using the AcqKnowledge Rate Detector algorithm, which creates a cardiotachometer tracing by calculating the interbeat interval and then extrapolating the BPM. For example, if there is a 600-millisecond interval between heart beats 1 and 2, then the HR is calculated as 100 BPM and then updated at each subsequent beat based on the interval between beats. Following Lang (1990), we visually examined all values greater than 100 BPM or less than 50 BPM, which is an average range for resting HR, and identified ectopic beats and artifacts in the ECG data (likely due to participant movement). In these cases, we visually identified the point on the ECG waveform at which we would have expected the heart beat to occur. Using the functionality provided by the software, we manually modified the waveform so that it would peak at this point, thereby allowing the software’s Rate Detector algorithm to include this point in the HR calculation accordingly.

The EMG data, measured in microvolts, were transformed using an integrate function to average over 100 samples using the root mean square method with no baseline removal. There were no transformations conducted on the SCR data, which was measured as increases in skin conductance in microsiemens. An example of the raw and transformed data can be found in Figure 3.

Physiological Measures and Metrics

As described previously, this research focuses on the psychophysiological constructs of attention and emotionally valenced responses as reflections of avatar self-relevance. The descriptive statistics for all metrics described below can be found in Table 1.

Table 1.

Descriptive Statistics.

	n	M	SD	Minimum	Maximum
Avatar-body connection	219	2.99	0.90	1.00	4.60
Avatar-emotion connection	219	3.07	1.07	1.00	6.00
HR deceleration (Δ BPM)	133	4.92	5.35	−2.23	36.72
SCR amplitude (µS)	135	0.83	1.21	0.00	6.80
Corrugator EMG (nV)	219	12.08	12.49	−2.94	59.72
SCR count (per minute)	76	6.73	3.85	0.00	16.22
Frequency video games	219	2.20	0.81	1.00	5.00
Computer hours per week	219	23.25	13.52	3.00	60.00

Note. Statistics are presented here for raw values, but values are centered for independent variables and covariates in subsequent analyses. HR = heart rate; BPM = beats per minute; SCR = skin conductance response; EMG = electromyogram.

Figure 3.

Example of raw (left) and transformed (right) physiological output.

Our measure of attention was an orienting response (OR), which is a physiological process that serves to enhance perception of stimuli—thereby reflecting increased attention to the stimuli (Sokolov, 1963)—and occurs in response to stimuli that are novel and/or significant to the individual (Graham, 1979; Sokolov, 1963), that is, relevant to the self. ORs are associated with HR deceleration and a phasic increase in skin conductance—or an SCR—and thus reflect both sympathetic and parasympathetic nervous system activity, with the parasympathetic nervous system dominantly affecting HR (deceleration) but inconsequential to skin conductance (Graham, 1979; Graham & Clifton, 1966; Öhman, Hamm, & Hugdahl, 2000).

HR metric

We operationalized HR deceleration by taking the difference between a one-second pre-hit mean (baseline) and a three-second post-stimulus (hit from the opponent) minimum. Although a full HR response can last up to six seconds (see Lang, 1994), this type of three-second initial HR change has been commonly used as an index of ORs (Bohlin & Kjellberg, 1979; Bradley & Lang, 2007; Graham, 1979), and a full six seconds of data were not available on many of the hits (described further below).

SCR metrics

As our primary SCR metric, we examined the amplitude of SCRs associated with each hit. We operationalized this as the onset-to-peak skin conductance change associated with each hit the participant received from the opponent, using 0.1 microsiemens as the minimum threshold. We treated the cases below this threshold as non-responses, for which we entered a value of 0. We should note that we associated an SCR with a hit only if the onset for the SCR occurred within four seconds following the hit. Although it normally takes at least one second for SCR onset to begin after a stimulus, we observed that many participants began responding within the zero-to-one-second window. Upon examination of the Sword play game, we noticed that the opponent would raise the sword about one second before hitting the player, and so we reasoned that SCR occurred in response to this movement, with onset within one second of the hit.

As a secondary operationalization of skin conductance activity, we measured the total count of SCRs that each participant exhibited during the play session, which is a metric that has been used in previous research. For example, threat of shock and stimulant medications have been found to increase the number of non-specific SCRs (Dawson & Nuechterlein, 1984). In the present case, the count includes SCRs not associated with any of the three hits to the avatar but may have occurred in response to the threat of being hit by opponents. During the play session, the player faced off against a primary computer-controlled opponent but was surrounded by two or three other opponents during the session. The game is designed this way so that the player can knock down multiple opponents in quick succession, but only the primary opponent (with a name displayed above its head) is able to hit the player’s avatar. During the play session, the non-primary opponents raise and lower their swords but do not strike the opponent, and even the primary opponent occasionally raises the sword and then lowers it without striking the player’s avatar. This threatening environment would have been associated with a heightened frequency of SCRs. Thus, our secondary operationalization of skin conductance activity was the total number of SCRs exhibited throughout the play session (controlling for session duration).

Corrugator EMG metric

Regarding emotional valence, given that this research involves the avatar receiving only negative treatment, we focus only on a measure of negative emotional valence, corrugator EMG activity (frowning), which is commonly observed following stimuli judged to be unpleasant (Larsen, Norris, & Cacioppo, 2003). Thus, we operationalized corrugator EMG activity as the difference between a one-second pre-hit mean (baseline) and a three-second post-hit maximum, except when the post-hit duration was less than three seconds, in which case we took the maximum value leading up to subsequent hit. We should note that there was a single outlier with a change value more than 10 standard deviations above the mean, so we replaced this value with the second highest value across the sample, which was approximately 2.5 standard deviations above the mean.

Sample restrictions

As we noted previously, there was variance in the interval between hits, with some hits occurring less than three seconds after the previous hit. We considered these intervals when analyzing the hit-associated metrics (HR deceleration, SCR amplitude, and EMG activity). Specifically, for our analyses of the HR and SCR amplitude, we included only the cases in which the pre-hit duration was greater than six seconds, which is an approximate amount of time required for recovery from a cardiac OR (Lang, 1994) and a SCR (Dawson, Schell, & Filion, 2007). Furthermore, given that the HR deceleration was measured up to three seconds after each hit, for our analyses of the HR metric, we restricted the sample to only those with a post-hit duration of greater than three seconds. We used the same three-second post-hit restriction for our analyses of SCR amplitude because by three seconds after stimulus, it was possible to identify the SCR peak and thus calculate the metric for the given hit, even if the SCR did not fully recover before the subsequent hit. Unlike the other measures, corrugator EMG activity occurs within a very short window of a stimulus (Dimberg & Petterson, 2000), and so given that the shortest interval between hits was 2.31 seconds, we included all cases in the analysis.

Metric interpretation

Across the hit-associated metrics, we interpreted the size of the response to the hit as reflective of avatar self-relevance. In other words, we inferred that participants who experienced greater avatar self-relevance would exhibit more HR deceleration, larger SCRs, and greater corrugator EMG activity in response to their avatar being hit by the opponent. In other words, we pre-supposed that avatar self-relevance would be reflected by more attention (HR deceleration and larger SCRs) as well as more negative emotional responses (greater corrugator EMG activity) to the negative treatment of the avatar. Furthermore, for the SCR count metric, we interpreted the number of responses (controlling for session duration) as reflective of attention, with more responses reflecting more attention and thus more avatar self-relevance.

Results

Manipulation Check

The manipulation check supported the expectation that participants in the customized avatar condition perceived their avatar as reflecting elements of their identities to a greater extent than participants who used a generic avatar, with those in the former condition reporting a greater avatar-identity connection (M = 3.24, SD = .73) than those in the latter (M = 2.43, SD = .93). This difference was significant, t(74) = 4.22, p < .001.

Confirming Responses

As a foundation for examining the hypotheses about the antecedents of avatar self-relevance, as reflected by physiological responses, we confirmed that participants across all conditions and psychological experiences of the avatar did respond physiologically to the hits. First, we constructed a cardiac response curve for all participants across all three hits and then separately between hits, measured at 0.5-second intervals. The graph for all hits together (Figure 4, left) reflects an average HR deceleration of about 1.5 BPM to a minimum at about 2.5 second following the hit, thereby suggesting ORs to the hits. This is supported by a one-way (HR at time point) univariate analysis of variance (ANOVA), which shows a significant linear trend component, F(1, 132) = 6.50, p = .01, η² = .05. The graph that separates responses by hit number (Figure 4, right) suggests that the trend is similar across all hits. A 3 (hit #) × 6 (time) ANOVA found no significant interaction between hit # and HR, indicating that the monophasic response was consistent across hits.

Figure 4.

Cardiac response curves for three seconds following hits to avatar.

In order to confirm that SCRs were associated with the hits, we examined the descriptive statistics for SCRs across all hits (M = 0.83µS, SD = 1.21). We then conducted a one-way (hit #) ANOVA on SCR amplitude and found a significant difference, with larger responses following Hit 1 (1.20 µS) than Hit 2 (0.52 µS), F(2, 132) = 4.07, p < .05, η² = .06. Given this difference, we examined responses to each hit together as well as separately in subsequent analyses.

In order to confirm corrugator EMG activity associated with the hits, we compared the pre-hit and post-hit values and found that the latter was significantly larger than the former, F(1, 218) = 109.36, p < .01, η² = .33. We then conducted a one-way (hit #) ANOVA on the EMG activity metric and found a significant difference, with larger responses for Hit 1 (15.05 nV) than Hit 2 (9.30 nV), F(2, 216) = 4.00, p < .05, η² = .04). Given this difference, we examined responses to each hit together as well as separately in subsequent analyses.

Analyses

We conducted a univariate analysis of covariance (ANCOVA) to examine the effects on the SCR count metric (Table 2) and a series of mixed model analyses to examine the effects on each of the hit-associated metrics (HR deceleration, SCR amplitude, and corrugator EMG activity; Tables 3-5). In addition to the experiment conditions, the measured independent variables (avatar connection measures) and control variables (experience with computing technology and video games) were centered and included in the analysis.The full mixed models contained the hit number (i.e., Hit 1, Hit 2, and Hit 3) as a repeated measure (also centered). These models utilized restricted maximum likelihood estimation of covariance parameters and a first-order autoregressive (AR1) covariance structure. We did not include the interaction terms between the independent variables and the hit number in the full models because this would have led to an excessive number of interactions that could potentially provide misleading results, given the sample sizes. Thus, in order to account for differences in responses between the three hits, we examined additional models with the variables of interest, holding each hit constant.

Table 2.

ANCOVA on SCR Count.

	df	F	Significance	η²
Avatar gender	1	0.84
Avatar customization	1	2.91	^†	0.04
Avatar customization × Avatar gender	1	0.12
Avatar-body connection	1	0.23
Avatar-emotion connection	1	5.12	*	0.07
Frequency video games	1	0.53
Computer hours per week	1	4.98	*	0.07
Session duration	1	14.57	***	0.18
Error	1
Total	76

Note. SCR = skin conductance response.

†

p < .10. *p < .05. **p < .01. ***p < .001.

Table 3.

Mixed-Level Model on Heart Rate Deceleration for All Hits and for Each Hit Individually.

	All hits			Hit 1			Hit 2			Hit 3
	F	T	Significance	F	t	Significance	F	t	Significance	F	t	Significance
Avatar gender	0.00	-0.51		0.01	-0.19		0.18	0.41		0.03	-0.51
Avatar customization	0.13	-0.75		0.17	-0.52		0.62	0.16		0.03	-0.47
Avatar customization × Avatar gender	0.53	0.73		0.13	0.36		1.01	-1.01		0.74	0.86
Avatar-body connection	0.00	-0.04		0.68	0.83		2.72	-1.65		0.03	-0.18
Avatar-emotion connection	0.13	-0.36		1.08	-1.04		4.42	2.10	*	0.37	-0.61
Frequency video games	1.67	1.29		0.27	0.52		3.79	1.95	^†	0.39	0.62
Computer hours per week	1.40	-1.18		0.13	0.37		0.44	-0.67		2.28	-1.51
Hit # (single fixed factor)	2.48
Hit 1 (compared with Hit 3)		-1.89	^†
Hit 2 (compared with Hit 3)		-1.98	^†

†

p < .10. *p < .05. **p < .01. ***p < .001.

The results of these analyses (Tables 2 -5) provide insights into the relationship between avatar self-relevance and physiological responses to viewing the avatar receive negative treatment. Hypothesis 1a was unsupported, although there were marginally significant differences in the number of SCRs (Table 2) and the amount of corrugator EMG activity in response to Hit 2 (Table 5) between those who customized the avatar and those who did not. Hypothesis 1b was supported by SCR amplitudes in response to Hit 3 (Table 4), with participants who used a female avatar exhibiting larger SCRs (M = 0.96 µS, SE = 0.20) compared with participants who used a male avatar (M = 0.29 µS, SE = 0.23). This suggests that using an avatar whose gender is consistent with the user’s gender contributes to avatar self-relevance.

Table 4.

Mixed-Level Model on SCR Amplitude for All Hits and for Each Hit Individually.

	All hits			Hit 1			Hit 2			Hit 3
	F	t	Significance	F	t	Significance	F	t	Significance	F	t	Significance
Avatar gender	1.25	1.31		0.01	0.34		1.09	0.22		4.54	2.45	*
Avatar customization	1.79	-0.37		1.00	-0.25		0.00	-0.56		0.07	0.85
Avatar custom × Avatar gender	0.59	-0.77		0.34	-0.58		0.59	0.77		2.31	-1.52
Avatar-body connection	0.00	0.02		0.08	0.27		0.78	0.88		0.00	0.04
Avatar-emotion connection	4.51	2.12	*	0.84	0.92		7.05	2.66	*	1.10	1.05
Frequency video games	2.14	-1.46		0.24	-0.49		3.29	-1.81	^†	2.41	-1.55
Computer hours per week	1.74	-1.32		0.09	-0.30		5.90	-2.43	*	1.34	-1.16
Hit # (single fixed factor)	3.53
Hit 1 (compared with Hit 3)		2.18	*
Hit 2 (compared with Hit 3)		-0.46

Note. SCR = skin conductance response.

†

p < .10. *p < .05. **p < .01. ***p < .001.

Table 5.

Mixed-Level Model on Corrugator EMG Activity for All Hits and for Each Hit Individually.

	All hits			Hit 1			Hit 2			Hit 3
	F	t	Significance	F	t	Significance	F	t		F	t	Significance
Avatar gender	1.12	0.17		1.24	1.75	^†	3.73	0.38		0.04	-0.52
Avatar customization	1.65	1.81		0.02	1.11		1.83	1.93	^†(p = .06)	0.46	0.07
Avatar customization × Avatar gender	1.74	-1.32		2.01	-1.42		1.98	-1.41		0.31	0.56
Avatar-body connection	4.88	-2.21	*	1.24	-1.12		4.39	-2.09	*	1.80	-1.34
Avatar-emotion connection	0.00	0.05		0.07	0.26		0.23	-0.48		1.25	1.12
Frequency video games	0.07	-0.26		0.16	0.40		0.02	-0.12		0.47	-0.69
Computer hours per week	0.01	-0.12		0.00	-0.06		0.00	0.04		0.02	-0.14
Hit # (single fixed factor)	5.55		**
Hit 1 (compared with Hit 3)		1.51
Hit 2 (compared with Hit 3)		-1.84	^†

Note. EMG = electromyogram.

†

p < .10. *p < .05. **p < .01. ***p < .001.

Providing support for Hypothesis 2, the more participants reported an avatar-emotion connection, the greater number of SCRs they exhibited throughout the play session (Table 2), the larger their HR decrease in response to Hit 2 (Table 3), and the larger their SCR amplitudes across all hits (Table 4), though this third finding appears to be driven by responses to Hit 2. This suggests that an avatar-emotion connection enhances avatar self-relevance.

Providing support for Hypothesis 3, participants who reported a stronger avatar-body connection exhibited weaker corrugator EMG activity (frowning) in response to watching their avatar get hit (Table 5), though this finding also appears to be driven by the response to Hit 2. This suggests that these participants experience less negative emotions when watching their avatar receive negative treatment and thus that an avatar-body connection during use reduces post-use avatar self-relevance.

Discussion

The present research puts forth the concept of avatar self-relevance and examines how this psychological experience (of perceiving the avatar as relevant to the self) is reflected by physiological responses to post-use observations of the avatar receiving negative treatment. Results from an experimental study suggest that avatar self-relevance after avatar use was higher for people who used a same-gendered (compared with other-gendered) avatar, experienced more avatar-emotion connection, or experienced less avatar-body connection during avatar use, as reflected by some (but not all) of the physiological metrics employed in the study. These findings suggest that avatar self-relevance is a useful construct to consider when examining avatar use effects, especially because of the potential for counterintuitive relationships (i.e., when avatar self-relevance is diminished by a salient disconnection from the avatar). Avatar customization in general was not found to contribute significantly to avatar self-relevance (though the relationship was marginally significant in some cases). As with all findings of no difference, this does not preclude the possibility that avatar customization contributes to avatar self-relevance in other contexts. Overall, this research supports the notion that avatar self-relevance moderates avatar use effects, offering a theoretical mechanism of avatar use effects that is situated in between self-perception and priming-oriented explanations.

The positive relationship found between avatar self-relevance and an avatar-emotion connection (i.e., the extent to which interactions between the avatar and virtual objects led to emotional responses) was the most consistently supported across the physiological metrics (i.e., HR deceleration, SCR amplitude, and SCR count). This suggests that an avatar-emotion connection established during avatar use persists beyond avatar use, at least while the user still retains the experience of controlling the avatar in recent memory. Furthermore, this supports the notion that controlling an avatar is an important antecedent of avatar self-relevance because the user experiences emotional responses to interactions between the avatar and virtual objects.

The finding that participants who experienced a greater avatar-body connection (i.e., the extent to which the avatar felt connected to the user’s body schema during use) exhibited less negative emotional valence (as reflected by corrugator EMG activity) in response to watching the avatar get hit supports the notion that for these participants, the disconnection from the avatar was more salient because it emphasized the separation between the self and the avatar, thus diminishing the avatar self-relevance they experienced after use. Conversely, for the participants who experienced less avatar-body connection during avatar use, the disconnection from the avatar did not represent a significant contrast, and thus their avatar self-relevance did not diminish as strongly after use. This relationship is somewhat counterintuitive: More avatar-body connection during use leads to less avatar self-relevance after use, and vice versa.

This relationship differs notably from the relationship between an avatar-emotion connection and avatar self-relevance. The distinction between the two appears to hinge on whether the disconnection from the avatar represents a large contrast for the user. In the case of an avatar-body connection, the disconnection from the avatar presents a significant contrast because the user’s inability to control the avatar is very obvious, and this hinders the avatar’s integration into conception of the self. In the case of an avatar-emotion connection, the disconnection seems to present only a minor contrast, possibly because it does not detract from the memory of recent emotional responses to interactions between the avatar and virtual objects (i.e., the basis for establishing an avatar-emotion connection), and thus avatar self-relevance persists beyond use.

Future research could test these explanations by incorporating designs that facilitate repeated measures throughout the entire avatar-use process as well as manipulations that influence the psychological experience of avatar use. For example, a study could include repeated measures of avatar self-relevance both while the user is designing and/or controlling the avatar (e.g., through a reaction time-based secondary task) and after use (as in this study) and then compare differences across these time periods with differences in avatar-emotion connection, avatar-body connection, and any other facets of the psychological experience of avatar use of interest. Such comparisons would support the present argument if they find that for participants who report a stronger avatar-body connection, avatar self-relevance is stronger during use but weaker after use, and conversely, that participants who report a stronger avatar-emotion connection experience more avatar self-relevance both during and after use.

Future research could also manipulate related independent variables to influence the psychological experience of avatar use. For example, avatar-body connection is conceptualized to be stronger when the user controls the avatar with a device that involves more natural movements (e.g., the Wii Remote) compared with one that is more symbolic (e.g., a keypad; Ratan, 2012). Thus, a study could manipulate avatar-body connection by varying control modality. Similarly, avatar-emotion connection is expected to diminish as the memory of the interactions between the avatar and objects in the avatar’s environment fades, so a study could vary the amount of time between avatar use and the post-use measure of avatar self-relevance with the expectation that decreases in avatar self-relevance over time should be moderated by extent of avatar-emotion connection. Such a study would also offer insights into the rate of avatar self-relevance decay, a topic which the present study was unable to address but which would be of interest to researchers in this area (e.g., Peña, 2011). These examples, as well as any other independent variables that manipulate the psychological experience of avatar use, would allow future research to make stronger causal claims about the findings than the present study, which established time order but only included random assignment for the customization variable (not avatar-body connection or avatar-emotion connection).

The present study’s finding that avatar self-relevance was related to using a same-gender (compared with other-gender) avatar (as reflected by SCR size), but not to avatar customization, confirms the expectation that gender is one of the most—if not the most—important identity characteristics that contributes to avatar self-relevance. The lack of a relationship between customization and avatar self-relevance may be explained by the specific experimental context in which participants were assigned to use a specific avatar gender. Given that avatar users—especially women—tend to prefer same-gender avatars (Ducheneaut et al., 2009; Huh & Williams, 2010; Yee et al., 2011), the (all-female) participants who customized and used a male (other-gender) avatar may have found the experience unnatural, thereby hindering their avatar self-relevance. This may have counteracted any relationship between customization and avatar self-relevance for participants who used same-gender avatars.

This issue highlights an inherent challenge in the present study’s design. Namely, in order to examine the effects of avatar gender consistency, participants were assigned to use specific genders, but gender is usually chosen or customized, not assigned, in virtual spaces. In this way, the present study sacrificed external for internal validity. Future research could provide more valid insights into the relationship between customization (of non-gender characteristics) and avatar self-relevance by allowing all participants to choose their avatar’s gender but only some participants to customize other avatar characteristics. We would expect such a study to support the expectation derived from previous research (Bailey et al., 2009; Ducheneaut et al., 2009; Lim & Reeves, 2009) that customization contributes to avatar self-relevance.

Theoretical Contributions

Together, the present findings suggest that the psychological experience of using an avatar influences avatar self-relevance after use, though not always in the positive direction. When avatar self-relevance is developed through an avatar-emotion connection or through gender consistency during avatar use, then avatar self-relevance persists after use. However, avatar self-relevance developed through an avatar-body connection counteracts post-use avatar self-relevance possibly because the former enhances the salience of the disconnection from the avatar, thereby diminishing the latter. This suggests that both avatar self-relevance and disconnection salience should be considered when examining avatar use effects.

Furthermore, this study suggests that avatar self-relevance, as well as the multiple factors that contribute to it, potentially moderate the relationship between avatar use and post-use avatar effects. This would help explain why previous research has found that controlling (compared with viewing) an avatar augments such post-use avatar effects (Yee & Bailenson, 2009; Yoon & Vargas, 2014). Specifically, controlling an avatar likely contributes to avatar self-relevance, possibly because it engenders an avatar-emotion connection. By definition, when the avatar is relevant to the self, then the schema of avatar-related concepts becomes relevant (or “closely connected”) to the schema of self-related concepts. Thus, the priming of one schema potentially activates the other. If the schemata are associated so closely that the avatar’s characteristics are perceived as part of the user’s self-related schema, then we may call this a shift in self-perception, in accordance with the self-perception theory-based explanation of the Proteus effect (Yee et al., 2009). However, if we consider the two schemata to be independent but closely associated temporarily, then we can still explain such avatar effects as the result of priming and spreading activation through schemata, in accordance with the priming theory–based explanation of the Proteus effect (Peña et al., 2009). In this way, the present approach is somewhat agnostic of the theory used to explain the Proteus effect. We do not want this to be considered a third, new approach, but instead a hybrid that draws from both perspectives and is situated in between them. Future research should continue to examine the role of avatar self-relevance in the Proteus effect, possibly by applying the notions of schema, associations, and self-perception into a novel measurement tool for avatar self-relevance.

Limitations

One notable limitation of the present design is the non-uniformity of the repeated stimuli, that is, the three times the participants observed their avatars being hit. The order of the hits clearly influenced the physiological responses to each hit, with Hit 2 offering the largest and most reliable results across the hit-associated metrics. This may have occurred because there was too much noise in the data in response to Hits 1 and 3. More specifically, given the novelty of Hit 1, participants across all conditions may have responded strongly to it (which is consistent with the SCR amplitude comparisons), thereby reducing the amount of variance caused by the psychological variables of interest. By Hit 2, the participant may have habituated to the novelty of the threat and thus responses may have been moderate enough across participants that psychological connections to the avatar accounted for variance between participants. But then with Hit 3, the stimulus was novel again because it was the game-ending blow that, unlike the other hits, was accompanied with dramatic music and a slow-motion depiction of the avatar falling to the ground, on which the participants may have been especially focused regardless of psychological connection to the avatar. Overall, this may suggest that the influence of avatar self-relevance on avatar use effects is limited in comparison with other psychological factors (e.g., novelty). Future research could examine this potential by measuring avatar self-relevance where the novelty of interactions between the avatar and virtual objects is more controlled.

Furthermore, although the psychophysiological processes of attention and emotional valence should reflect avatar self-relevance, the physiological data did not provide consistent results across the expected relationships of interest. For example, an avatar-emotion connection was associated with ORs but not with changes in emotional valence, while an avatar-body connection was associated with changes in emotional valence responses but not with ORs. It is unclear whether such inconsistency is due to methodological limitations or meaningful differences, for example, that different physiological responses reflect different types of avatar self-relevance. Future research could examine this issue by comparing antecedents of avatar self-relevance across additional physiological measures. One especially interesting approach would be to conduct studies that considerexisting functional magnetic resonance imaging (fMRI) research on the neurological processes involved when personally relevant objects are incorporated into the sense of self (Kim & Johnson, 2013, 2014).

Conclusion

Noting these caveats, the present research still contributes to the theoretical understanding of avatar use effects, which has implications for the meaningful use of avatars. We have introduced the concept of avatar self-relevance, utilized a psychophysiological approach to measuring it after avatar use, illustrated that it is influenced by the psychological experience of controlling the avatar, and connected these findings to the theoretical mechanisms of avatar use effects.

These theoretical implications are potentially relevant to the design of virtual worlds and video games, especially those that intend to influence the users after avatar use. For example, the design and control options for avatars in a virtual classroom may influence post-avatar use engagement in course material (e.g., homework). The present research suggests that if the avatar is integrated into the post-use material (e.g., as an image on the homework), students will pay more attention to (i.e., orient toward) the material if they had previously experienced a stronger avatar-emotion connection but weaker avatar-body connection during avatar use. Such a model may be particularly effective if the avatar gets beaten with a sword for every incorrect answer, but future research should examine this empirically (and ethically).

Overall, this article illustrates an important facet of avatar use in our daily lives. Namely, we do not leave our avatars in virtual cages when we disconnect from them. Just as we put a little bit of ourselves into our avatars, we take a little bit of them with us into the unmediated world, facilitating the experience of “Mii is Me.” The extent to which this happens depends on the psychological experience of avatar use. Given the growing ubiquity of virtual worlds, video games, and other avatar-facilitating media in our everyday lives, it is important that we continue to develop this understanding of avatar uses and effects.

Footnotes

Appendix

Acknowledgements

For their indispensable support on this article and patience through the process, we would like to thank Anne Schell, Christelle Williams, Olga Chandra, Michael Roloff, and the reviewers.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Rabindra A. Ratan is an assistant professor in the Department of Media & Information at Michigan State University. His research focuses on the psychological experience of using mediated self-representations (e.g., avatars) with an emphasis on outcomes of use related to education-related behaviors, health-related behaviors, and stereotypes.

Michael Dawson is a professor in the Department of Psychology at the University of Southern California. His research focuses on the psychophysiology of schizophrenia, psychophysiological measures of conditioning, emotion, and attention, and the neurophysiology of emotion learning.

References

Armel

K. C.

Ramachandran

V. S.

(2003). Projecting sensations to external objects: Evidence from skin conductance response. Proceedings of the Royal Society B: Biological Sciences, 270, 1499-1506. doi:10.1098/rspb.2003.2364

Bailenson

J. N.

Blascovich

(2004). Avatars. In Bainbridge

W. S.

(Ed.), Berkshire encyclopedia of human-computer interaction (pp. 64-68). Great Barrington, MA: Berkshire Publishing Group.

Bailey

Wise

Bolls

(2009). How avatar customizability affects children’s arousal and subjective presence during junk food–sponsored online video games. CyberPsychology & Behavior, 12, 277-283. doi:10.1089/cpb.2008.0292

Bartlett

C. P.

Harris

R. J.

Bruey

(2008). The effect of the amount of blood in a violent video game on aggression, hostility, and arousal. Journal of Experimental Social Psychology, 44, 539-546. doi:10.1016/j.jesp.2007.10.003

Berti

Frassinetti

(2000). When far becomes near: Remapping of space by tool use. Journal of Cognitive Neuroscience, 12, 415-420. doi:10.1162/089892900562237

Bessière

Seay

A. F.

Kiesler

(2007). The ideal elf: Identity exploration in World of Warcraft. CyberPsychology & Behavior, 10, 530-535. doi:10.1089/cpb.2007.9994

Bohlin

Kjellberg

(1979). Orienting activity in two-stimulus paradigms as reflected in heart rate. In Kimmel

H. D.

van Olst

E. H.

Orlebeke

J. F.

(Eds.), The orienting reflex in humans (pp. 169-197). Hillsdale, NJ: Lawrence Erlbaum.

Botvinick

Cohen

(1998). Rubber hands “feel” touch that eyes see. Nature, 391, 756. doi:10.1038/35784

Bradley

M. M.

Lang

P. J.

(2007). Emotion and motivation. In Cacioppo

J. T.

Tassinary

L. G.

Berntson

G. G.

(Eds.), Handbook of psychophysiology (3rd ed., pp. 581-607). New York, NY: Cambridge University Press.

10.

Chandler

Konrath

Schwarz

(2009). Online and on my mind: Temporary and chronic accessibility moderate the influence of media figures. Media Psychology, 12, 210-226. doi:10.1080/15213260902849935

11.

Cramer

(1997). Basic statistics for social research: Step-by-step calculations and computer techniques using Minitab. London, England: Routledge.

12.

Damasio

(1999). The feeling of what happens: Body and emotion in the making of consciousness (1st ed.). Fort Worth, TX: Harcourt College.

13.

Dawson

M. E.

Nuechterlein

K. H.

(1984). Psychophysiological dysfunctions in the developmental course of schizophrenic disorders. Schizophrenia Bulletin, 10, 204-232. doi:10.1093/schbul/10.2.204

14.

Dawson

M. E.

Schell

A. M.

Filion

D. L.

(2007). The electrodermal system. In Cacioppo

J. T.

Tassinary

L. G.

Berntson

G. G.

(Eds.), Handbook of psychophysiology (3rd ed., pp. 159-181). New York, NY: Cambridge University Press.

15.

Dimberg

Petterson

(2000). Facial reactions to happy and angry facial expressions: Evidence for right hemisphere dominance. Psychophysiology, 37, 693-696. doi:10.1111/1469-8986.3750693

16.

Drachen

Nacke

L. E.

Yannakakis

Pedersen

A. L.

(2010). Correlation between heart rate, electrodermal activity and player experience in first-person shooter games. In Spencer

S. N.

(Ed.), Proceedings of the 5th ACM SIGGRAPH Symposium on Video Games (pp. 49-54). New York, NY: ACM. doi:10.1145/1836135.1836143

17.

Ducheneaut

Wen

M.-H.

Yee

Wadley

(2009). Body and mind: A study of avatar personalization in three virtual worlds. Proceedings of CHI 2009. doi:10.1145/1518701.1518877

18.

Fox

Bailenson

Binney

(2009). Virtual experiences, physical behaviors: The effect of presence on imitation of an eating avatar. Presence: Teleoperators and Virtual Environments, 18, 294-303. doi:10.1162/pres.18.4.294

19.

Graham

(1979). Distinguishing among orienting, defense, and startle reflexes. In Kimmel

H. D.

Van Olst

E. H.

Orlebeke

J. F.

(Eds.), The orienting reflex in humans (pp. 137-167). Hillsdale, NJ: Lawrence Erlbaum.

20.

Graham

F. K.

Clifton

R. K.

(1966). Heart-rate change as a component of the orienting response. Psychological Bulletin, 65, 305-320. doi:10.1037/h0023258

21.

Hazlett

R. L.

(2006). Measuring emotional valence during interactive experiences: Boys at video game play. Proceedings of CHI 2006. doi:10.1145/1124772.1124925

22.

Higgins

E. T.

(1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319-340. doi:10.1037/0033-295X.94.3.319

23.

Huh

Williams

(2010). Dude looks like a lady: Gender swapping in an online game. In Bainbridge

(Ed.), Online worlds: Convergence of the real and the virtual (pp. 161-174). London, England: Springer.

24.

Ivory

J. D.

Kalyanaraman

(2007). The effects of technological advancement and violent content in video games on players’ feelings of presence, involvement, physiological arousal, and aggression. Journal of Communication, 57, 532-555. doi:10.1111/j.1460-2466.2007.00356.x

25.

Kilteni

Normand

J. M.

Sanchez-Vives

M. V.

Slater

(2012). Extending body space in immersive virtual reality: A very long arm illusion. PLoS ONE, 7(7). doi:10.1371/journal.pone.0040867

26.

Kim

Johnson

M. K.

(2013). Extended self: Medial prefrontal activity during transient association of self and objects. Social Cognitive and Affective Neuroscience, 7(2), 199-207. doi:10.1093/scan/nst082

27.

Kim

Johnson

M. K.

(2014). Extended self: Spontaneous activation of medial prefrontal cortex by objects that are ‘mine’. Social Cognitive and Affective Neuroscience, 9(7), 1006-1012. doi:10.1093/scan/nst082

28.

Klasen

Weber

Kircher

T. T. J.

Mathiak

K. A.

Mathiak

(2012). Neural contributions to flow experience during video game playing. Social Cognitive and Affective Neuroscience, 7, 485-495. doi:10.1093/scan/nsr021

29.

Klimmt

Hefner

Vorderer

Roth

Blake

(2010). Identification with video game characters as automatic shift of self-perceptions. Media Psychology, 13, 323-338.

30.

Lang

(1990). Involuntary attention and physiological arousal evoked by structural features and emotional content in TV commercials. Communication Research, 17, 275-299. doi:10.1177/009365090017003001

31.

Lang

(1994). What can the heart tell us about thinking. In Lang

(Ed.). (2014). Measuring psychological responses to media messages (pp. 161-174). New York, NY: Routledge.

32.

Lang

Potter

R. F.

Bolls

P. D.

(2009). Where psychophysiology meets the media. In Bryant

Oliver

M. B.

(Eds.), Media effects: Advances in theory and research (3rd ed., pp. 185-206). New York, NY: Routledge.

33.

Larsen

J. T.

Norris

C. J.

Cacioppo

J. T.

(2003). Effects of positive and negative affect on electromyographic activity over zygomaticus major and corrugator supercilii. Psychophysiology, 40, 776-785.

34.

Lee

J.-E. R.

Nass

(2012). Distinctiveness-based stereotype threat and the moderating role of coaction contexts. Journal of Experimental Social Psychology, 48, 192-199. doi:10.1016/j.jesp.2011.06.018

35.

Lim

Reeves

(2009). Being in the game: Effects of avatar choice and point of view on psychophysiological responses during play. Media Psychology, 12, 348-370. doi:10.1080/15213260903287242

36.

Maravita

Iriki

(2004). Tools for the body (schema). Trends in Cognitive Sciences, 8, 79-86. doi:10.1016/j.tics.2003.12.008

37.

Markus

Nurius

(1986). Possible selves. American Psychologist, 41, 954-969. doi:10.1037/0003-066X.41.9.954

38.

Mathiak

Weber

(2006). Toward brain correlates of natural behavior: fMRI during violent video games. Human Brain Mapping, 27, 948-956. doi:10.1002/hbm.20234

39.

Messinger

P. R.

Stroulia

Lyons

Smirnov

Bone

(2008). On the relationship between my avatar and myself. Journal of Virtual Worlds Research, 1(2), 1-17.

40.

Öhman

Hamm

Hugdahl

(2000). Cognition and the autonomic nervous system: Orienting, anticipation, and conditioning. In Cacioppo

J. T.

Tassinary

L. G.

Berntson

G. G.

(Eds.), Handbook of psychophysiology (2nd ed., pp. 533-575). New York, NY: Cambridge University Press.

41.

Peña

Blackburn

(2013). The priming effects of virtual environments on interpersonal perceptions and behaviors. Journal of Communication, 63, 703-720. doi:10.1111/jcom.12043

42.

Peña

Hancock

J. T.

Merola

N. A.

(2009). The priming effects of avatars in virtual settings. Communication Research, 36, 838-856. doi:10.1177/0093650209346802

43.

Peña

J. F.

(2011). Integrating the influence of perceiving and operating avatars under the automaticity model of priming effects. Communication Theory, 21, 150-168. doi:10.1111/j.1468-2885.2011.01380.x

44.

Persky

Blascovich

(2007). Immersive virtual environments versus traditional platforms: Effects of violent and nonviolent video game play. Media Psychology, 10, 135-156. doi:10.1080/15213260701301236

45.

Potter

R. F.

Choi

(2006). The effects of auditory structural complexity on attitudes, attention, arousal, and memory. Media Psychology, 8, 395-419. doi:10.1207/s1532785xmep0804_4

46.

Ratan

R. A.

(2011). Self-presence in online gamers: Differences in gender and genre. Proceedings of the International Society for Presence Research 12th Annual International Workshop on Presence. ISBN: 978-0-9792217-4-3.

47.

Ratan

R. A.

(2012). Self-presence, explicated: Body, emotion, and identity extension into the virtual self. In Luppicini

(Ed.), Handbook of research on technoself (pp.322-336). New York, NY: IGI Global.

48.

Ratan

R. A.

Hasler

B. S.

(2010). Exploring self-presence in collaborative virtual teams. PsychNology Journal, 8, 11-31.

49.

Ratan

R. A.

Sah

Y. J.

(2014). The spawn of presence: Examining the relationship between presence and self-presence. Challenging Presence: Proceedings of the 15th International Society for Presence Research, 123-130.

50.

Ravaja

(2004). Contributions of psychophysiology to media research: Review and recommendations. Media Psychology, 6, 193-235. doi:10.1207/s1532785xmep0602_4

51.

Ravaja

Saari

Kallinen

Laarni

(2006). The role of mood in the processing of media messages from a small screen: Effects on subjective and physiological responses. Media Psychology, 8, 239-265. doi:10.1207/s1532785xmep0803_3

52.

Ravaja

Saari

Salminen

Laarni

Kallinen

(2006). Phasic emotional reactions to video game events: A psychophysiological investigation. Media Psychology, 8, 343-367. doi:10.1207/s1532785xmep0804_2

53.

Ravaja

Turpeinen

Saari

Puttonen

Keltikangas-Järvinen

(2008). The psychophysiology of James Bond: Phasic emotional responses to violent video game events. Emotion, 8, 114-120. doi:10.1037/1528-3542.8.1.114

54.

Relevance. (n.d.). In Oxford dictionaries online. Retrieved from http://www.oxforddictionaries.com/us/definition/english/relevant

55.

Self. (n.d.). In Oxford dictionaries online. Retrieved from http://www.oxforddictionaries.com/us/definition/english/self

56.

Slater

Perez-Marcos

Ehrsson

H. H.

Sanchez-Vives

M. V.

(2009). Inducing illusory ownership of a virtual body. Frontiers in Neuroscience, 3, 214-220. doi:10.3389/neuro.01.029.2009

57.

Sokolov

E. N.

(1963). Higher nervous functions: The orienting reflex. Annual Review of Physiology, 25, 545-580. doi:10.1146/annurev.ph.25.030163.002553

58.

Vasalou

Joinson

A. N.

(2009). Me, myself and I: The role of interactional context on self-presentation through avatars. Computers in Human Behavior, 25(2), 510-520.

59.

Wang

Lang

(2012). Reconceptualizing excitation transfer as motivational activation changes and a test of the television program context effects. Media Psychology, 15, 68-92. doi:10.1080/15213269.2011.649604

60.

Weber

Ritterfeld

Mathiak

(2006). Does playing violent video games induce aggression? Empirical evidence of a functional magnetic resonance imaging study. Media Psychology, 8, 39-60. doi:10.1207/S1532785XMEP0801_4

61.

Wise

Reeves

(2007). The effect of user control on the cognitive and emotional processing of pictures. Media Psychology, 9, 549-566. doi:10.1080/15213260701283186

62.

Yee

Bailenson

J. N.

(2009). The difference between being and seeing: The relative contribution of self-perception and priming to behavioral changes via digital self-representation. Media Psychology, 12, 195-209. doi:10.1080/15213260902849943

63.

Yee

Bailenson

J. N.

Ducheneaut

(2009). The Proteus effect: Implications of transformed digital self-representation on online and offline behavior. Communication Research, 36, 285-312. doi:10.1177/0093650208330254

64.

Yee

Ducheneaut

Yao

Nelson

(2011). Do men heal more when in drag? Conflicting identity cues between user and avatar. Proceedings of CHI 2011, 773-776.

65.

Yoon

Vargas

P. T.

(2014). Know thy avatar: The unintended effect of virtual-self representation on behavior. Psychological Science, 25(4), 1043-1045. doi:10.1177/0956797613519271