Proteus effect
More actions
The Proteus effect is the tendency of people to change their behavior and attitudes to match the appearance of their avatar, the digital body that represents them in a virtual environment. The term was introduced by Nick Yee and Jeremy Bailenson of Stanford University in a 2007 paper in Human Communication Research, which described "the hypothesis that an individual's behavior conforms to their self-representation independent of how others perceive them".[1] In their first experiments, participants in virtual reality who were given more attractive avatars walked closer to a stranger and disclosed more about themselves, and participants given taller avatars negotiated more aggressively, even though the person they were talking to could not see the manipulation.[1]
The effect has since been tested in dozens of experiments. A 2020 meta-analysis of 46 experiments found a relatively consistent effect size of between .22 and .26, which its authors described as "small-but-approaching-medium",[2] and a follow-up meta-analysis of 56 experiments found stronger effects in studies that used head-mounted displays than in studies that used flat screens.[3] Researchers have suggested that such body-driven changes could matter for VR applications in learning, training, psychotherapy and rehabilitation, although most of the evidence comes from short laboratory experiments.[4]
Definition
Yee and Bailenson named the effect after the Greek god Proteus, whose name is the source of the adjective "protean".[1] Their 2007 paper defined it as a process in which a person's behavior conforms to their own digital self-representation "independent of how others perceive them".[1] In a 2009 follow-up they summarized it as people inferring "their expected behaviors and attitudes from observing their avatar's appearance".[5] Later literature uses similar wording: Beyea and colleagues describe it as "the phenomenon that people tend to conform behaviorally and attitudinally to their avatars' identity characteristics".[3]
The definition separates the Proteus effect from behavioral confirmation, in which other people's expectations about a person's appearance change how they treat that person, and that treatment in turn changes the person's behavior. Yee and Bailenson designed their studies so that the confederate who interacted with each participant could not see the manipulated feature: in the attractiveness study the confederate saw an untextured face, and in the height study the participant's avatar always appeared to the confederate at the confederate's own height.[1] The authors described this exclusion of behavioral confirmation as deliberate, since it isolated the effect of changing a person's own self-representation.[1] Later reviews treat this independence from other people's perceptions as part of the definition.[6]
The effect is distinct from, but closely linked to, the sense of embodiment and body ownership studied in VR. Embodiment research asks whether a user feels that a virtual body is their own; Proteus effect research asks whether that body's appearance changes what the user does or believes.[6] It is also related to presence, since several studies have tested whether stronger presence or ownership makes the effect larger.[7]
History
Original experiments (2007)
The concept came out of Nick Yee's doctoral work at Stanford. His dissertation, The Proteus Effect: Behavioral Modification via Transformations of Digital Self-Representation, was submitted to the Department of Communication in June 2007 with Bailenson as principal adviser.[8] The journal paper published that year reported two experiments carried out in a collaborative virtual environment, where each participant's rendering of the scene could differ from what the other person saw.[1] As an example of how easily self-representation can be changed online, the paper's introduction cited Second Life, whose documentation advertised more than 150 sliders for customizing an avatar.[1]
Participants wore an nVisor SX head-mounted display from NVIS with two 1280 by 1024 pixel panels refreshing at 60 Hz; head orientation was tracked by an orientation sensor and position by an optical tracking system, and the virtual room matched the dimensions of the physical lab.[1] In the first experiment, 32 undergraduates (16 men and 16 women) were given an avatar with either an attractive or an unattractive face of their own gender, chosen in a pretest, and spent 60 to 75 seconds looking at it in a virtual mirror before talking with an opposite-gender confederate. Participants with attractive faces walked closer to the confederate (a mean final distance of 0.98 m, against 1.74 m for unattractive faces) and gave more pieces of information about themselves when asked to introduce themselves.[1]
In the second experiment, 50 undergraduates were given an avatar 10 cm shorter than, the same height as, or 10 cm taller than the confederate, whose avatar had a base height of 172 cm. They then played a version of the ultimatum game, splitting a hypothetical pool of US$100 over four rounds. Participants with tall avatars made the second of their two offers significantly more in their own favor than participants with short avatars, and participants with short avatars were about twice as likely (72%) to accept the confederate's unfair 75/25 split as those with normal (31%) or tall (38%) avatars. The height manipulation had no significant effect on the participants' first offer.[1] In both experiments, coders reviewing participants' guesses about the study's purpose found that no one mentioned attractiveness or height.[1]
Online games and offline behavior (2009)
In a 2009 paper in Communication Research, Yee, Bailenson and Nicolas Ducheneaut of the Palo Alto Research Center extended the work in two directions. The first study used an automated census of three World of Warcraft servers over seven days, covering 76,843 characters, and found that both the height and the rated attractiveness of a character's race were significant predictors of character level, although the model explained little of the variance (adjusted R2 = .01, which the authors called small). The authors noted that this study was correlational and could not rule out players choosing avatars that matched their own personalities.[5] The second study gave 40 undergraduates a tall or short avatar in immersive VR, ran the negotiation task in VR, then repeated it face to face after the headset was removed. Participants who had used tall avatars made their first offer more in their own favor both in VR and in the later face-to-face round, while the second offer and acceptance of the unfair split showed no significant height effect. The authors described this as "tentative evidence that the Proteus Effect may persist in subsequent face-to-face interactions".[5]
A 2008 Stanford Magazine feature on Bailenson's Virtual Human Interaction Lab reported the height and attractiveness findings for a general audience and quoted Bailenson as saying: "It only takes 90 seconds of exposure to a mirror image transformed in age, height or gender to cause drastic changes in behavior."[9]
Priming account
In 2009 Jorge Peña, Jeffrey Hancock and Nicholas Merola examined the effect from a priming perspective: the avatar activates concepts associated with its appearance, which then shape thoughts and behavior automatically, while inconsistent concepts are inhibited. They argued that Yee and Bailenson's self-perception explanation was less able to account for that inhibition of positive thoughts, and called for studies to establish whether the two mechanisms act together or independently.[10] In their desktop experiments, participants who used black-cloaked avatars in group discussions developed more aggressive intentions and less group cohesion than those using white-cloaked avatars, and participants given a Ku Klux Klan-associated avatar wrote more aggressive stories in a Thematic Apperception Test than a control group and less affiliative stories than participants with avatars dressed as doctors.[10]
Yee and Bailenson responded with a 2009 study in Media Psychology that separated "being" an avatar from "seeing" it. Seventy-three undergraduates were given attractive or unattractive avatars and either watched their own reflection in a virtual mirror or watched a recording of a previous participant from the same attractiveness condition, so that both groups saw the same kind of visual stimulus. Afterwards they chose partners on a mock dating website. Participants who had embodied an attractive avatar through the mirror chose more attractive partners than those who had embodied an unattractive one, while the playback groups showed no significant difference.[11] The authors concluded that "embodiment produced significantly larger behavioral changes than mere observation of the same visual stimuli", while acknowledging that more work was needed "to pin down the exact mechanism behind the effect".[11]
Proposed mechanisms
Several explanations have been offered, and a 2024 theoretical review by Anna Martin Coesel, Beatrice Biancardi and Stéphanie Buisine concluded that no single one accounts for every finding.[6]
| Explanation | Proposed process | Origin in the literature |
|---|---|---|
| Self-perception theory | Users infer their own attitudes from external cues, including the look of their avatar, much as an outside observer would | Cited by Yee and Bailenson (2007), based on Daryl Bem's 1972 theory[1] |
| Deindividuation | Anonymity in the virtual setting makes users rely more on situational identity cues, such as the avatar's appearance, than on personal norms | Also cited by Yee and Bailenson (2007)[6] |
| Priming | The avatar activates related concepts and inhibits conflicting ones, without any self-inference | Peña, Hancock and Merola (2009)[10] |
| Cognitive dissonance | A drive to keep attitudes and behavior coherent with external cues, including the avatar, leads users to act in line with it | Proposed as an additional hypothesis by Martin Coesel et al. (2024)[6] |
| Embodiment | A strong sense of agency, self-location and body ownership over the avatar makes the effect possible or stronger | Discussed in several studies; reviewed by Martin Coesel et al. (2024)[6] |
| Perspective-taking | Embodying a member of another group increases empathy and can lead to behavior that runs against stereotypes | A related literature that often predicts the opposite outcome to the Proteus effect[6] |
The role of embodiment is still debated. Kilteni, Bergstrom and Mel Slater found that the stronger the illusion of body ownership in one of their conditions, the larger the behavioral change.[4] Reinhard and colleagues, however, found that reported body ownership did not affect the strength of an avatar-age effect, although spatial presence did.[7] A 2024 study by Louise Dupraz and colleagues manipulated agency and self-location with visuomotor synchrony and first- or third-person perspective, and found that "the Proteus effect was not stronger the more participants embodied the elderly avatar".[12]
Evidence and effect size
Ratan, Beyea, Li and Graciano's 2020 meta-analysis covered 46 quantitative experiments in which avatars with specific characteristics were randomly assigned to participants. It found an effect size between .22 and .26 depending on the subset of studies, with nearly all variance explained, and concluded that the Proteus effect "is a reliable phenomenon" that is "relatively large compared to other digital media effects examined in previous meta analyses".[2]
A second meta-analysis by Beyea, Ratan and colleagues, published in PRESENCE: Virtual and Augmented Reality, added newer studies for a total of 56 and examined how study design affects the result.[3]
| Subset | Studies (k) | Weighted effect size (r) |
|---|---|---|
| All studies | 56 | .23 (95% CI .20 to .26) |
| Virtual reality (head-mounted display) | 25 | .29 |
| Flat screen (computer or console) | 31 | .20 |
| Commercial software | 17 | .22 |
| Custom-built software | 36 | .24 |
Source: Beyea et al. (2022), Table 3 and section 5.[3]
The difference between VR and flat-screen studies was significant, supporting the authors' hypothesis that VR, which tends to produce a stronger sense of embodiment, yields larger effects. The difference between commercial games and custom research software was not.[3] The same paper found signs of possible publication bias: effect sizes correlated negatively with sample size (r = -.51), a trim-and-fill analysis lowered the estimate to r = .19, a selection model adjusted it to .20, and a PET-PEESE adjustment produced a nonsignificant estimate. The authors noted that VR studies had smaller samples on average (a mean of 53.32 participants, against 112.07 for flat-screen studies) and that this should be considered when interpreting the VR result.[3]
Applications in VR
In VR studies, letting participants move their bodies and watch the avatar follow in a virtual mirror is a common way of increasing embodiment.[3] Studies have reported effects on motor behavior, cognition, perception and attitudes:
- Music. Thirty-six participants played a West African djembe drum that was registered with a real drum. Those embodied in a casually dressed, dark-skinned body moved more while drumming than a baseline group and than those embodied in a formally dressed, light-skinned body; a path analysis tied the change to the strength of body ownership and the body's perceived appropriateness for the task.[4]
- Cognitive performance. Domna Banakou, Sameer Kishore and Mel Slater embodied 15 men in a virtual body resembling Albert Einstein and 15 in a normal body of similar age to their own. The Einstein group performed better on a cognitive task once prior IQ was taken into account, with the largest improvement among participants with low self-esteem, and showed reduced implicit bias against older people.[13]
- Walking speed. Participants who had embodied older avatars took significantly longer to walk a set distance after leaving VR than those who had embodied young avatars or a control group, but only during the first half of the walk, which the authors read as a sign that the effect decays quickly.[7]
- Pain. In two experiments with 20 and 44 adults using a head-mounted display and a thermal pain stimulator, participants reported pain scores 15.982% lower when using a muscular avatar than when using a normal one, with differences by gender and by whether the avatar's gender matched the participant's.[14]
- Creativity. In a study described in the 2024 review, engineers who brainstormed with avatars dressed like stereotypical inventors proposed more technology-centered ideas, while those whose avatars looked like ordinary public transport users proposed more user-centered ideas.[6]
- Self-objectification. Jesse Fox, Bailenson and Liz Tricase placed 86 women in a fully immersive virtual environment with sexualized or nonsexualized avatars bearing either the participant's own face or an unknown face. Participants with sexualized avatars reported more body-related thoughts, and those who saw their own faces, particularly on sexualized avatars, expressed more rape myth acceptance.[15]
The effect is not limited to head-mounted displays. In a Nintendo Wii exergame study, men operating normal-weight avatars showed more wrist activity than men using obese avatars, although avatar body size did not affect waist activity.[16] Kilteni and colleagues suggested that such body-driven changes could matter for "learning, education, training, psychotherapy and rehabilitation" in VR,[4] and Martin Coesel and colleagues point to applications that tailor workers' creative processes to specific demands.[6] These proposals overlap with virtual reality therapy, although the studies above are laboratory experiments rather than clinical treatments.
A related line of VR research uses embodiment for perspective-taking. Tabitha Peck, Sofia Seinfeld, Salvatore Aglioti and Mel Slater found that light-skinned women embodied in a dark-skinned virtual body, with full-body visuomotor synchrony and a virtual mirror, showed a larger reduction in implicit racial bias than those embodied in light-skinned or purple-skinned bodies or with no body.[17] The 2024 review notes that this framework often predicts the opposite of the Proteus effect, since stereotype-consistent behavior is what the Proteus effect predicts, and suggests that the framing of the task (for example, explicitly asking participants to imagine life as the person) may explain which outcome appears.[6] Similarly, Rosenberg, Baughman and Bailenson found that participants given the "superpower" of flight in VR later helped an experimenter pick up spilled pens more often than participants who rode as helicopter passengers; the authors suggested that the power of flight "primed concepts and prototypes associated with superheroes".[18]
Limitations and open questions
Findings are not uniform. The 2024 review lists studies that manipulated avatar gender without finding behavioral differences, and suggests that people's complex attitudes toward gender stereotypes may interfere with the expected effect.[6] Some results suggest the effect may be short-lived. In the walking study the slowdown appeared only during the first half of the walk, which the authors read as a possible sign of fast decay.[7] In Yee and colleagues' 2009 height study only the first offer was affected; the authors said this might indicate a quickly dissipating effect, but noted that dissipation did not fully explain why the effect reappeared in the face-to-face round.[5] The publication-bias analyses in the second meta-analysis add uncertainty about the true size of the effect, although PET-PEESE was the only one of its tests that suggested a negligible effect.[3][3]
The review also notes that the use of negative stereotypes about race, gender or body size in such studies has been criticized for a lack of consideration of the sensitive nature of these topics.[6] Its authors argue that most research has focused on demonstrating the effect rather than explaining it, and that the social, affective and cognitive processes behind it remain unclear.[6]
See also
References
- ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 1.10 1.11 1.12 Nick Yee, Jeremy Bailenson (2007). "The Proteus Effect: The Effect of Transformed Self-Representation on Behavior". Human Communication Research, vol. 33, no. 3, pp. 271-290. Author's accepted manuscript, nickyee.com. doi:10.1111/j.1468-2958.2007.00299.x. https://nickyee.com/pubs/Yee%20&%20Bailenson%20-%20Proteus%20Effect%20(in%20press).pdf. Retrieved 2026-10-04.
- ↑ 2.0 2.1 Rabindra Ratan, David Beyea, Benjamin J. Li, Luis Graciano (2020). "Avatar characteristics induce users' behavioral conformity with small-to-medium effect sizes: a meta-analysis of the proteus effect". Media Psychology, vol. 23, no. 5, pp. 651-675. https://doi.org/10.1080/15213269.2019.1623698. Retrieved 2026-10-04.
- ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 David Beyea, Rabindra Ratan, Yiming Lei, Hanjie Liu, Gabriel E. Hales, Chaeyun Lim (2022). "A New Meta-Analysis of the Proteus Effect: Studies in VR Find Stronger Effect Sizes". PRESENCE: Virtual and Augmented Reality, vol. 31, pp. 189-202. MIT Press. doi:10.1162/pres_a_00392. https://static1.squarespace.com/static/62c876b674dbb453d4f28b36/t/6689dc303487e13dd6031194/1720356596383/2022-a-new-meta-analysis-of-the-proteus-effect_fullarticle.pdf. Retrieved 2026-10-04.
- ↑ 4.0 4.1 4.2 4.3 Konstantina Kilteni, Ilias Bergstrom, Mel Slater (2013). "Drumming in Immersive Virtual Reality: The Body Shapes the Way We Play". IEEE Transactions on Visualization and Computer Graphics, vol. 19, no. 4, pp. 597-605. https://doi.org/10.1109/TVCG.2013.29. Retrieved 2026-10-04.
- ↑ 5.0 5.1 5.2 5.3 Nick Yee, Jeremy N. Bailenson, Nicolas Ducheneaut (2009). "The Proteus Effect: Implications of Transformed Digital Self-Representation on Online and Offline Behavior". Communication Research, vol. 36, no. 2, pp. 285-312. Virtual Human Interaction Lab, Stanford University (author's manuscript). doi:10.1177/0093650208330254. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/yee-proteus-implications.pdf. Retrieved 2026-10-04.
- ↑ 6.00 6.01 6.02 6.03 6.04 6.05 6.06 6.07 6.08 6.09 6.10 6.11 6.12 Anna Martin Coesel, Beatrice Biancardi, Stéphanie Buisine (2024). "A theoretical review of the Proteus effect: understanding the underlying processes". Frontiers in Psychology, vol. 15. https://doi.org/10.3389/fpsyg.2024.1379599. Retrieved 2026-10-04.
- ↑ 7.0 7.1 7.2 7.3 René Reinhard, Khyati Girish Shah, Corinna A. Faust-Christmann, Thomas Lachmann (2020). "Acting your avatar's age: effects of virtual reality avatar embodiment on real life walking speed". Media Psychology, vol. 23, no. 2, pp. 293-315. https://doi.org/10.1080/15213269.2019.1598435. Retrieved 2026-10-04.
- ↑ Nick Yee (2007-06). "The Proteus Effect: Behavioral Modification via Transformations of Digital Self-Representation (PhD dissertation)". Stanford University, Department of Communication. https://nickyee.com/pubs/Dissertation_Nick_Yee.pdf. Retrieved 2026-10-04.
- ↑ Kara Platoni (2008-01). "Seeing Is Believing". Stanford Magazine. https://stanfordmag.org/contents/seeing-is-believing. Retrieved 2026-10-04.
- ↑ 10.0 10.1 10.2 Jorge Peña, Jeffrey T. Hancock, Nicholas A. Merola (2009). "The Priming Effects of Avatars in Virtual Settings". Communication Research, vol. 36, no. 6, pp. 838-856. doi:10.1177/0093650209346802. https://socialmedialab.sites.stanford.edu/sites/g/files/sbiybj22976/files/media/file/pena-cr-priming-effects.pdf. Retrieved 2026-10-04.
- ↑ 11.0 11.1 Nick Yee, Jeremy N. Bailenson (2009). "The Difference Between Being and Seeing: The Relative Contribution of Self-Perception and Priming to Behavioral Changes via Digital Self-Representation". Media Psychology, vol. 12, no. 2, pp. 195-209. https://doi.org/10.1080/15213260902849943. Retrieved 2026-10-04.
- ↑ Louise Dupraz, Marine Beaudoin, Michel Guerraz, Julien Barra (2024). "Does the avatar embodiment moderate the Proteus effect?". International Journal of Human-Computer Studies, vol. 187, article 103272. https://doi.org/10.1016/j.ijhcs.2024.103272. Retrieved 2026-10-04.
- ↑ Domna Banakou, Sameer Kishore, Mel Slater (2018). "Virtually Being Einstein Results in an Improvement in Cognitive Task Performance and a Decrease in Age Bias". Frontiers in Psychology, vol. 9, article 917. https://doi.org/10.3389/fpsyg.2018.00917. Retrieved 2026-10-04.
- ↑ Youchan Yim, Zongheng Xia, Yuki Kubota, Fumihide Tanaka (2024). "The proteus effect on human pain perception through avatar muscularity and gender factors". Scientific Reports, vol. 14, article 11332. https://doi.org/10.1038/s41598-024-61409-4. Retrieved 2026-10-04.
- ↑ Jesse Fox, Jeremy N. Bailenson, Liz Tricase (2013). "The embodiment of sexualized virtual selves: The Proteus effect and experiences of self-objectification via avatars". Computers in Human Behavior, vol. 29, no. 3, pp. 930-938. Virtual Human Interaction Lab, Stanford University. doi:10.1016/j.chb.2012.12.027. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/fox-chb-sexualized-virtual-selves.pdf. Retrieved 2026-10-04.
- ↑ Jorge Peña, Subuhi Khan, Cassandra Alexopoulos (2016). "I Am What I See: How Avatar and Opponent Agent Body Size Affects Physical Activity Among Men Playing Exergames". Journal of Computer-Mediated Communication, vol. 21, no. 3, pp. 195-209. https://doi.org/10.1111/jcc4.12151. Retrieved 2026-10-04.
- ↑ Tabitha C. Peck, Sofia Seinfeld, Salvatore M. Aglioti, Mel Slater (2013). "Putting yourself in the skin of a black avatar reduces implicit racial bias". Consciousness and Cognition, vol. 22, no. 3, pp. 779-787. https://doi.org/10.1016/j.concog.2013.04.016. Retrieved 2026-10-04.
- ↑ Robin S. Rosenberg, Shawnee L. Baughman, Jeremy N. Bailenson (2013). "Virtual superheroes: using superpowers in virtual reality to encourage prosocial behavior". PLOS ONE, vol. 8, no. 1, e55003. https://doi.org/10.1371/journal.pone.0055003. Retrieved 2026-10-04.