Abstract
Virtual environments like online videogames offer increasingly more people the opportunity to socialize purely within the digital plane. These interactions, often done through customizable avatars, have brought about the concept of the “virtual-self,” understood as the multiple identities that can be expressed in virtual environments. This can take many forms and fulfill different psychological roles, from portraying the individual’s ideal-self to allowing them to explore what it feels like to be someone completely different. In this study, we used a constructivist perspective to put forward a typology of the different ways in which players construe their virtual identities in online videogames. We administered the repertory grid technique to 202 online videogame players to measure their perceived discrepancies between their actual-self (how they see themselves), ideal-self (how they wish to be), and virtual-self (how they see their main character in a game of their choice). After conducting a cluster analysis, we found three main patterns of virtual-self construal. The projection-type, where players with a high actual-ideal discrepancy created a virtual character resembling their ideal-self; the exploration-type, where players with a lower actual-ideal discrepancy tended to create a character that was different from both their actual and ideal selves; and the proximal-type, where players created characters that were similar to their actual-self. This typology can be a useful resource for any videogame research that wishes to include an identity perspective, as well as in the diagnosis and treatment of internet gaming disorder.
Introduction
The virtual-self can be described as the traits and characteristics individuals express in virtual environments such as social media1,2 or online videogames,3,4 especially in Massive Multiplayer Online Role-Playing Games (MMORPGs), given their vast scope, customizable characters, and direct avatar-to-avatar interaction. The anonymity of online platforms allows users to express new aspects of themselves and exhibit behaviors they might not express offline. Previous research shows players often create their virtual-self projecting, idealized traits on it, especially those with low self-esteem.5,6 Others enjoy “identity tourism,” adopting alternative, nonideal identities for enjoyment. 7 In this case, the virtual-self has a more performative function. While previous studies have attempted to put forward a typology of virtual-self construal,3,8 most of them used self-administered instruments like the Big Five Inventory. 9 These can have limitations in capturing the nuances of virtual identity, often leading to biased self-reports. 10
Virtual identity research can benefit from tools like personal construct theory,11–14
and the repertory grid technique (RGT),15–17
which assesses an individual’s identity constructs and how they relate to their actual, ideal, or virtual selves more accurately and minimizes social desirability bias.
18
In previous work, we proposed a typology of the different types of virtual identity that online videogame players can construe, according to the distance between their virtual, actual, and ideal selves. We proposed four types of virtual-self construal
19
:
Projection-type: Players with a high actual-ideal discrepancy and low virtual-ideal discrepancy (see Figure 1A). This could be understood as a compensatory strategy to experience in the virtual world the traits that the person would like to have and test them within a safer virtual environment.
20
Exploration-type: People within this category show a low actual-ideal discrepancy, whereas their virtual-self has a high discrepancy with both the actual and ideal-self (see Figure 1B). This would be an example of identity tourism, having a more performative or playful function.7,21,22 Proximal-type: This category represents people who, regardless of how they perceive themselves, have a low actual-virtual discrepancy (see Figure 1C). Rather than seeing their virtual character as a different facet of their identity, people within this type consider it an extension of their offline self. Unspecified-type: Although not contemplated by previous research, we included this category as a mathematical possibility where all three elements of the self were roughly equidistant from one another (see Figure 1D).

Visual representation of the different types of virtual identity construal proposed in our typology. The projection-type
Our main hypothesis was that the optimal clustering model for online videogame players according to their virtual-self construal would match our four-category model.
The proximal-type group would have a significantly lower actual-virtual discrepancy than the others, while the exploration-type and projection-type groups would differ in terms of their virtual-ideal discrepancy (with the exploration-type having a significantly higher discrepancy). Additionally, we would identify a fourth group, characterized by having the same, or very similar, values in all three discrepancies.
Methods
Participants
We reached out to MMORPG players on internet forums, websites, and social media. After conducting a power analysis, we determined a target sample of 200 participants to provide our intended analyses with a sensitivity to effects of η2 = 0.06 with 85% power (α = 0.05).
19
The inclusion/exclusion criteria were as follows:
Inclusion criteria: Being over 17 years old and being an active MMORPG player (some MMORPG players may have periods of inactivity as long as new content is not being released for their preferred game). Exclusion criteria: Playing under the influence of alcohol or other recreational drugs, playing videogames professionally, or having been diagnosed with a severe mental health disorder.
The study started with a brief online survey, and those who met our criteria were contacted to schedule an online interview. Of the 716 people who answered the survey, 202 were finally included in the study. Figure 2 shows the flowchart with the participants who were excluded on each stage and the reasons for their removal.

Flowchart of the participant selection process.
We interviewed players of 29 different games from around the world. Graphs with our participants’ countries of origin and the games they played can be found in the supplementary material. Table 1 summarizes participant sociodemographic and gaming habits data.
Sociodemographic and Game-Related Data
Instruments
Online survey: In an initial survey, we asked participants a series of questions to assess the inclusion/exclusion criteria. We also asked for some sociodemographic and playing habits information, such as which MMORPG they played the most, how long they had been playing, and for how many hours a week.
Repertory grid technique (RGT): The RGT11,15–17 was designed to assess an individual’s construct system. These constructs are expressed in two opposite poles (e.g., happy-unhappy; selfish-generous) and written down as the rows of a two-dimensional grid (see Figure 3). On the grid columns, participants wrote the most relevant people in their life (family, friends, etc.), as well as their actual, ideal, and virtual selves, their most-played MMORPG character (see Figure 3). Finally, participants rated every element (columns) on the grid for each one of their constructs (rows) using a 7-point Likert scale, determined by which pole the corresponding individual would lean toward. A more detailed description of this procedure can be found in the protocol for this study. 19

Example of the grid resulting from the RGT interview.
The resulting data matrix of the grid (see Figure 3 for an example) was mathematically analyzed to extract different variables regarding the individual’s construct system.
23
Actual-ideal discrepancy: The Euclidian distance between the scores from the actual and ideal-self elements, usually interpreted as a self-esteem indicator.
15
According to a validation study with a Spanish sample, the average actual-ideal discrepancy in the general population was estimated at 0.23 (SD = 0.09).
24
Actual-virtual discrepancy: The Euclidian distance between the scores from the actual and virtual-self elements. It measures how similar our participants see themselves to their characters. People with a low discrepancy would see their characters as similar to themselves. In our sample, the average actual-virtual discrepancy was 0.40 (SD = 0.11). Virtual-ideal discrepancy: The Euclidian distance between the scores from the virtual and ideal-self elements. It is an indicator of how idealized the person’s virtual-self is since a low discrepancy indicates a character that reflects their ideal-self. In our sample, the average virtual-ideal discrepancy was 0.32 (SD = 0.12).
Procedure
The protocol for this study was approved by the Bioethics Commission of the Universitat de Barcelona, registered at ClinicalTrials.gov (NCT04551638), and published before the beginning of participant recruitment. 19 We made an online call on internet forums and websites with a high population of MMORPG players.25–27 People who accessed the link were first shown the informed consent form, and after signing it, they were led to a survey where they entered their sociodemographic and gameplay data. The survey was carried out using Qualtrics 28 and had an average duration of 24 minutes.
We contacted participants who met our inclusion criteria to schedule a date for an online interview that was conducted mostly using Discord. 29 During the interviews, a predoctoral researcher administered the RGT, which took around 90 minutes. We used the EYME software 30 to assist in the data gathering, and the relevant variables were computed from each grid using the GRIDCOR6.0 software. 31
Results
Cluster analysis
We conducted a k-means cluster analysis using the three self-discrepancy variables as criteria. The NbClust R package32,33 reported the optimal number of clusters for our data was 3. Clusters 1 and 3 had approximately the same size, whereas cluster 2 was the largest (see Table 2). Cluster 1 was characterized by a low actual-ideal discrepancy and high actual-virtual and virtual-ideal discrepancies. Hence, participants within cluster 1 tended to perceive themselves as closer to how they would like to be and construed a virtual identity that was distant from both their actual and ideal selves. This cluster’s scores align with the “exploration-type” described in our typology.
Mean Scores for the Euclidian Distances of the Three Self-Discrepancy Variables (Actual-Ideal, Actual-Virtual, and Virtual-Ideal) Grouped as per the Resulting Clusters from the k-Means Analysis
People in cluster 2 showed the lowest actual-virtual discrepancy. This cluster aligns with the “proximal-type” described in our typology since the mean actual-virtual discrepancy is considerably lower than in the other two. Last, people in cluster 3 present a high actual-ideal and actual-virtual-self discrepancy. However, they present a low discrepancy between their virtual and ideal selves. This indicates a tendency for people to see themselves as distant from how they would like to be and to create a virtual-self that compensates for that discrepancy, making it resemble their ideal-self, aligning with the “projection-type” of our typology.
Cluster validation
One-way analysis of variances showed significant differences in all the discrepancy variables between clusters. There was a statistically significant difference in actual-ideal discrepancy (F[2, 199] = 89.86, p < 0.001). Pairwise comparisons using the Bonferroni correction showed that people within the projection-type had a significantly higher actual-ideal discrepancy than those within the exploration and proximal types (p < 0.001 in both cases), meaning that people in that cluster tended to have a more negative impression of themselves than the rest. There was also a statistically significant difference between clusters in their actual-virtual discrepancy (F[2, 199] = 94.88, p < 0.001). Pairwise comparisons showed that people within the proximal-type had a significantly lower actual-virtual discrepancy than the projection or exploration types (p < 0.001 in both cases), meaning that they perceived their virtual character as more similar to themselves than the other players. Finally, the analysis of variance showed a statistically significant difference between clusters in their virtual-ideal discrepancy (F[2, 199] = 142.10, p < 0.001). Pairwise comparisons showed that people within the exploration-type had a significantly higher virtual-ideal discrepancy than those in the projection or proximal types (p < 0.001 in both cases), meaning that their virtual-self was further away from their ideal-self. The differences between clusters are represented in Figure 4.

Boxplots with the mean differences between clusters in the three self-discrepancy variables (actual-ideal, actual-virtual, and virtual-ideal). Over each box is shown the average score of the self-discrepancy variable within the indicated cluster. The significance of the mean differences is indicated by the brackets above, where *** indicates p < 0.001, ** indicates 0.001 < p < 0.049, * indicates 0.05 < p < 0.1, and “NS” indicates p > 0.1.
Discussion
The present study aimed to examine the validity of a typology consisting of four different types of virtual-self construal that can be found in MMORPG players. The results provide some evidence for our proposed typology. We found three possible types of virtual-self construal. First, we found players within the proximal-type, whose virtual identity does not differ from their offline one (virtual-self as an extension of their actual-self rather than a different identity). In this case, the virtual-self gives continuity or extends the self toward the virtual environment. When the virtual-self becomes more differentiated from the actual-self, we identified two more types of virtual-self construal. On one hand, we grouped under the projection-type those players who were dissatisfied with their actual-self and chose to create a virtual identity that resembled their ideal-self. For them, the virtual serves as a compensatory strategy for their perceived deficiencies. This is consistent with previous studies where a similar tendency was observed in videogame players.5,6 On the other hand, players included in the exploration-type are considerably satisfied with themselves (low actual-ideal discrepancy) and created a virtual-self that was different from both their actual and ideal selves. This type of virtual identity allows players to experiment with alternative selves rather than being a projection of their ideal-self, serving a performative or exploratory role. This phenomenon aligns with the “identity tourism” phenomenon,7,21,22 where people experience satisfaction from taking a different, not necessarily idealized, identity. Another possibility within this type of player is that they are trying to express their “negative true self.” They might be trying to achieve a sense of self-congruence within the virtual environment by expressing negative traits of themselves that usually would have to be repressed for the offline self-due to social norms. 1
The typology described in this study may help further our understanding of virtual identity. This knowledge could be beneficial in detecting and treating Internet Gaming Disorder (IGD). Past research has already identified links between pathological gaming and players’ perceptions of their characters, as well as the psychological function of their virtual-self.4,8,34,35 Because of that, we believe that the understanding of virtual-self construal could be beneficial to IGD research. Pathological gaming could manifest differently in players with different types of virtual selves or even have different prevalences. Research on virtual-self construal could also be useful in the treatment of IGD, as players with different types of virtual selves may respond better to different therapeutic approaches. On a similar note, this typology could be useful for the development of serious games aimed at improving individuals’ well-being by experiencing different aspects of their selves through virtual avatars. Some research has already been done in this area, although it is still a very new field.36–38
Limitations and Future Research
This study’s limitations include a lengthy interview process and a lack of economic compensation that could have biased participant willingness.
Our typology represents an identity-based perspective for research regarding videogames. Future research on virtual identity should incorporate this typology to understand how virtual identity affects players and their online behavior. It would also be worthwhile to test its applicability in other virtual environments like social media for broader insights. Results in this type of research could inform commercial and health applications, furthering our understanding of how virtual identity affects player experience or its possible use in therapeutic environments.
Conclusions
In this study, we put forward a typology that classifies the different ways in which online videogame players construe their virtual-self. This three-type classification reflects how similar or different the virtual-self is created with regard to the actual-self and ideal-self, and it can serve as an indicator of the psychological function of virtual identity. First, the projection-type of virtual-self construal serves a compensatory role in players’ lives, allowing people with many perceived shortcomings in their offline reality to experience their desired traits through their videogame character. The exploration-type has a more ludic purpose of playfully exploring different identities that are not necessarily ideal in the safe context of online videogames. Finally, the proximal-type has a much more utilitarian function by extending the players’ offline selves into the virtual space, allowing them to interact with it and other players without changing anything about themselves.
This typology could be a useful addition in research focused on the diagnosis and treatment of IGD and other forms of pathological gaming, as well as in the development of therapeutic tools that use the virtual-self to improve different aspects of its users. Furthermore, while this typology has been validated for online videogame players, it could be worthwhile to replicate this study with other types of virtual environments, such as social media or dating apps, to assess whether the distribution of our three clusters changes on different platforms. It might be even possible to observe new types of virtual selves altogether that are exclusive to different types of virtual environments.
Footnotes
Authors’ Contributions
A.G.: Conceptualization, methodology, investigation, software, formal analysis, visualization, and writing—original draft. A.M.: Conceptualization, writing—reviewing and editing, supervision, project administration; G.F.: Conceptualization, software, writing—reviewing and editing, supervision, and project administration.
Author Disclosure Statement
No competing financial interests exist.
Funding Statement
GF and AM have received funding from the Catalan government (Agaur, Generalitat de Catalunya) as an emergent research group (ref. 2021SGR 00666).
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
