User-Avatar discrepancy scale: a comparative measurement of self and avatar views

Figures & data

Table 1. Available measures accessing the connection between the user and their avatar.

Measure	Author	Description
Player Identification Scale (PIS)	Van Looy et al. 2012	The PIS utilises 28 items to evaluate the connection players feel with their avatars, emphasising the shift in players’ self-view through their interactions with their avatars. This instrument divides into three components: similarity identification, wishful identification, and embodied presence. The scoring ranges from 17 to 85, where higher scores demonstrate more stronger avatar identification.
User-Avatar Questionnaire (UAQ)	Blinka 2008	The UAQ assess the UAB from a psychodynamic perspective, incorporating 12 items on a five-point Likert scale. The scale items assess three dimensions: identification, immersion, and compensation. The scores are calculated by adding the values for each item in each subscale, with higher scores reflecting higher experience levels of each respective aspect.
Player-Avatar Identification (PAI) Scale	Li, Liau, and Khoo 2013	The PAI employs 15 questions to evaluate how players relate to their avatars. Responses to statements like ‘It feels like I'm physically part of the game's world when I play’ are captured on a five-point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree).
Self-Presence Questionnaire (SPQ)	Ratan and Dawson 2016	The SPQ evaluates self-presence (SP) across three areas: physical (PSP), emotional (CSP), and identity (ESP), alongside an aggregate SP assessment. It includes 13 questions, such as ‘Do you feel your avatar acts as an extension of your own body in the game?’ answered on a 5-point Likert scale from 0 (not at all) to 4 (completely). Scores from these items are totalled to calculate the overall SP score, from 0 to 52, and scores for each SP area: PSP (from five questions, 0–20), CSP (from five questions, 0–20), and ESP (from three questions, 0–12), where higher scores reflect more intense experiences of SP in each category.

Table 2. Socio-Demographics of the sample.

Variables	Frequency (N = 477)	Valid Percentage (%)
Sex
Male	438	91.8
Female	39	8.2
Highest level of Education
Unfinished elementary school	88	18.8
Elementary school	350	73.4
Vocational Secondary	11	2.3
High school	18	2.4
Occupation
Student (Elementary)	136	29.3
Student (Secondary)	320	68.7
Student (University)	2	.4
Employed	3	.6
Partially Employed	2	.4
Unemployed	3	.6
Type of gaming
Online multiplayer	372	78
Browser multiplayer	24	5
Online singleplayer	68	14.3
Offline singleplayer	12	2.5
Game genre
Battle Royale (e.g. Fortnite)	195	40.9
Role-playing games (e.g. Elden Ring)	182	38.2
First-person shooter (e.g. Apex Legends)	39	8.2
Simulation (e.g. Railway Empire)	35	7.3
Strategy (e.g. Battlestar Galactica)	6	1.3
Other	19	4

Table 3. The User-Avatar Discrepancy Scale (UADS) with all items, prior to the analyses conducted.

Using the pairs of adjectives, how do see yourself? Please try to assign a suitable response for each item.

Using the pairs of adjectives, how do see your gaming avatar? Please try to assign a suitable response for each item

Table 4. Fit Indices, Factor Loadings, and Reliabilities of the EFA in the Calibration Sample.

Model	KMO	Bartlett's Test	Chi-square (χ²)	RMSEA	BIC	TLI	Reliability
Model 1	0.73	p < 0.01	264	0.08	-222	0.59	α ­= 0.72	ω = 0.74
Model 2 (Items 4, 11, 14 removed)	0.77	p < 0.01	173	0.09	-120	0.67	α ­= 0.75	ω = 0.76
Model 3 (Items 3, 4, 5, 11, 12, 13, 14 removed)	0.79	p < 0.01	51.4	0.08	-57.5	0.85	α ­= 0.75	α ­= 0.74
Model 4 (Items 3, 4, 5, 7, 11, 12, 13, 14 removed)	0.78	p < 0.01	41.9	0.09	-34.4	0.84	α ­= 0.74	α ­= 0.74

Note. The Kaiser-Meyer-Olkin (KMO) is an assumption that measures sampling adequacy. Bartlett’s test for sphericity checks the overall significance of all correlations within the matrix. Fit indices considered include chi-square, the root mean square error of approximation (RMSEA), the Bayesian information criteria (BIC), and the Tucker-Lewis Index (TLI). Considering overall fit, it has been supported that if the RMSEA is less than 0.16, another incremental measure of fit, such as TLI, may not be that informative, as it is in fact expected to be less than 0.90 (this applies for model 3 selected here; Harry et al., 2019).

Table 5. The factor loadings and reliabilities of the one-factor EFA in the calibration sample.

Model		Model 1					Model 2 Items 4, 11, & 14 removed					Model 3 Items 3, 4, 5, 11, 12, 13, & 14 removed					Model 4 Items 3, 4, 5, 7, 11, 12, 13, & 14 removed
Item no.	Item Content	Factor	Load-ing ^b	Uniqueness	α	ω	Factor	Load-ing	Uniqueness	α	ω	Factor	Load-ing	Uniqueness	α	ω	Factor	Load-ing	Uniqueness	α	ω
1	Strong / Weak	1	0.49	0.76	0.7	0.72	1	0.49	0.76	0.73	0.74	1	0.55	0.74	0.72	0.73	1	0.52	0.73	0.71	0.72
2	Fast / Slow	1	0.52	0.73	0.7	0.72	1	0.54	0.71	0.73	0.74	1	0.56	0.68	0.72	0.72	1	0.56	0.69	0.7	0.71
3 ^a	Brave / Cowardly	1	0.31	0.9	0.71	0.74	1	0.31	0.91	0.75	0.76
4	Loud / Silent	1	0.16	0.98	0.73	0.75
5	Social / Solitary	1	0.38	0.86	0.7	0.73	1	0.34	0.89	0.75	0.76
6	Active / Passive	1	0.62	0.62	0.68	0.71	1	0.62	0.62	0.71	0.73	1	0.56	0.68	0.71	0.72	1	0.54	0.71	0.7	0.71
7	Observant / Oblivious	1	0.41	0.83	0.7	0.72	1	0.41	0.83	0.74	0.75	1	0.41	0.84	0.74	0.74
8 ^a	Important / Unimportant	1	0.58	0.66	0.69	0.71	1	0.58	0.66	0.72	0.73	1	0.59	0.65	0.71	0.72	1	0.6	0.64	0.7	0.7
9	Good looking / Ugly	1	0.62	0.62	0.69	0.71	1	0.63	0.6	0.71	0.75	1	0.61	0.63	0.7	0.71	1	0.61	0.63	0.68	0.7
10 ^a	Cheerful / Gloomy	1	0.42	0.82	0.7	0.73	1	0.43	0.81	0.73	0.75	1	0.41	0.83	0.74	0.74	1	0.39	0.85	0.73	0.71
11	Masculine / Feminine	1	0.1	0.62	0.73	0.75
12	Honest / Dishonest	1	0.3	0.91	0.71	0.73	1	0.3	0.91	0.74	0.76
13	Just / Unjust	1	0.32	0.9	0.71	0.73	1	0.31	0.9	0.74	0.75
14 ^a	Bold / Diffident	1	0.19	0.96	0.72	0.74
15 ^a	Helpless / Dominant	1	0.49	0.76	0.7	0.72	1	0.49	0.77	0.73	0.74	1	0.53	0.72	0.72	0.73	1	0.56	0.69	0.7	0.71

Note. ^aReversed scaled item. ^bLoading on the primary factor.

Table 6. Fit indices, factor loadings, and reliabilities of the one-factor CFA in the validation sample.

		Fit indices ^a						Reliability
Model	Estimator	df	p-value (X^2)	RMSEA	CFI	TLI	SRMR	α	ω
Model 1	DWLS ^b	20	0.06	0.06	0.96	0.94	0.07	0.69	0.68
Model 2	ML ^c	20	0	0.1	0.81	0.73	0.07	0.69	0.68
Model 3	WLS ^d	20	0.01	0.07	0.69	0.57	0.11	0.69	0.73

Note. ^aThe fit indices provide a conventional cut-off criterion that evaluates the model’s fit. This includes the model chi-square (χ²), the Root Mean square Error of Approximation (RMSEA), the Comparative Fit Index (CFI), Tucker Lewis Index (TLI), and the Standardized Root Mean Square Residual (SRMR). ^bDiagonally weighted least squares (DWLS) is a robust WLS method that is based on the polychronic correlation matrix of the variables in the analysis. An estimated polychoric correlation captures the linear relationship between two normal, latent response variables. It uses only the diagonal of weights in inversion, and all weights in the estimation of fit and standard error (Mîndrilă, 2010). This method is preferable for smaller samples and ordinal data. ^cMaximum likelihood (ML) is the most used estimator in CFA due to its useful statistical properties (i.e., asymptotic unbiasedness, normality, consistency, and maximal efficiency; Li, 2016). The use of ML requires the assumption that the observed variables follow a continuous and multivariate normal distribution given the covariates in the population (Jöreskog, 1969; Satorra, 1990). ^dWeighted least squares (WLS) is an asymptotically distribution-free estimator and is suggested to be used when the data is continuous but non-normal and a consistent estimate of the asymptotic covariance matrix of sample-based variances and covariances is used (Browne, 1984).

Table 7. User Avatar Discrepancy questionnaire item fit data.

Item Number	Item Content	S-χ^2	df	p	RMSEA
1	Strong/Weak	90.3	99	.77	0
2	Fast/Slow	94.43	95	.49	0
6	Active/Passive	123.89	118	.34	.01
7	Observant/Oblivious	145.9	108	.009	.03
8	Important/ Unimportant	115.82	101	.15	.02
9	Good looking/Ugly	125.26	120	.35	.01
10	Cheerful/Gloomy	122.64	124	.52	0
15	Helpless/Dominant	90.12	84	.31	.01

Note. The S-χ² describes the item-fit statistic for each item and behaves similarly to X² in CFA, with insignificant rates showing no deviation of the item modelling from the data. RMSEA values here refer to per-item fit.

Table 8. User Avatar Discrepancy questionnaire item discrimination (α) and difficulty (β) parameters.

Item Number	Item Content	α	β^1	β^2	β^3	β^4	β^5	β^6	β^7	β^8	β^9	β^10	β^11	β^12	λ
1	Strong/Weak	1.2	−5.75	4.84	−3.67	−2.72	2.05	-.64	.45	1.67	2.62	4.53	5.79		.58
2	Fast/Slow	1.23	−5.66	−4.53	−3.48	−3.48	−1.79	-.35	.65	1.68	2.87	4.27	5.1		.59
6	Active/Passive	1.06	−4.81	−4.81	−3.28	−2.63	−1.86	−0.44	.47	1.46	2.42	3.29	4.36		.53
7	Observant/Oblivious	.8	−6.71	−5.32	−4.58	−3.33	−2.4	−1.39	.51	1.94	3.21	4.68	6.67	8.06	.43
8	Important/ Unimportant	1.28	−5.42	−4.55	−3.25	−2.66	−1.87	-.5	.39	1.46	2.33	2.96	4.21		.56
9	Good looking/Ugly	1.16	−4.93	−3.85	−3.01	−2.19	−1.41	-.32	.49	1.26	2.36	3.18	3.86		.6
10	Cheerful/Gloomy	.79	−7.23	−6.72	−5.64	−3.49	−2.19	-.94	.61	1.64	2.61	3.8	4.53	6.75	.43
15	Helpless/Dominant	1.31	−5.42	−4.31	−3.53	−2.35	−1.03	.14	1.18	2.47	3.26	3.86			.61

Note. α = discrimination parameter; β^x = difficulty parameter. β¹ represents the range of values of – 5 to – 4, with negative values suggesting better views of their avatar than themselves. β² represents the range of values of – 4 to – 3. β³ represents the range of values of – 3 to – 2. β⁴ represents the range of values of – 2 to – 1. β⁵ represents the range of values of – 1–0. β⁶ represents the range of values from 0 to 1, with positive values suggesting better views of themselves than their avatar. β⁷ represents the range of values from 1 to 2. β⁸ represents the range of values from 2 to 3. β⁹ represents the range of values from 3 to 4. β¹⁰ represents the range of values from 4 to 5. β¹¹ represents the range of values from 4 to 6. β¹² represents the range of values from 6 to 7. Interestingly, the positive views of self seem to escalate for observant/oblivious (item 7) and cheerful/gloomy (item 10). The λ defines the amount of variance of an item explained by the latent factor.

Figure 1. The user-avatar discrepancy category characteristic curve. The figure is expressed in a nonlinear regression line relating to discrimination (α) and difficulty (β) parameters. Discrimination (α) describes how well an item can differentiate between participants according to their latent trait levels. Thus, items more strongly related to the latent variable present steeper slopes. β reflects the level of the latent trait, theta (θ), at which a participant has an equal probability (50:50) of endorsing an item. For example, ‘easier’ items have lower β values and are presented closer to the horizontal axis.

Figure 2. The user-avatar discrepancy item information curve. Information refers to the ability of an item to accurately estimate scores on the theta (θ). Item-level information demonstrates how well each item contributes to scoring estimation precision, with higher levels of information indicating more accurate score estimates.

Figure 3. The Test Characteristic Curve (TCC) provides a visual representation of expected discrepancy scores (blue line) as a function of the latent trait. The figure demonstrates that the latent trait (i.e. the discrepancy) increases steeply, as the total score reported increases and becomes positive.

Figure 4. The Test Information Function (TIF) curve demonstrates the relationship between the stand errors (SE; dotted line) and reliability indices. The blue line represents the level of scale information for varying levels of the latent trait and the red line is the level of error for similarly varying levels of the latent trait.

Table 9. Correlation matrix of the UADS items with the AEQ Addiction and Engagement factors.

UADS Items		Addiction-Engagement Questionnaire (AEQ)
		Addiction	Engagement
1	Strong/Weak	.141**	.122*
2	Fast/Slow	.055	.165**
6	Active/Passive	.161**	.217***
7	Observant/Oblivious	.105*	.136*
8	Important/ Unimportant	.159**	.05
9	Good looking/Ugly	.153**	.047
10	Cheerful/Gloomy	.165**	.017
15	Helpless/Dominant	.13**	.112
	UADS scale	.207***	.206***

***p < .001.

** p < .01.

* p < .05.

Van Looy, J., C. Courtois, M. De Vocht, and L. De Marez. 2012. “Player Identification in Online Games: Validation of a Scale for Measuring Identification in MMOGs.” Media Psychology 15 (2): 197–221. https://doi.org/10.1080/15213269.2012.674917.

 Web of Science ®Google Scholar

Blinka, L. 2008. “The Relationship of Players to Their Avatars in MMORPGs: Differences Between Adolescents, Emerging Adults and Adults.” Cyberpsychology 2 (1), Article 5.

 Google Scholar

Li, D. D., A. K. Liau, and A. Khoo. 2013. “Player–Avatar Identification in Video Gaming: Concept and Measurement.” Computers in Human Behavior 29 (1): 257–263. https://doi.org/10.1016/j.chb.2012.09.002.

 Web of Science ®Google Scholar

Ratan, R. A., and M. Dawson. 2016. “When Mii is Me: A Psychophysiological Examination of Avatar Self-Relevance.” Communication Research 43 (8): 1065–1093. https://doi.org/10.1177/0093650215570652.

 Web of Science ®Google Scholar

Harry, M. L., J. Acevedo, and T. M. Crea. 2019. “Assessing the Factor Structure of the Spanish Language Parent Strengths and Difficulties Questionnaire (SDQ) in Honduras.” PLoS One 14 (3): e0214394. https://doi.org/10.1371/journal.pone.0214394.

 PubMed Web of Science ®Google Scholar

Mîndrilă, D. 2010. “Maximum Likelihood (ML) and Diagonally Weighted Least Squares (DWLS) Estimation Procedures: A Comparison of Estimation Bias with Ordinal and Multivariate non-Normal Data.” International Journal for Digital Society 1 (1): 60–66. https://doi.org/10.20533/ijds.2040.2570.2010.0010.

 Google Scholar

Li, C. H. 2016. “Confirmatory Factor Analysis with Ordinal Data: Comparing Robust Maximum Likelihood and Diagonally Weighted Least Squares.” Behavior Research Methods 48 (3): 936–949. https://doi.org/10.3758/s13428-015-0619-7.

 PubMed Web of Science ®Google Scholar

Jöreskog, K. G. 1969. “A General Approach to Confirmatory Factor Analysis.” Psychometrika 34 (2): 183–202. https://doi.org/10.1007/BF02289343.

 Web of Science ®Google Scholar

Satorra, A. 1990. “Robustness Issues in Structural Equation Modeling: A Review of Recent Developments.” Quality and Quantity 24 (4): 367–386. https://doi.org/10.1007/BF00152011.

 Web of Science ®Google Scholar

Browne, M. W. 1984. “Asymptotically Distribution-Free Methods for the Analysis of Covariance Structures.” British Journal of Mathematical and Statistical Psychology 37 (1): 62–83. https://doi.org/10.1111/j.2044-8317.1984.tb00789.x.

 PubMed Web of Science ®Google Scholar

Data availability statement

All data and syntax have been made publicly available at Git Hub and can be accessed at https://vas08011980.github.io/UABDISSCALE/UserAvatarBondDiscrepancyQuestionnaireFile.html.