TYPE Original Research

PUBLISHED 07 January 2025

DOI 10.3389/fpsyg.2024.1444039

The Bergen Facebook addiction

OPEN ACCESS scale: a reliability generalization
meta-analysis
EDITED BY
Saleem Alhabash,
Michigan State University, United States

REVIEWED BY
Bo-Ching Chen, Jian-Ling Ma 1*, ZhengCheng Jin 1 and Chang Liu 2*
CTBC Business School, Taiwan 1
Chongqing University of Posts and Telecommunications, Chongqing, China, 2 Yangtze Normal
Paolo Soraci,
University, Fuling District, China
Other, Roma, Italy

*CORRESPONDENCE
Jian-Ling Ma The Bergen Facebook addiction scale (BFAS) is a screening instrument frequently
jianling_ma@163.com
Chang Liu
used to evaluate Facebook addiction. However, its reliability varies considerably
294332805@qq.com across studies. This study aimed to evaluate the reliability of the BFAS and its
RECEIVED 13June 2024 adaptation, the Bergen Social Media Addiction Scale (BSMAS), and to identify
ACCEPTED 21 November 2024 which study characteristics are associated with this reliability. We performed a
PUBLISHED 07 January 2025
reliability generalization meta-analysis involving 173,641 participants across 127
CITATION articles, which reported 147 Cronbach’s alpha values for internal consistency.
Ma J-L, Jin Z and Liu C (2025) The Bergen
The random-effects model revealed that the pooled Cronbach’s alpha values
Facebook addiction scale: a reliability
generalization meta-analysis. were 0.8535 (95% CI [0.8409, 0.8660]) for the BFAS and 0.8248 (95% CI [0.8116,
Front. Psychol. 15:1444039. 0.8380]) for the BSMAS. Moderator analyses indicated that the mean and standard
doi: 10.3389/fpsyg.2024.1444039
deviation of the total scores accounted for 10.06 and 36.7% of the total variability
COPYRIGHT in the BFAS alpha values, respectively. For the BSMAS, the standard deviation of
© 2025 Ma, Jin and Liu. This is an
open-access article distributed under the
the total scores and sample size accounted for 13.54 and 10.22% of the total
terms of the Creative Commons Attribution variability alpha values, respectively. Meta-ANOVA analyses revealed that none
License (CC BY). The use, distribution or of the categorical variables significantly affected the estimated alpha values for
reproduction in other forums is permitted,
provided the original author(s) and the
either the BFAS or BSMAS. Our findings endorse the BFAS and BSMAS as reliable
copyright owner(s) are credited and that the instruments for measuring social media addiction.
original publication in this journal is cited, in
accordance with accepted academic
KEYWORDS
practice. No use, distribution or reproduction
is permitted which does not comply with Facebook addiction, Facebook addiction scale, reliability, reliability generalization,
these terms.
meta-analysis

1 Introduction
Social media addiction is a psychological condition characterized by an excessive focus
on social media platforms. Individuals with this addiction feel a strong compulsion to use
social media and invest substantial time and energy into it, often at the expense of their social
activities, learning, interpersonal relationships, mental health, and overall well-being
(Andreassen and Pallesen, 2014). Research has consistently highlighted the detrimental health
effects of social media addiction, including sleep disturbances (Ho, 2021; Marino et al., 2018),
impaired decision-making (Delaney et al., 2018), and increased risk of depression (Ho, 2021;
Mamun and Griffiths, 2019; Seabrook et al., 2016). Therefore, accurately assessing social media
addiction is crucial for understanding its underlying mechanisms and potential harmful
effects. The Bergen Facebook addiction scale (BFAS; Andreassen et al., 2012) is a widely
utilized tool for assessing Facebook addiction. It is a self-report scale designed primarily for
college students and is based on six criteria: salience, tolerance, mood modification, relapse,
withdrawal, and conflict, as defined by Brown (1993) and Griffiths (1996). The BFAS includes
a 6-item short version and an 18-item standard version. Each item is rated on a 5-point Likert
scale (1 = very rarely, 5 = very often). The total score is calculated by summing individual item
scores, with higher scores indicating greater levels of Facebook addiction. The higher the total
score, the more severe the addiction to the Facebook platform. Preliminary findings indicate

Frontiers in Psychology 01 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

that the BFAS demonstrates good validity. Total BFAS scores correlate generalization studies. The research protocol was registered with the
well with other measures of Facebook activity, neuroticism, and International Prospective Register of Systematic Reviews (PROSPERO
extraversion, and show a negative relationship with conscientiousness. ID: CRD42021295390) to ensure transparency and adherence to
Additionally, higher BFAS scores are associated with delayed sleep systematic review standards.
onset and wake times (Andreassen et al., 2012). Given the proliferation
of social media platforms beyond Facebook, researchers have adapted
the BFAS to assess addiction across various platforms through the 2.1 Study search strategy
Bergen Social Media Addiction Scale (BSMAS; Schou Andreassen
et al., 2016). Both the BFAS and BSMAS have been translated into Systematic searches were conducted in the EBSCO, Elsevier,
several languages, including German (Brailovskaia et al., 2018), Springer, ProQuest, Wiley Online Library, and CNKI databases using
Spanish (Elphinston et al., 2022), Portuguese (da Veiga et al., 2019), keywords such as ‘Facebook addiction,’ ‘social media addiction,’ and
and Chinese (Yam et al., 2019), due to their demonstrated validity. related terms. No search limits were applied. In addition, backward
In classical test theory, reliability refers to how consistently a searches were performed from recent qualitative reviews and key
measurement tool produces results. It is typically defined as the ratio of studies to identify additional relevant articles. The final search was
true score variance to the total variance, reflecting the proportion of completed on December 30, 2021.
variance in scores due to the true score rather than measurement error
(Higgins et al., 2003; Rodriguez and Maeda, 2006). Cronbach’s alpha is
commonly used to assess reliability because it provides a measure of 2.2 Study selection criteria
internal consistency, indicating how well the items in a scale measure
the same underlying construct. Researchers frequently use it as the To be included in this reliability generalization meta-analysis, studies
reliability indicator for the BFAS. The BFAS itself has a good Cronbach’s had to meet the following criteria: (1) Published in English or Chinese;
alpha (0.83). Studies using the BFAS have also found high internal (2) Empirically reported Cronbach’s alpha values for the scales used; (3)
consistency reliability in specific contexts. However, there are several Published in a peer-reviewed academic journal or as a dissertation to
issues with reporting this reliability indicator in studies. For the BFAS, ensure quality. Figure 1 illustrates the study selection process.
Cronbach’s alpha ranges from 0.66 (Błachnio et al., 2017; Brailovskaia
et al., 2023) to 0.94 (Satici, 2019; Soraci et al., 2023). Similarly, for
BSMAS, it varies from 0.66 (Chung et al., 2019) to 0.92 (Brailovskaia 2.3 Data extraction and coding
et al., 2019; Hoşgör et al., 2021). These variations highlight significant
discrepancies in reported reliability. Another major issue is that, when Characteristics were only extracted from studies that reported the
some studies have used the BFAS or BSMAS, they report reliability target Cronbach’s alpha values. To examine how study characteristics
values from previous research rather than calculating them from their influenced alpha values, the following category moderators were
own data, which can lead to inaccurate or misleading conclusions. This coded: COVID-19, administration, country, test language, study aim,
phenomenon of omitting or improperly reporting reliability values is study nature, test length, participant group, sampling method, and
an issue of reliability induction (Henson and Thompson, 2002). It is social media platform. Continuous variables included publication
clear that reliability is context-dependent and can vary based on sample year, sample size, mean and standard deviation of sample age, female
and testing conditions. Discrepancies in reliability estimates as well as proportion, and mean and standard deviation of the total score.
reliability induction threaten the reliability of statistical analyses and Missing data for studies was recorded as such, and no imputation was
research conclusions based on such indicators. Therefore, although the performed. The coding manual, developed by the first and second
BFAS and BSMAS are widely used, no study has systematically explored authors, included detailed guidelines for extracting and categorizing
the variability in the reliability of the two tools in different test scenarios study characteristics. The coding process involved dual coding of a
and estimated their overall reliability. random sample of 40 studies to ensure accuracy. Disagreements
Reliability generalizability analysis is a method that evaluates the between coders were assessed using the intraclass correlation
average reliability of a measurement tool, explores variability in coefficient (ranging from 0.99 to 1) for continuous variables and kappa
reliability across studies, and identifies factors that affect reliability coefficients (ranging from 0.87 to 1.00) for categorical variables.
(Henson and Thompson, 2002; Vacha-Haase, 1998). This study uses
this approach to address gaps in the current research on the BFAS and
BSMAS. This study aims to (1) estimate the average internal 2.4 Data analysis
consistency reliability of the BFAS and BSMAS, (2) assess the
variability in reliability across different studies, (3) identify research The current study used Cronbach’s alpha values for the reliability
characteristics that might influence reliability, and (4) address issues generalization meta-analysis. The transformation method was not
related to reliability induction. employed based on recommendations by Thompson and Vacha-Haase
(2000), who suggested that it may not be necessary for the analyses
conducted. Random-effect models using a frequentist framework were
2 Materials and methods chosen for their ability to account for variability between studies, in line
with standard practice in meta-analysis. Inverse variance was used as
The review methods and reporting followed the Reliability the weighting method. The between-study variance estimator is a
Generalization Meta-analysis (REGEMA) guidelines (Sánchez-Meca restricted maximum likelihood method. The confidence limits of the
et al., 2021), which outline best practices for conducting reliability overall reliability estimates were computed using the method proposed

Frontiers in Psychology 02 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

FIGURE 1
Flowchart of selection and inclusion of articles.

by researchers (Knapp and Hartung, 2003). Heterogeneity was assessed The asymmetry of the funnel plots indicates that there are potential
using the Q test and the I2 index, which indicates the percentage of total coefficient alphas that were not included in the current meta-analysis,
variation across studies due to heterogeneity rather than chance and the number of these coefficient alphas can be estimated using the
(Thompson and Vacha-Haase, 2000). I2 values of 25, 50, and 75% trim-and-fill method. The fail-safe number (Durlak and Lipsey, 1991)
correspond to low, moderate, and high heterogeneity, respectively was calculated to assess the robustness of the meta-analytic findings
(Higgins et al., 2003). To explain the variance of alpha values, moderator against publication bias. Meta-analytic results were considered reliable
analysis was applied. Specifically, meta-analyses of variances (meta- if the fail-safe number exceeded the critical value of 5 × k + 10, where
ANOVA) and meta-regression analyses were applied for categorical k represents the number of studies included in the analysis. If the fail-
and continuous variables, respectively. Moreover, the adjustments safe number falls below this critical value, publication bias or file
method proposed by Knapp and Hartung were used (Knapp and drawer problems may exist.
Hartung, 2003) to examine the statistical significance of the moderator All statistical analyses were performed using the metafor
variable and explain the residual heterogeneity. The QW and QE indices (Viechtbauer, 2010) package (V3.8) in the R program 4.1.2. for Windows.
were used to examine model misspecification of meta-ANOVA and
meta regression, respectively. Furthermore, the present study also
employed R2 as an index to quantify the degree of variance explained 3 Results
by the moderator variables. If more than one moderator contributed to
the variance of the coefficient alpha, a multiple meta-regression analysis 3.1 Description of sample
was conducted to identify the unique contributions of the moderators.
Publication bias was assessed using funnel plots for both BFAS In total, 127 articles, 147 reliability values, and 173,641 subjects
and BSMAS, and the trim-and-fill method (Duval and Tweedie, 2000) were included in the formal meta-analysis (reliability-induced articles
was used to estimate and adjust for any asymmetry in the funnel plots. were excluded). Among the 127 articles that reported the target

Frontiers in Psychology 03 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

reliability values, the distribution of the number of reliability values in BSMAS; more specific results are presented in Table 1. To further
the top five countries were: China (21), Germany (20), the United States investigate the sources of heterogeneity in the overall reliability
(19), Italy (18) and Turkey (11); Concerning the language of the scale, estimates, we first analyzed the moderating effect of the version; the
the top five languages were English (38), Chinese (20), German (20), results showed that the moderating effect of the version was
Italian (18) and Turkish (11), and the distribution of the number of significant, QM = 10.03, p < 0.0001, QE = 5851.985, p < 0.0001,
subjects was: Norway (47,283), China (46,712), Italy (11,435), the τ2 = 0.0027, R2 = 7.04%. The high heterogeneity observed indicates
United States (10,137), and Hungary (4073). From the perspective of substantial variability in reliability estimates across different studies,
the language used in the scale, the top five languages in terms of the suggesting that factors such as sample characteristics and study
number of subjects were English (70221), Chinese (45797), Italian conditions significantly influence reliability. The overall heterogeneity
(11435), Hungarian (7043), and German (6840). From the perspective of each version was tested to further investigate the heterogeneity of
of sample distribution, the mixed sample, undergraduate student the reliability values of different versions. The estimation results are
sample, unknown sample, and adult sample were 96,756, 35,871, presented in Table 1.
34,600, 3,499, and 2,915, respectively. The number of subjects using the The meta-analysis revealed that the BFAS and BSMAS had high
BFAS was 122,018, and the number using the BSMAS was 51,623. internal consistency, with Cronbach’s alpha values of 0.85 and 0.82,
Sixteen articles had reliability-induction issues. Specifically, 12 respectively, indicating strong reliability. The results also indicate
studies using BFAS had been induced, eight studies omitted reports, heterogeneity in the reliability values of both the BFAS and BSMAS;
and four studies introduced initial reliability. The BSMAS had four therefore, further analyses are required.
studies that had been induced, three studies omitted reports, and one
study introduced initial reliability.
3.3 Moderation analysis

3.2 Overall reliability estimates and test of 3.3.1 Meta-ANOVA for category variables
heterogeneity The summary results of the meta-ANOVA for the categorical
variables are shown in Table 2. For all category variables, the estimated
The averaged estimated Cronbach’s alpha value is 0.8407 (95% CI average alpha values of the BFAS and BMAS were not statistically
[0.8313, 0.8500]) without considering the version of BFAS and significant. For COVID-19, administration, country, test language,
study aim, study nature, test length, participant group, sampling
TABLE 1 Mean alpha reliability of BFAS and BSMAS and Heterogeneity method, and social media platform, none of the variables exerted an
analysis results. effect on the average internal consistency reliability of the BFAS and
BSMAS. Tables 3 present the reliability estimates between the different
95%CI
levels of the category variables.
n α+ LL UL τ2 Q I2
(%) 3.3.2 Meta-regression for continuous variables
Total 147 0.8407 0.8313 0.8500 0.003 5852.7054*** 98.84 As shown in Table 4, the mean and standard deviation of the total
BFAS 82 0.8535 0.8409 0.8660 0.003 2934.9131*** 97.95 score for BFAS account for 10.06 and 36.7% of the variance of alpha
values, respectively. Together, these two variables explained 66.57% of
BSMAS 65 0.8248 0.8116 0.8380 0.002 2917.0719*** 99.04
the variance in alpha values. For the BSMAS, the standard deviation
n = number of Cronbach’s alpha coefficients; α+ = estimated average alpha value;
of the total score and sample size accounted for 13.54 and 10.22% of
H2 = sampling variability index; Tau2 = estimated total heterogeneity; Q = heterogeneity
statistics in the distribution of the Cronbach’s alpha. I2: total heterogeneity index. the variance in alpha values, respectively. However, these two variables
***p < 0.001. explain only 8.51% of the variance.

TABLE 2 Meta-ANOVA for category variables.

moderator BFAS BSMAS

R 2
F values p R 2
F values p
COVID-19 0.000 0.1275 0.7220 0.000 0.0431 0.8362

Administration 0.0123 0.9998 0.3977 0.000 1.0456 0.3916

Country 0.0626 1.1402 0.3350 0.0982 1.5716 0.1173

Test language 0.0000 0.9785 0.4947 0.0747 1.5000 0.1543

Study aim 0.0000 0.4171 0.5203 0.0000 0.0102 0.9898

Study nature 0.0000 0.0014 0.9701 0.0000 1.0319 0.3136

Test length 0.0000 0.7980 0.3744 - - -

Participant group 0.0000 0.5398 0.6565 0.000 0.2115 0.8800

Sampling method 0.0157 1.4653 0.2305 0.000 0.9129 0.4401

Social media platform 0.0000 0.2597 0.8542 0.0551 2.5412 0.0870

R = proportion of variance explained by the moderator variable; F values = test of moderators; p = p value for the statistical tests of F values.
2

Frontiers in Psychology 04 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

TABLE 3 Comparison of coefficient alpha estimates of the different levels of category variables.

Moderator BFAS BSMAS

n α+ 95%CI n α+ 95%CI
COVID-19
No 80 0.8538 [0.8410,0.8666] 56 0.8242 [0.8098,0.8385]

Yes 2 0.8394 [0.7601,0.9187] 9 0.8281 [0.7930,0.8632]

Administration
Online + Paper-pencile 1 0.8100 [0.6980,0.9220] 1 0.7700 [0.6649,0.8751]

NR 11 0.8401 [0.8052,0.8751] 13 0.8310 [0.8009,0.8610]

Online 45 0.8615 [0.8444,0.8786] 38 0.8309 [0.8137,0.8480]

Paper-pencile 23 0.8417 [0.8183,0.8652] 11 0.8000 [0.7665,0.8334]

Telephone — — — 1 0.8000 [0.6943,0.9057]

Study aim
Correlation 79 0.8542 [0.8414,0.8670] 60 0.8247 [0.8107,0.8388]

Experiment 3 0.8297 [0.7554,0.9040] 3 0.8271 [0.7588,0.8953]

Psychometric — — — 1 0.8180 [0.7108,0.9252]

Study nature
Applied 76 0.8534 [0.8402,0.8665] 58 0.8271 [0.8132,0.8410]

Confirm 6 0.8543 [0.8082,0.9004] 7 0.8050 [0.7638,0.8461]

Test length
Long version 5 0.8752 [0.8252,0.9252] — — —

Short version 77 0.8520 [0.8390,0.8650] 65 0.8248 [0.8116,0.8380]

Participant group
Adolescents 8 0.8517 [0.8111,0.8922] 8 0.8153 [0.7783,0.8524]

Adults 2 0.8401 [0.7590,0.9211] 2 0.8450 [0.7723,0.9177]

Mixed 38 0.8594 [0.8403,0.8784] 33 0.8275 [0.8093,0.8457]

Undergraduate 30 0.8416 [0.8205,0.8628] 18 0.8264 [0.8013,0.8515]

Sampling method
Convience 61 0.8527 [0.8383,0.8671] 53 0.8206 [0.8059,0.8353]

NR 7 0.8263 [0.7819,0.8706] 1 0.8300 [0.7234,0.9366]

Random 9 0.8550 [0.8174,0.8926] 7 0.8550 [0.8159,0.8941]

Snow balling 5 0.8965 [0.8460,0.9469] 4 0.8226 [0.7702,0.8750]

Social media platform

Facebook 55 0.8538 [0.8381,0.8695] 2 0.8210 [0.7497,0.8923]

Instagram 3 0.8440 [0.7776,0.9104] 1 0.7000 [0.5882,0.8118]

Snap chat 1 0.9000 [0.7869,1.0131] — — —

Social media 23 0.8516 [0.8279,0.8753] 62 0.8268 [0.8137,0.8400]

3.4 Publication bias methods were calculated for this study; the BFAS values for the
methods were 21,193,629 (Rosenthal), 65 (Owen), and 44,612,242
To investigate the publication bias of the BFAS and BSMAS, (Rosenberg), indicating that the reliability generalization results are
corresponding funnel plots were drawn; the results are shown in relatively reliable.
Figures 2, 3. The trim-and-fill method results showed that the number
of studies on right-side BFAS trimming was zero (SE = 5.1902),
indicating no publication bias; for BSMAS, the number of studies on 4 Discussion
right clipping was zero (SE = 4.6142), indicating no publication bias.
The BFAS internal consistency reliability measurements for the The purpose of this study was to conduct a meta-analysis of the BFAS
Rosenthal (24,493,065), Owen (82), and Rosenberg (22,335,449) and BSMAS’s internal consistency reliability using a reliability

Frontiers in Psychology 05 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

TABLE 4 Results of simple meta-regression analysis by the continuous moderator variables.

Moderator n b QM p R2 QE
BFAS
Publication year 82 −0.0033 0.8094 0.3710 0.0000 2819.2302****

Mean of age 72 0.0008 0.6733 0.4147 0.0017 2613.9072****

SD of age 71 0.0021 1.1027 0.2973 0.0081 2527.3068****

Female proportion 72 −0.0006 0.2098 0.6484 0.0000 2720.5324****

Mean of total score 33 0.0027 4.2112 0.0487 0.1006 1245.9475****

SD of total score 33 0.0142 16.8116 0.0003 0.3670 961.4841****

Sample size 82 0.0000 1.0619 0.3059 0.0084 2203.5410****

Mean of total score + SD of total score 33 - 25.8499 0.0001 0.6657 590.3401****

BSMAS
Publication year 65 −0.0064 1.7842 0.1864 0.0217 2785.5586****

Mean of age 60 0.0008 0.7624 0.3862 0.0039 2485.9618****

SD of age 60 0.0030 3.1002 0.0836 0.0360 2584.7461****

Female proportion 59 0.0706 1.8243 0.1821 0.005 2419.3051****

Mean of total score 35 −0.0092 3.5271 0.0692 0.0624 606.2396****

SD of total score 35 0.0446 6.4296 0.0161 0.1354 905.6529****

Sample size 65 0.0000 6.7404 0.0117 0.1022 1755.9959****

SD of total score + sample size 46 - 2.3875 0.1039 0.0851 618.0904****

n = number of the Cronbach’s alpha coefficient for each subgroup of moderator variable; b = unstandardized regression coefficient; QM = significance test of moderator regression coefficient; p = p
value of significance test; R2 = total amount proportion of variance accounted for; QE = statistic for test of residual heterogeneity. ****p < 0.0001. Bold values indicate that the p values is significant.

FIGURE 2
Funnel plot of BFAS.

Frontiers in Psychology 06 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

FIGURE 3
Funnel plot of BSMAS.

generalization method. The study found that the internal consistency from a side perspective. However, the BSMAS is more heterogeneous
reliabilities of the BFAS and BSMAS were 0.8535 (95% CI [0.8409, than the BFAS, which may be related to the measured social media
0.8660]) and 0.8248 (95% CI [0.8116, 0.8380]), respectively. Second, there platforms. The BFAS was specifically designed to investigate Facebook
was high heterogeneity between the BFAS and BSMAS studies. Third, platform addiction, whereas the BSMAS measures a wider range
Category variables did not significantly moderate the reliability of either of platforms.
scale. However, for the BFAS, the mean and standard deviation of total In addition to examining the average estimation reliability and
scores were significant moderators, whereas for the BSMAS, only the heterogeneity of the BFAS and BSMAS, this study examined the
standard deviation of total scores and sample size played a role. Fourth, impact of research characteristic variables (continuous and category
reliability-induction issues were noted in both scales. For the BFAS, eight variables) on their reliability. The results showed that, for the BFAS, the
studies failed to report reliability values, and four studies reported initial moderating effect of category variables on the average estimate of
reliability values incorrectly. For the BSMAS, three studies omitted reliability was not statistically significant, while the mean and standard
reliability reporting, and one incorrectly reported initial values. deviation of the test scores affected its reliability. For the BSMAS, only
According to the evaluation criteria for internal consistency the standard deviation of the test scores and sample size affected its
reliability; 0.9 indicates good internal consistency reliability, above 0.8 internal consistency reliability. The year that the study was published,
is ideal, above 0.7 is recommended for modification, and below 0.7 mean age of the subjects, standard deviation of the subjects’ age,
should be reworked (Sijtsma, 2009). This study’s results show that both proportion of women in the sample, and total test score had no impact
BFAS and BSMAS have an estimated reliability of more than 0.8, on internal consistency. For the BFAS, the mean and standard
which is ideal. This finding is consistent with the initial reliability deviation of the test scores independently explained about 10.06 and
values of the two measures, indicating that this tool is reliable for 36.7% of the variance, respectively, while the two together explained
measuring an individual’s social media addiction. Moreover, there was about 66.57% of the variance. For the BSMAS, the standard deviation
a significantly high heterogeneity between the studies using the two of test scores and sample size explained approximately 13.54 and
scales. This indicates that there are significant differences in research 10.22% of the variance, respectively, but both were not significant
tools across a wide range of populations, samples, age groups, together. The effects of test scores and standard deviations on the
countries, regions, and publication years. Despite this, the average reliability estimates have also been reported in other reliability
estimation reliability of the two tools still reached a high level, generalizability studies (Blázquez-Rincón et al., 2022; Liang et al.,
indicating the stability of the measurement results of the two tools 2021; López-Pina et al., 2015). Consistent with classical test theory,

Frontiers in Psychology 07 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

which posits that greater variation in observation scores enhances should be considered. Future research should aim to validate these
reliability, our findings align with previous studies showing that the findings in diverse languages and contexts, and examine additional
variability in test scores influences reliability estimates (Vacha- factors that may impact the reliability of these tools.
Haase, 1998).
The present study identified reliability-induction issues associated
with both the BFAS and BSMAS. This phenomenon can be attributed, Data availability statement
in part, to a misunderstanding regarding the nature of reliability—
specifically, whether it pertains to the measurement instrument itself or The original contributions presented in the study are included in
the outcomes derived from the testing process. Our findings underscore the article/supplementary material, further inquiries can be directed
a prevalent misconception that reliability is an intrinsic quality of the to the corresponding author.
testing tool, rather than a characteristic that is contingent upon specific
testing conditions and sample populations. It is important to note that
the reliability of psychological assessments is not an inherent property Ethics statement
of the instrument; rather, it is a feature of the results obtained from the
test. Consequently, administering the same assessment to different The studies involving humans were approved by Ethics
sample groups will inevitably yield varying reliability estimates due to committee of Chongqing University of Posts and
factors such as differences in research samples, testing environments, Telecommunications approved present study. The studies were
cultural contexts, and linguistic backgrounds. Therefore, it is imperative conducted in accordance with the local legislation and institutional
for researchers to consistently report the reliability of a testing requirements. The participants provided their written informed
instrument as it pertains to their specific study context. consent to participate in this study.
This study had certain limitations. The main limitations were as
follows: Firstly, the research was restricted to English and Chinese
publications, which could potentially limit the broader applicability Author contributions
of the findings. Secondly, there was a notable underrepresentation of
studies from South America and Africa, and most studies lacked J-LM: Conceptualization, Data curation, Formal analysis,
racial demographic data, which might restrict the thoroughness of the Methodology, Software, Writing – original draft. ZJ: Methodology,
analysis. Thirdly, since race was not reported in the majority of Software, Writing – review & editing. CL: Investigation, Supervision,
studies, it was not included as a variable in this analysis. Fourthly, the Validation, Visualization, Writing – original draft, Writing – review &
exploratory model indicated that variations in test scores were the editing.
primary source of error. Nevertheless, a significant portion of the
variations remained unexplained. Future research should consider
including studies from a wider range of languages and regions. It Funding
should also explore the influence of racial and cultural factors, as well
as investigate other possible factors that could moderate reliability. The author(s) declare that financial support was received for
The widespread use of social media has led to a prevalent issue of the research, authorship, and/or publication of this article. The
addiction, making the assessment of social media addiction a hot present study was supported by Humanities & Social Sciences
topic in the field of cyberpsychology. Effectively evaluating social Program of Chongqing Education Committee (Chang Liu,
media addiction is crucial for guiding adolescents to use social media 22SKSZ070; Jianling Ma, 21SKGH064; and ZhengCheng Jin,
responsibly and for intervening in cases of addiction. This paper 23SKGH121) and the 2020 Key Social Sciences Program of
employs the generalizability theory to conduct reliability analyses on Chongqing Education Science 13th-Five-Year-Plan (Jianling MA,
the widely used BFAS scale and its variants, explores the sources of 2020-GX-114).
reliability variation, and further clarifies and confirms that the scale
demonstrates high reliability. It is therefore suitable for a broad
application in assessing social media addiction and can be used in Conflict of interest
clinical settings to identify participants who meet the criteria
for addiction. The authors declare that the research was conducted in the
absence of any commercial or financial relationships that could
be construed as a potential conflict of interest.
5 Conclusion
In summary, this pioneering study offers the first generalized Publisher’s note
assessment of the internal consistency reliability of the BFAS and
BSMAS. Our results demonstrate that both tools have average All claims expressed in this article are solely those of the
reliability estimates exceeding 0.8, confirming their stability and authors and do not necessarily represent those of their affiliated
dependability as evaluation instruments for social media organizations, or those of the publisher, the editors and the
addiction. This robust reliability underscores their suitability for reviewers. Any product that may be evaluated in this article, or
use in a wide range of research and clinical settings. However, the claim that may be made by its manufacturer, is not guaranteed or
study’s limitations, such as language and regional constraints, endorsed by the publisher.

Frontiers in Psychology 08 frontiersin.org

Ma et al. 10.3389/fpsyg.2024.1444039

References
Andreassen, C. S., and Pallesen, S. (2014). Social network site addiction - an overview. Hoşgör, H., Ülker Dörttepe, Z. U., and Memiş, K. (2021). Social media addiction and
Curr. Pharm. Des. 20, 4053–4061. doi: 10.2174/13816128113199990616 work engagement among nurses. Perspect. Psychiatr. Care 57, 1966–1973. doi: 10.1111/
ppc.12774
Andreassen, C. S., Torsheim, T., Brunborg, G. S., and Pallesen, S. (2012). Development
of a Facebook addiction scale. Psychol. Rep. 110, 501–517. doi: 10.2466/02.09.18. Knapp, G., and Hartung, J. (2003). Improved tests for a random effects meta-
PR0.110.2.501-517 regression with a single covariate. Stat. Med. 22, 2693–2710. doi: 10.1002/sim.1482
Błachnio, A., Przepiorka, A., Senol-Durak, E., Durak, M., and Sherstyuk, L. (2017). Liang, J. H., Shou, Y. Y., Wang, M. C., Deng, J. X., and Luo, J. (2021). Alabama
The role of personality traits in Facebook and internet addictions: a study on Polish, parenting questionnaire-9: a reliability generalization meta-analysis. Psychol. Assess. 33,
Turkish, and Ukrainian samples. Comput. Hum. Behav. 68, 269–275. doi: 10.1016/j. 940–951. doi: 10.1037/pas0001031
chb.2016.11.037
López-Pina, J. A., Sánchez-Meca, J., López-López, J. A., Marín-Martínez, F.,
Blázquez-Rincón, D., Durán, J. I., and Botella, J. (2022). The fear of COVID-19 scale: Núñez-Núñez, R. M., Rosa-Alcázar, A. I., et al. (2015). The Yale-Brown obsessive
a reliability generalization meta-analysis. Assessment 29, 940–948. doi: compulsive scale: a reliability generalization meta-analysis [Meta-analysis; research
10.1177/1073191121994164 support, non-U.S. Gov't]. Assessment 22, 619–628. doi: 10.1177/1073191114551954
Brailovskaia, J., Margraf, J., and Teismann, T. (2023). Repetitive negative thinking Mamun, M. A. A., and Griffiths, M. D. (2019). The association between Facebook
mediates the relationship between addictive Facebook use and suicide-related outcomes: addiction and depression: a pilot survey study among Bangladeshi students. Psychiatry
a longitudinal study. Curr. Psychol. 42, 6791–6799. doi: 10.1007/s12144-021-02025-7 Res. 271, 628–633. doi: 10.1016/j.psychres.2018.12.039
Brailovskaia, J., Rohmann, E., Bierhoff, H. W., Schillack, H., and Margraf, J. (2019). Marino, C., Gini, G., Vieno, A., and Spada, M. M. (2018). The associations between
The relationship between daily stress, social support and Facebook addiction disorder. problematic Facebook use, psychological distress and well-being among adolescents and
Psychiatry Res. 276, 167–174. doi: 10.1016/j.psychres.2019.05.014 young adults: a systematic review and meta-analysis. J. Affect. Disord. 226, 274–281. doi:
10.1016/j.jad.2017.10.007
Brailovskaia, J., Schillack, H., and Margraf, J. (2018). Facebook addiction disorder in
Germany. Cyberpsychol. Behav. Soc. Netw. 21, 450–456. doi: 10.1089/cyber.2018.0140 Rodriguez, M. C., and Maeda, Y. (2006). Meta-analysis of coefficient alpha. Psychol.
Methods 11, 306–322. doi: 10.1037/1082-989X.11.3.306
Brown, R. I. F. (1993). “Some contributions of the study of gambling to the study of
other addictions” in Gambling behavior and problem gambling. eds. W. R. Eadington Sánchez-Meca, J., Marín-Martínez, F., López-López, J. A., Núñez-Núñez, R. M.,
and J. Cornelius (Reno, NV: Univer. of Nevada Press), 341–372. Rubio-Aparicio, M., López-García, J. J., et al. (2021). Improving the reporting quality of
reliability generalization meta-analyses: the REGEMA checklist. Res. Synth. Methods 12,
Chung, K. L., Morshidi, I., Yoong, L. C., and Thian, K. N. (2019). The role of the dark 516–536. doi: 10.1002/jrsm.1487
tetrad and impulsivity in social media addiction: findings from Malaysia. Personal.
Individ. Differ. 143, 62–67. doi: 10.1016/j.paid.2019.02.016 Satici, S. A. (2019). Facebook addiction and subjective well-being: a study of the
mediating role of shyness and loneliness. Int. J. Ment. Heal. Addict. 17, 41–55. doi:
da Veiga, G. F., Sotero, L., Pontes, H. M., Cunha, D., Portugal, A., and Relvas, A. P. 10.1007/s11469-017-9862-8
(2019). Emerging adults and Facebook use: the validation of the Bergen Facebook
addiction scale (BFAS). Int. J. Ment. Heal. Addict. 17, 279–294. doi: 10.1007/ Schou Andreassen, C. S., Billieux, J., Griffiths, M. D., Kuss, D. J., Demetrovics, Z.,
s11469-018-0018-2 Mazzoni, E., et al. (2016). The relationship between addictive use of social media and
video games and symptoms of psychiatric disorders: a large-scale cross-sectional study.
Delaney, D., Stein, L. A. R., and Gruber, R. (2018). Facebook addiction and impulsive Psychol. Addict. Behav. 30, 252–262. doi: 10.1037/adb0000160
decision-making. Addict. Res. Theory 26, 478–486. doi: 10.1080/16066359.2017.1406482
Seabrook, E. M., Kern, M. L., and Rickard, N. S. (2016). Social networking sites,
Durlak, J. A., and Lipsey, M. W. (1991). A practitioner's guide to meta-analysis. Am. depression, and anxiety: a systematic review. JMIR Mental Health 3:e50. doi: 10.2196/
J. Community Psychol. 19, 291–332. doi: 10.1007/BF00938026 mental.5842
Duval, S., and Tweedie, R. (2000). Trim and fill: a simple funnel-plot-based method Sijtsma, K. (2009). On the use, the misuse, and the very limited usefulness of
of testing and adjusting for publication bias in meta-analysis. Biometrics 56, 455–463. Cronbach's alpha. Psychometrika 74, 107–120. doi: 10.1007/s11336-008-9101-0
doi: 10.1111/j.0006-341X.2000.00455.x
Soraci, P., Ferrari, A., Barberis, N., Luvarà, G., Urso, A., Del Fante, E., et al. (2023).
Elphinston, R. A., Gullo, M. J., and Connor, J. P. (2022). Validation of the Facebook Psychometric analysis and validation of the Italian Bergen Facebook addiction scale. Int.
addiction questionnaire. Personal. Individ. Differ. 195:111619. doi: 10.1016/j. J. Ment. Heal. Addict. 21, 451–467. doi: 10.1007/s11469-020-00346-5
paid.2022.111619
Thompson, B., and Vacha-Haase, T. (2000). Psychometrics is datametrics: the test is
Griffiths, M. (1996). Nicotine, tobacco and addiction. Nature 384:18. doi: not reliable. Educ. Psychol. Meas. 60, 174–195. doi: 10.1177/0013164400602002
10.1038/384018a0
Vacha-Haase, T. (1998). Reliability generalization: exploring variance in measurement
Henson, R. K., and Thompson, B. (2002). Characterizing measurement error in scores error affecting score reliability across studies. Educ. Psychol. Meas. 58, 6–20. doi:
across studies: some recommendations for conducting “reliability generalization” 10.1177/0013164498058001002
studies. Meas. Eval. Couns. Dev. 35, 113–127. doi: 10.1080/07481756.2002.12069054
Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. J.
Higgins, J. T., Thompson, S. G., Deeks, J. J., and Altman, D. G. (2003). Measuring Stat. Softw. 36, 1–48. doi: 10.18637/jss.v036.i03
inconsistency in meta-analyses. Br. Med. J. 327, 557–560. doi: 10.1136/bmj.327.7414.557
Yam, C. W., Pakpour, A. H., Griffiths, M. D., Yau, W. Y., Lo, C.-L. M., Ng, J. M. T., et al.
Ho, T. T. Q. (2021). Facebook addiction and depression: loneliness as a moderator and (2019). Psychometric testing of three Chinese online-related addictive behavior
poor sleep quality as a mediator. Telematics Inform. 61:101617. doi: 10.1016/j. instruments among Hong Kong university students. Psychiatry Q. 90, 117–128. doi:
tele.2021.101617 10.1007/s11126-018-9610-7

Frontiers in Psychology 09 frontiersin.org