Video game training does not enhance cognitive ability: A comprehensive meta-analytic investigation

Classified as probably not a substantial claim.

The search found 63 studies regarding the effects of video game training on cognitive ability, including 3,286 participants and 359 effect sizes.

“Finally, we found 63 studies regarding the effects of video game training on cognitive ability, including 3,286 participants and 359 effect sizes.”

May combine two claims.

Outcome measure and Skill measure were significant moderators in Meta-Analysis 1 (p = .005 and p < .001, respectively), while Type of video game was marginally significant (p = .059).

“Outcome measure and Skill measure were significant moderators (p = .005and p < .001,respectively).Type of video game was marginally significant (p = .059).”

Probably not a substantial claim: may not pass S3 (load-bearing).

The search found 66 studies reporting correlations between cognitive ability and video game skill, including 8,141 participants and 310 effect sizes.

“We found 66 studies reporting correlations between cognitive ability and video game skill, including 8,141 participants and 310 effect sizes.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The search found 98 studies reporting comparisons between players and nonplayers, including 6,166 participants and 315 effect sizes.

“We found 98 studies reporting comparisons (i.e., quasi-experimental design) between players and non-players, including 6,166 participants and 315 effect sizes.”

May combine two claims.

Trim-and-fill analysis for the main model of Meta-Analysis 1 filled 16 studies left of the mean, yielding an estimated correlation of r = .05, 95% CI [.03; .08].

“The trim-and-fill analysis filled 16 studies left of the mean.The estimated correlation was 𝑟̅ = .05,95% CI [.03; .08].”

Probably not a substantial claim: may not pass S3 (load-bearing).

The degree of heterogeneity between effect sizes in the main model of Meta-Analysis 1 was I2 = 52.19.

“The degree of heterogeneity between effect sizes was I 2 = 52.19.”

May combine two claims.

Viechtbauer and Cheung's analysis detected four influential effect sizes in the main model of Meta-Analysis 1; excluding them yielded r = .06, 95% CI [.04; .08], k = 306, p < .001, I2 = 41.08.

“Finally, Viechtbauer and Cheung's (2010) analysis detected four influential effect sizes.The overall correlation without these effect sizes was 𝑟̅ = .06,95% CI [.04; .08],k = 306, p < .001,I 2 = 41.08.”

The quote only partly states this claim.

In the sub-model using frequency of video game playing as the skill measure, the overall correlation was r = .03, 95% CI [.00; .05], k = 156, p = .024.

“The random-effects meta-analytic overall effect size was 𝑟̅ = .03,95% CI [.00; .05],k = 156, p = .024.”

May combine two claims.

In Meta-Analysis 1, overall correlations for five outcome measures were near zero in four measures, with a small correlation of r = .18 for spatial ability.

“The results provided near-zero correlations in four measures and a small correlation (𝑟̅ = .18)in spatial ability.”

May combine two claims.

In Meta-Analysis 2, overall effect sizes for the five outcome measures were positive, ranging from 0.19 to 0.41.

“The results showed positive effect sizes (range 0.19 to 0.41) in all the measures.”

May combine two claims.

The overall effect sizes of the main model and sub-models showed significantly positive effect sizes, indicating that video game players outperformed non-players in all five broad measures of cognitive ability.

“The overall effect sizes of the main model and sub-models showed significantly positive effect sizes, indicating that video game players outperformed non-players in all the five broad measures of cognitive ability.”

The quote only partly states this claim.

For action video game players vs non-video game players, the random-effects overall effect size was g=-0.12, 95% CI [-0.25, 0.01], k=88, p=.072.

“the random-effects metaanalytic overall effect size was 𝑔̅ = -0.12,95% CI [-0.25; 0.01], k = 88, p = .072.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In sensitivity analysis for RTs, the random-effects overall effect size was g=0.09.

“Regarding RTs, the random-effects meta-analytic overall effect size was 𝑔̅ = 0.09, 95% CI”

May combine two claims.

In the moderator analysis, random allocation and hours of training were not significant moderators (p=.321 and p=.826).

“Random allocation and hours of training were not significant moderators, p = .321and p = .826,respectively.”

May combine two claims.

The RVE analyses provided similar overall correlations in all meta-analytic models, and so did the sensitivity analysis.

“Moreover, the RVE analyses provided similar overall correlations in all the meta-analytic models (for details, see the Supplemental material available online), and so did the sensitivity analysis.”

May combine two claims.

The only exception to the weak-correlation pattern in Meta-Analysis 1 was spatial ability; overall correlations between video game skill and spatial ability were all significant, ranging from .09 to .30.

“The only exception to this pattern of results was spatial ability.The overall correlations between video game skill and spatial ability were all significant with a range of values between .09and .30.”

May combine two claims.

Outcome measure and Age were marginally significant moderators (p=.058 and p=.056).

“Outcome measure and Age were marginally significant, p = .058and p = .056,respectively.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The degree of heterogeneity in Meta-Analysis 2 main model was low, I2 = 33.79.

“The degree of heterogeneity between effect sizes was low, I 2 = 33.79.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The advantage of video game players over non-players was slightly more pronounced in RT-related measures (g=0.46).

“the sensitivity analyses showed that the advantage of video game players over non-players was slightly more pronounced in RT-related measures (𝑔̅ = 0.46).”

May combine two claims.

Neither Outcome measure nor Type of game was a significant moderator in Meta-Analysis 2 (p = .174 and p = .101, respectively).

“Neither Outcome measure, nor Type of game was significant (p = .174and p = .101,respectively).”

May combine two claims.

Viechtbauer and Cheung's analysis detected two influential effect sizes in Meta-Analysis 2 main model; excluding them yielded g = 0.32, 95% CI [0.27; 0.38], k = 313, p < .001, I2 = 30.59.

“Finally, Viechtbauer and Cheung's (2010) analysis detected two influential effect sizes.The overall effect size without these effect sizes was 𝑔̅ = 0.32, 95% CI [0.27; 0.38], k = 313, p < .001,I 2 = 30.59.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The sensitivity analyses did not show any pattern of interest.

“Finally, the sensitivity analyses did not show any pattern of interest.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The RVE analyses provided similar results.

“The RVE analyses provided similar results”

May combine two claims.

Neither action nor non-action video game training exerted any substantial effect on participants' cognitive ability.

“neither action nor non-action video game training exerted any substantial effect on the participants' cognitive ability”

May combine two claims.

The meta-analysis of experimental studies reported null effects for video game training on older adults' cognitive ability and thus does not support the hypothesis that video game training could slow cognitive decline in older adults.

“While the meta-analysis of experimental studies reported null effects for video game training on the older adults' cognitive ability and thus does not support this hypothesis”

May combine two claims.

Trim-and-fill analysis for accuracy-related measures filled 16 studies to the left of the mean and yielded an estimated effect size of g = -0.13, 95% CI [-0.26, 0.00].

“Trim-and-fill analysis filled 16 studies left of the mean.The estimated effect size was 𝑔̅ = -0.13,95% CI [-0.26; 0.00].”

The quote only partly states this claim.

In the sub-model comprising all effect sizes for older video game players, the random-effects pooled effect size was g = -0.08, 95% CI [-0.21, 0.04], k = 92, p = .184.

“In the sub-model comprising all effect sizes, the random-effects meta-analytic overall effect size was 𝑔̅ = -0.08,95% CI [-0.21; 0.04], k = 92, p = .184.”

May combine two claims. The quote only partly states this claim.

In the meta-analysis with 102 effect sizes, the unadjusted pooled effect size was g = 0.23, with a 95% confidence interval of 0.15 to 0.31 and p < .001.

“[0.15; 0.31], k = 102, p < .001.”

• Also stated in: § Conclusion · hedged

  “Thus, the publication bias analyses once again suggested that the unadjusted overall effect size (𝑔̅ = 0.23) was an overestimation.”

Probably not a substantial claim: may not pass S3 (load-bearing).

For those RT-related measures, heterogeneity was I2 = 23.10.

“The degree of heterogeneity was I 2 = 23.10.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In that meta-analysis, the between-effect-size heterogeneity was I2 = 33.15.

“The degree of heterogeneity between effect sizes was I 2 = 33.15.”

May combine two claims.

Trim-and-fill analysis for RT-related measures filled 16 studies to the left of the mean and yielded an estimated effect size of g = 0.34, 95% CI [0.22, 0.45].

“The trim-and-fill analysis filled 16 studies left of the mean.The estimated effect size was 𝑔̅ = 0.34, 95% CI [0.22; 0.45].”

May combine two claims.

Selection model estimates for RT-related measures were g = 0.40, 0.34, 0.42, and 0.36 for moderate one-tailed, severe one-tailed, moderate two-tailed, and severe two-tailed selection, respectively.

“The estimates of the selection model analysis were 𝑔̅ = 0.40, 𝑔̅ = 0.34, 𝑔̅ = 0.42, and 𝑔̅ = 0.36 for moderate one-tailed selection, severe one-tailed selection, moderate two-tailed selection, and severe two-tailed selection, respectively.”

May combine two claims. The quote only partly states this claim.

For RT-related measures in the meta-analysis with 44 effect sizes, the pooled effect size was g = 0.09, 95% CI [-0.05, 0.24], p = .200.

“0 [- 0 .05; 0.24], k = 44, p = .200.”

• Also stated in: § Conclusion · plain

  “Thus, considering only RT-related measures did not provide substantially different results compared to the main model (𝑔̅ = 0.09 and 𝑔̅ = 0.07, respectively).”

May combine two claims.

Excluding those five influential effect sizes, the pooled effect size was not different from the unadjusted effect size, g = 0.10, 95% CI [0.05, 0.15], k = 234, p < .001, I2 = .00.

“The overall effect size without these effect sizes was not different from the unadjusted effect size, 𝑔̅ = 0.10, 95% CI [0.05; 0.15], k = 234, p < .001,I 2 = .00.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Selection model estimates for that sub-model were g = -0.18, -0.32, -0.09, and -0.07 for moderate one-tailed, severe one-tailed, moderate two-tailed, and severe two-tailed selection, respectively.

“The estimates of the selection model analysis were 𝑔̅ = -0.18,𝑔̅ = -0.32,𝑔̅ = -0.09,and 𝑔̅ = -0.07 for moderate one-tailed selection, severe one-tailed selection, moderate two-tailed selection, and severe two-tailed selection, respectively.”

The quote only partly states this claim.

One influential effect size was detected for accuracy-related measures.

“One influential effect size was detected.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). The quote only partly states this claim.

For older video game players, the overall effect size for intelligence/reasoning-related measures was significantly negative (g = -0.63).

“The overall effect size referring to intelligence/reasoning-related measures was significantly negative (𝑔̅ = -0.63).”

Probably not a substantial claim: may not pass S3 (load-bearing).

In that sub-model, heterogeneity between effect sizes was I2 = 48.90.

“The degree of heterogeneity between effect sizes was I 2 = 48.90.”

Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing), S4 (based). The quote only partly states this claim.

For older video game players, publication bias analyses estimated values similar to the unadjusted effect sizes.

“The publication bias analyses estimated values similar to the unadjusted effect sizes.”

The quote only partly states this claim.

One influential effect size was detected in that sub-model.

“One influential effect size was detected.”

May combine two claims.

Trim-and-fill analysis for that sub-model filled 21 studies to the left of the mean and yielded an estimated pooled effect size of g = -0.28, 95% CI [-0.41, -0.16].

“Trim-and-fill analysis filled 21 studies left of the mean.The estimated overall effect size was 𝑔̅ = -0.28,95% CI [-0.41; -0.16].”

May combine two claims.

Trim-and-fill analysis for those RT-related measures filled ten studies to the right of the mean and yielded an estimated effect size of g = 0.21, 95% CI [0.07, 0.35].

“The trim-and-fill analysis filled ten studies right of the mean.The estimated effect size was 𝑔̅ = 0.21, 95% CI [0.07; 0.35].”

Probably not a substantial claim: may not pass S3 (load-bearing).

For those RT-related measures, heterogeneity was I2 = 7.89.

“The degree of heterogeneity was I 2 = 7.89.”

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

For older video game players, no influential case was detected for the intelligence/reasoning-related measure.

“No influential case was detected.”

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

For accuracy-related measures, heterogeneity was I2 = 46.35.

“The degree of heterogeneity was I 2 = 46.35.”

May combine two claims.

Trim-and-fill analysis in that meta-analysis added 25 studies to the left of the mean and produced an adjusted pooled effect size of g = 0.12, 95% CI [0.04, 0.20].

“Trim-and-fill analysis filled 25 studies left of the mean.The estimated overall effect size was 𝑔̅ = 0.12, 95% CI [0.04; 0.20].”

May combine two claims.

In the meta-analysis with 234 effect sizes, trim-and-fill analysis filled 21 studies to the left of the mean and yielded an estimated pooled effect size of g = 0.04, 95% CI [-0.01, 0.10].

“filled 21 studies left of the mean.The estimated overall effect size was 𝑔̅ = 0.04, 95% CI [-0.01; 0.10].”

May combine two claims.

Excluding that influential effect size, the pooled effect size was g = -0.06, 95% CI [-0.18, 0.05], k = 91, p < .001, I2 = 43.88.

“The overall effect size without these effect sizes was 𝑔̅ = -0.06,95% CI [-0.18; 0.05], k = 91, p < .001,I 2 = 43.88.”

May combine two claims.

Selection model estimates for accuracy-related measures were g = -0.04, -0.14, 0.02, and 0.02 for moderate one-tailed, severe one-tailed, moderate two-tailed, and severe two-tailed selection, respectively.

“The estimates of the selection model analysis were 𝑔̅ = -0.04,𝑔̅ = -0.14, 𝑔̅ = 0.02, and 𝑔̅ = 0.02 for moderate one-tailed selection, severe one-tailed selection, moderate twotailed selection, and severe two-tailed selection, respectively.”

The quote only partly states this claim.

No influential effect sizes were detected for RT-related measures.

“No influential effect sizes were detected.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Five influential effect sizes were detected in that meta-analysis.

“Five influential effect sizes were detected.”

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

For older video game players, four overall effect sizes sorted by outcome measure were small or null.

“Four overall effect sizes were small or null.”

The quote only partly states this claim.

For RT-related measures in the meta-analysis with 71 effect sizes, the pooled effect size was g = 0.46, 95% CI [0.36, 0.56], p < .001.

“0 [ 0 . 36; 0.56], k = 71, p < .001.”

• Also stated in: § Conclusion · hedged

  “The overall effect size referring to RT-related measures thus appeared to be slightly larger than the one of the main model (𝑔̅ = 0.46 and 𝑔̅ = 0.33, respectively).”

Classified as probably not a substantial claim.

There is an inverse relationship between the size of positive treatment effects and the design quality of experiments.

“The key point risen by Simons and colleagues is that there is an inverse relationship between the size of the positive effects of the treatments and the design quality of the experiments.”

• cites Simons et al. (2016)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Working memory capacity strongly correlates with fluid intelligence.

“A classical result in cognitive psychology is that WM capacity strongly correlates with fluid intelligence (Kane, Hambrick, & Conway, 2005).”

• cites Kane, Hambrick, & Conway, 2005Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Action video game players have been found to outperform nonplayers in a variety of attentional and perceptual tasks.

“Action video game players have been found to outperform nonplayers in a variety of attentional and perceptual tasks (Green, Li, & Bavelier, 2010).”

• cites Green, Li, & Bavelier, 2010Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

A systematic review by Simons et al. claimed that there is no convincing evidence that brain-training programs provide benefits beyond the trained skill or task.

“Finally, a systematic review by Simons et al. (2016) has claimed that there is no convincing evidence that brain-training programs provide benefits that go beyond the trained skill or task.”

• cites Simons et al. (2016)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim. May combine two claims.

Powers et al.'s first meta-analysis comparing players with nonplayers reported medium to large effect sizes showing that video game players were superior in visual processing, executive functioning, and spatial imagery.

“The first meta-analysis, comparing players to non-players, reported medium to large effect sizes showing that video-game players were superior to non-players in measures of visual processing, executive functioning, and spatial imagery, among others.”

• cites Powers, Brooks, Aldrich, Palladino, and Alfieri (2013)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Various types of cognitive training, such as working memory training, improve performance only in the trained tasks.

“substantial research into the psychology of expertise and a recent series of meta-analytic reviews have suggested that various types of cognitive training (e.g., working memory training) benefit performance only in the trained tasks.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A recent meta-analysis found no evidence of positive effects of music training on a broad range of cognitive abilities or academic attainment.

“In the same vein, a recent meta-analysis (Sala & Gobet, 2017c) found no evidence of positive effects of music training on a broad range of cognitive abilities (e.g., intelligence, spatial ability, and phonological processing) or academic attainment (e.g., mathematics and literacy).”

• cites Sala & Gobet, 2017cNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Chess training exerts a small effect on academic achievement and cognitive ability.

“While a recent meta-analysis (Sala & Gobet, 2016) has shown that chess training exerts a small effect on academic achievement and cognitive ability, almost all the studies in the field tested the alleged benefits of chess training using a passive control group only (Sala, Foley, & Gobet, 2017).”

• cites Sala & Gobet, 2016Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The first and second authors independently coded each effect size.

“The first and second authors coded each effect size independently.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Wang et al. meta-analyzed 19 studies and found a positive effect of action video game training on healthy adults' cognitive ability of g = 0.57.

“Finally, Wang et al. (2016) meta-analyzed 19 studies and found a positive effect of action video game training on healthy adults' cognitive ability (𝑔̅ = 0.57).”

• cites Wang et al. (2016)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

For each independent sample, effect sizes from the same type of measure and the same test were merged into a single effect size to reduce statistical dependence and improve reliability.

“For each independent sample, those effect sizes referring to the same type of measure (e.g., reaction times) and extracted from the same test (e.g., different stimulus onset asynchronies in the UFOV task) were merged into one effect size.”

• cites Schmidt & Hunter, 2015Not checked yet

  Not checked in the free run.

May combine two claims.

All publication bias estimates for action video games were small (between r = .03 and r = .13) and did not substantially differ from the overall correlation of .11.

“All the estimates of the publication bias analyses were small (between 𝑟̅ = .03and 𝑟̅ = .13)and thus did not substantially differ from the overall correlation in this model (𝑟̅ = .11).”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

A positive correlation between video game skill and cognitive ability may suggest that video game training exerts positive effects on cognitive ability.

“As stated in the introduction, a positive correlation between video game skill and cognitive ability may suggest that video game training exerts positive effects on cognitive ability.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Viechtbauer and Cheung's influential cases analysis was performed in every meta-analytic model to test robustness of the overall results.

“Finally, to evaluate whether some effect sizes had an unusually large influence on the metaanalytic means, we performed Viechtbauer and Cheung's (2010) influential cases analysis in every meta-analytic model.”

• cites Viechtbauer and Cheung's (2010)Not checked yet

  Not checked in the free run.

May combine two claims.

The two publication bias analyses for the video-game-scores sub-model provided a different pattern of results, with all estimated overall correlations small but greater than zero.

“The two publication bias analyses thus provided a different pattern of results.All the estimated overall correlations were small but greater than zero.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Inter-rater agreement for coding effect sizes was Cohen's kappa = .85, 95% CI [.82; .88].

“The Cohen's kappa was κ = .85,95% CI [.82; .88].”

May combine two claims.

Toril et al.'s meta-analysis included several exergaming and brain-training studies; when these are excluded, the number of studies is reduced to 12.

“Toril et al.'s (2014) meta-analysis included several exergaming and brain-training studies (e.g., Ackerman, Kanfer, & Calderwood, 2010;Anguera et al., 2013;Maillot et al., 2012).When these exergaming and brain-training studies are excluded, the number of the studies is reduced to 12.”

• cites Ackerman, Kanfer, & Calderwood, 2010Not checked yet

  Not checked in the free run.

• cites Anguera et al., 2013Not checked yet

  Not checked in the free run.

• cites Maillot et al., 2012Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable).

Using multiple methods to detect publication bias is fundamental for triangulating the true effect size estimate.

“The use of multiple methods for the detection of publication bias is fundamental for triangulating the true effect size estimate (Kepes, Bushman, & Anderson, 2017;Kepes & McDaniel, 2015).”

• cites Kepes, Bushman, & Anderson, 2017Not checked yet

  Not checked in the free run.

• cites Kepes & McDaniel, 2015Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

PET-PEESE estimators were calculated.

“Finally, the PET-PEESE estimators were calculated (Stanley and Doucouliagos, 2014).”

• cites Stanley and Doucouliagos, 2014Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

A systematic search strategy following the PRISMA statement was used to find relevant studies.

“We used a systematic search strategy to find the relevant studies (PRISMA statement; Moher, Liberati, Tetzlaff, & Altman, 2009).”

• cites Moher, Liberati, Tetzlaff, & Altman, 2009Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The absence of correlation or between-group differences does not necessarily imply that there is no causation.

“In addition, it must be noted that, just like correlation and between-group differences do not imply a causal relationship between two variables, the absence of correlation or between-group differences does not necessarily imply that there is no causation.”

May combine two claims.

Toril et al. did not apply an appropriate correction for statistically dependent effect sizes and simply merged dependent effect sizes into a single effect size irrespective of whether they measured the same cognitive ability.

“Finally, no appropriate correction for statistically dependent effect sizes was applied.The dependent effect sizes were simply merged into a single effect size irrespective to whether they measure the same cognitive ability or not.”

May combine two claims.

Partially dependent effect sizes were treated as statistically independent in the main analyses, a limitation that was nearly absent when effect sizes were sorted by type of video game.

“Since Cheung and Chan's ( 2004) method cannot be used for partially dependent samples, we ran our analyses as if these effect sizes were statistically independent.This relatively minor limitation was nearly absent when the effect sizes were sorted by type of video game.”

May combine two claims.

All publication bias estimates for non-action video games were close to zero and did not substantially differ from the unadjusted overall correlation of .07.

“All the estimates of the publication bias analyses were close to zero and did not substantially differ from the unadjusted overall correlation (𝑟̅ = .07).”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Toril et al. reviewed 20 studies on video game training effects on older adults' overall cognitive ability and reported a positive overall effect size of g = 0.37.

“Toril, Reales, and Ballesteros (2014) reviewed 20 studies regarding the effects of video game training on older adults' overall cognitive ability and reported a positive overall effect size (𝑔̅ = 0.37).”

• cites Toril, Reales, and Ballesteros (2014)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Duval and Tweedie's trim-and-fill analysis and Vevea and Woods's selection model analysis were used to investigate whether results were affected by publication bias.

“To investigate whether the results were affected by publication bias, we used Duval andTweedie's (2000) trim-and-fill analysis andVevea andWoods's (2005) selection model analysis.”

• cites Duval andTweedie's (2000)Not checked yet

  Not checked in the free run.

• cites Vevea andWoods's (2005)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Three sensitivity analyses were run including experiments that reported effect sizes for both accuracy and reaction times in the same cognitive tasks, analyzing accuracy and reaction times separately.

“To check for this potentially confounding variable, we ran three sensitivity analyses (one for each meta-analysis) including all the experiments that reported effect sizes for both accuracy and RTs in the same cognitive tasks, and analyzed accuracy and RTs separately.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Both accuracy and reaction time measures were reported to check for possible trade-off effects.

“Several studies reported the participants' performance on both accuracy and reaction times (RTs) in the same tasks.We reported both these measures to check for possible trade-off effects.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Toril et al. did not calculate a quantitative estimation of publication bias.

“Moreover, no quantitative estimation of publication-bias was calculated.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The first and second authors independently coded each effect size in Meta-Analysis 1, with Cohen's kappa = .96, 95% CI [.94; .99].

“The first and second authors coded each effect size independently.The Cohen's kappa was κ = .96,95% CI [.94; .99].”

Probably not a substantial claim: may not pass S3 (load-bearing).

A parallel set of analyses using Robust Variance Estimation was conducted to test whether outcomes were sensitive to the meta-analytic technique used to model dependent effect sizes.

“Furthermore, we carried out a parallel set of analyses using the Robust Variance Estimation method (RVE; Hedges, Tipton, & Johnson, 2010;Tanner-Smith & Tipton, 2014).”

• Also stated in: § Statistical Dependence of the Samples · plain

  “The whole procedure was implemented to test whether the outcomes of the meta-analytic models were sensitive to the type of meta-analytic technique used to model dependent effect sizes.”

• cites Hedges, Tipton, & Johnson, 2010Not checked yet

  Not checked in the free run.

• cites Tanner-Smith & Tipton, 2014Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Before this study, no meta-analysis had assessed the correlation between video game skill and cognitive ability in populations of video game players.

“To the best of our knowledge, no such meta-analysis had been carried out before.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

If video game experience or skill is not correlated with cognitive ability, or if video game players are not better than nonplayers, then it is hard to claim that video game training enhances cognitive ability.

“If video game experience/skill is not correlated with cognitive ability, or video game players are not better than non-video game players, then it is hard to claim that video game training enhances cognitive ability.”

May combine two claims.

Powers and Brooks's investigation did not have enough studies to produce reliable models; the other six categories of video game genre included only three to nine studies.

“Finally, although Powers and Brooks's (2014) investigation was an interesting attempt to understand the impact of different types of video game training in more detail, there were not enough studies to produce reliable models.In factaction video games excludedthe other six categories of videogame genre included only three to nine studies.”

Probably not a substantial claim: may not pass S3 (load-bearing).

For statistically dependent effect sizes from different constructs or tests, Cheung and Chan's correction was applied, reducing their weight by using an adjusted smaller N.

“For those effect sizes that were statistically dependent and referred to different constructs or were extracted from different tests, we applied Cheung and Chan's (2004) correction for statistically dependent samples.”

• cites Cheung and Chan's (2004)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Publication bias estimates for mixed video games did not significantly differ from the overall correlation.

“The estimates of the publication bias analyses thus did not significantly differ from the overall correlation in this model.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

It seems improbable that nonmonotonic curvilinear relationships or Simpson's paradox conditions occur in the case of video game training.

“However, it seems improbable that either or both of these conditions occur in the case of video game training.”

• AssumptionThe authors state this as an assumption: If video game training enhances cognitive function, then prolonged exposure to video games in natural groups should produce clear-cut differences between players and nonplayers and positive correlations between video game skill and cognitive skills.

Probably not a substantial claim: may not pass S3 (load-bearing).

The authors resolved every discrepancy in coding.

“The authors resolved every discrepancy.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The first and second authors independently coded each effect size in Meta-Analysis 2, with Cohen's kappa = .95, 95% CI [.92; .99].

“The first and second authors coded each effect size independently.The Cohen's kappa was κ = .95,95% CI [.92; .99].”

Classified as probably not a substantial claim.

The results do not exclude the possibility that individuals with higher cognitive abilities are more likely to play video games.

“However, the results do not exclude the possibility that individuals with higher cognitive abilities are more likely to play video games.”

Probably not a substantial claim: may not pass S3 (load-bearing).

It is possible that the correlation between cognitive ability and hours of play per week was more affected by measurement error than the correlation between cognitive ability and video game scores.

“Thus, it is possible that the correlation between cognitive ability and hours of play per week was more affected by measurement error than the correlation between cognitive ability and video game scores.”

• AssumptionRests on an assumption: Score tends to be a more reliable measure of video game skill than the weekly frequency of play reported in a questionnaire.

Probably not a substantial claim: may not pass S3 (load-bearing).

A sensitivity analysis was run including only studies meeting Green and Bavelier (2007)'s criteria for action video game players and non-video game players.

“For this reason, we ran a sensitivity analysis including only those studies meeting Green and Bavelier (2007)'s criteria: at least five hours of action video game play per week in the past six months for action video-game players and no action-video game play for non-video game players.”

• cites Green and Bavelier (2007)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

It is possible that cognitive benefits occur regardless of skill, as long as individuals engage in video game playing.

“However, it is possible that such benefits occur regardless of skill, as long as individuals engage in video game playing.”

Classified as probably not a substantial claim.

No direction of causality can be inferred from correlations.

“That said, no direction of causality can be inferred from correlations.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A finding that video game players outperform non-video game players may suggest, yet not necessarily imply, that video game training positively impacts cognitive ability.

“Like in the previous meta-analysis, this outcome may suggest, yet not necessarily imply, that video game training positively impacts cognitive ability.”

May combine two claims.

Overall, the results suggest that video game skill is not related or only weakly related to cognitive ability in general.

“Overall, the results suggest that video game skill is not related or only weakly related to cognitive ability in general.”

Classified as probably not a substantial claim. May combine two claims.

If individuals with higher cognitive abilities are more likely to play video games, no causal relationship in which playing video games leads to superior cognitive abilities needs to be postulated to account for these results.

“If so, no causal relationship in which playing video games leads to superior cognitive abilities needs to be postulated to account for these results.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The convergence of results confirms the reliability of the meta-analytic models.

“This convergence of results confirms the reliability of the meta-analytic models.”

Probably not a substantial claim: may not pass S3 (load-bearing).

To correct effect sizes for upward bias, Cohen's ds were converted into Hedges's gs.

“Finally, to correct the effect sizes for upward bias, we used CMA to convert Cohen's ds into Hedges's gs (Hedges & Olkin, 1985).”

• cites Hedges & Olkin, 1985Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Video game players are usually categorized according to the video game genre they play the most.

“Video game players are usually categorized according to the video game genre they play the most (e.g., Green & Bavelier, 2003).”

• cites Green & Bavelier, 2003Not checked yet

  Not checked in the free run.

May combine two claims.

After undergoing video game training, people may get better at solving cognitive tasks similar to the training task, and yet not show any genuine improvement in cognitive ability.

“after undergoing video game training, people may get better at solving cognitive tasks similar to the training task, and yet not show any genuine improvement in cognitive ability.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Non-video game players are often defined as individuals not playing a specific video game genre.

“Conversely, non-video game players are often defined as individuals not playing a specific video game genre.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Chess masters can recall entire chess positions, even when the material is presented only for a few seconds.

“chess masters can recall entire chess positions, even when the material is presented only for a few seconds (e.g., Gobet & Simon, 2000).”

• cites Gobet & Simon, 2000Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The correlation between chess skill and performance on tests of short-term memory is modest (r=.22; Burgoyne et al., 2016).

“the correlation between chess skill and performance on tests of short-term memory is modest (𝑟̅ = .22;Burgoyne et al., 2016).”

• cites Burgoyne et al., 2016Not checked yet

  Not checked in the free run.

May combine two claims.

Domain-general cognitive abilities are malleable to training, but the benefits, when any, are domain-specific.

“Domain-general cognitive abilities are malleable to training, but the benefits, when any, are domain-specific (Chase & Ericsson, 1982;Gobet, 2016).”

• cites Chase & Ericsson, 1982Not checked yet

  Not checked in the free run.

• cites Gobet, 2016Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing).

Like chess players, video game players may possess exceptional cognitive abilities only with domain-specific material.

“Like chess players, video game players may possess exceptional cognitive abilities only with domain-specific material.”

May combine two claims.

The three meta-analyses offer a consistent picture: weak correlations between skill and cognitive ability, small differences between video game players and non-players, and no or negligible differences between training and control groups.

“The three meta-analyses offer a consistent picture: weak correlations between skill and cognitive ability, small differences between video game players and non-players, and no or negligible differences between the participants who underwent video game training and the participants in the control groups.”

May combine two claims.

Playing video games in general, or some genre in particular, may be associated with specific cognitive abilities predicting, to some extent, a player's skill.

“Playing video games in general, or some genre in particular, may be associated with specific cognitive abilities predicting, to some extent, a player's skill.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Video game players are often engaged in more than one genre, and non-video game players rarely do not play any video game.

“However, video game players are often engaged in more than one genre, and non-video game players rarely do not play any video game.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

The field of video game playing may present characteristics analogous to other domains of expertise, such as chess and music.

“the field of video game playing may present characteristics analogous to other domains of expertise, such as chess and music.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The meta-analysis controlled for potential baseline confounding by including random allocation as a moderator.

“This moderator was included to control for potential confounding effects due to differences at baseline level.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

These findings are in line with substantial research into expertise and cognitive training in domains such as music, chess, WM, and brain training.

“These findings are in line with substantial research into expertise and cognitive training in domains such as music, chess, WM, and brain training.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The pooled effect size for RT-related measures (g = 0.46) was slightly larger than the main model's pooled effect size (g = 0.33).

“The overall effect size referring to RT-related measures thus appeared to be slightly larger than the one of the main model (𝑔̅ = 0.46 and 𝑔̅ = 0.33, respectively).”

Classified as probably not a substantial claim.

The meta-analysis included only studies whose training video game was not purposely designed to improve cognitive ability.

“The training video game was not purposely designed to improve cognitive ability (e.g., Lumosity® brain-training video games).”

Classified as probably not a substantial claim.

The meta-analysis included only studies that had at least one control group.

“The study included at least one control group”

Classified as probably not a substantial claim.

The meta-analysis included only studies whose participants had no or negligible experience with the training video game at the beginning of the experiment.

“The study included participants with no (or negligible) experience, at the beginning of the experiment, in the video game(s) used during training”