Because of the range of methodologies used across studies, we used a random-effects model, and conducted statistical analyses using the Comprehensive Meta-analysis, Version 2 (CMA) program (Borenstein, Hedges, Higgins, & Rothstein, 2005). A random-effects model is deemed appropriate as participant samples and experimental factors across studies cannot be assumed to be functionally equivalent. Thus, it cannot be assumed that all effect sizes will share a common effect. Cohen’s d values are reported here as calculated by the CMA program as a measure of effect size. Cohen’s d’s between .20 and .50 indicate small effects, d’s between .50 and .80 indicate moderate effects, and d’s greater than .80 indicate large effects (Cohen, 1988).

Overall Effects

The overall effects matched those previously reported by Powers and colleagues (2013): a meta-analysis at the level of studies (N = 46) generated a small to medium mean effect size, d = .48, 95% CI [.35, .60], with marginal heterogeneity observed across studies, Q (45) = 58.26, p = .089, I^{2} = 23.78. A meta-analysis at the level of comparisons (N = 265) generated a small mean effect size, d = .30, 95% CI [.23, .37], with significant heterogeneity, Q (264) = 684.80, p = .001, I^{2} = 61.45.

Training game type, analyzed at the level of comparisons, moderated effect sizes, Q (6) = 35.33, p = .001 (see Table 21.2). All game types showed small effects except for Wii games, which showed large effects, and strategy/role-play games, which showed negligible, nonsignificant effects.

Information-processing domain, analyzed at the level of comparisons, moderated effect sizes, Q (3) = 29.93, p = .001 (see Table 21.3). A large effect of training was observed in the domain of motor skills, but this result was based on only 16 comparisons from seven studies. Small effects of training were observed in domains of perceptual processing and spatial imagery; in contrast, negligible effects were found for executive functions.

Table 21.2Summary of Effect Sizes Moderated by Training Game Type at the Level of Comparisons

Training game type

Cohen’s d

95% CI

Z

p-value

(Z)

k

N Q (df) p-value

(Q)

Arcade-Style

.31

[.20, .42]

5.33

.001

59

2514

First-Person Shooter

.23

[.07, .39]

2.78

.005

61

2018

Puzzle (Non-Tetris)

.31

[-.001, .63] 1.95

.051

23

923

Sport/Racing

.36

[.13, .60]

3.02

.003

9

366

Strategy/Role-Play

.06

CO

t—1

LO

О

1.07

.284

35

1334

Tetris

.28

[.13, .44]

3.50

.001

51

1507

Wii

.95

[.66, 1.23]

6.58

.001

20

684

Between-classes

effect

35.33 (6) .001

Table 21.3Summary of Effect Sizes Moderated by Information-Processing Domain at the Level of Comparisons