Table 2.

Effect size estimates when different potential moderators were considered.

Moderator	k	Cohen’s dz	95% CI	p	$I^2$	p subgroup
Analysis method
JND 84%	3	1.04	0.2–1.87	0.03	29%
JND 75%	8	1.23	0.9–1.56	< 0.001	0%
Slope	7	1.5	1.03–1.98	< 0.001	35%
Accuracy	19	1.53	1.33–1.73	< 0.001	13%
A (flip model)	2	1.89	−0.31 – 4.1	0.06	0%
Total	39					0.007
Visual input
Open	18	1.5	1.28–1.72	< 0.001	35%
Closed	14	1.39	1.17–1.61	< 0.001	0%
Total	32					0.46
Response modality
Hands	28	1.36	1.2–1.52	< 0.001	14%
Feet external	7	1.94	1.55–2.33	< 0.001	0%
Feet anatomical	3	1.27	0.27–2.27	0.03	40%
Verbal	1	1.8	0.47–3.13	0.007	-
Total	39					0.009
Task instruction on speed
Unspeeded	17	1.4	1.15–1.6	< 0.001	1%
Time-restricted	19	1.51	1.31–1.72	< 0.001	31%
Total	36					0.49

k, number of studies; CI, confidence interval. p-values reported for each moderator characteristic indicate whether the crossing effect is significant when including only experiments with that characteristic. p-values for the “Total” tests show whether the moderator has a significant overall effect.