Table 3.

Summary of multiple linear regression analyses. The best fits were obtained for data from the Switch (models 1 and 2) and Nogo tasks (model 2). None of the 3 models fitted data from the Oddball task.

Regression models and predictors	Overall R2 (adjusted R2), F statistic, degrees of freedom (regression, residual), and partial standardized beta (β)
	Best fit, single task	Best fit, across tasks
Model 1: Switch costs	R2 = 0.52 (0.41), F3,13 = 4.7*	-
Cue-locked early P3†	β = –0.26 ns	-
Cue-locked late P3†	β = –0.39* ns	-
1st target P3†	β = –0.67** ns	-
Model 2: Restart costs	R2 = 0.68 (0.60), F3,12 = 8.4**	R2 = 0.27 (0.19), F3,29 = 3.5*
Cue-locked early P3	β = 1.10**	β = 0.46*
Cue-locked late P3	β = 0.02 ns	β = 0.10 ns
1st target P3	β = –0.44 ns	β = –0.69 ns
Model 3: Mean RTs	R2 = 0.58 (0.48), F3,13 = 5.9*	R2 = 0.37 (0.31), F3,29 = 5.8***
Cue-locked early P3	β = –0.17 ns	β = 0.30 ns
Cue-locked late P3	β = 0.04 ns	β = 0.02 ns
1st target P3	β = –0.67**	β = –0.77***

(ns, non significant).

*P < 0.05.

**P < 0.01.

***P < 0.001.

†Switch minus repeat trial differences in P3 amplitude were used instead of mean P3 amplitudes.