Table 1. Correlations between model features and task performance.

Model feature	VG		SC		NR
	r	p	r	p	r	p
Transition value	0.30	0.39	0.86*•	0.001	0.27	0.45
	[-0.27, 0.80]		[0.68, 0.95]		[-0.52, 0.69]
Functional effect (global brain)	-0.04	0.91	0.39	0.26	0.23	0.52
	[-0.52, 0.44]		[-0.27, 0.80]		[-0.31, 0.72]
Functional effect (task circuit)	0.42	0.22	0.73*	0.017	0.74*•	0.016
	[0.12, 0.82]		[0.01, 0.90]		[0.20, 0.94]
Functional effect (outside the task circuit)	-0.06	0.87	0.38	0.29	0.18	0.62
	[-0.54, 0.43]		[-0.26, 0.79]		[-0.35, 0.69]

The variables r and p denote the Pearson correlation coefficient and associated p-value, respectively. The 90% confidence interval for r is reported below each correlation. Here a * denotes that the observed correlation is significant under FDR correction for multiple comparisons across tasks (for p < 0.05) and a ^• denotes a significant correlation across the two scales of the brain parcellation studied in this paper (S1 Table). VG = verb generation, SC = sentence completion, and NR = number reading.