Table 1.

Predictive accuracy for choices involving new items, within subject.

		Predicted probability of the more likely item			LPM
Subject	Success rate	Mean	Std dev	p-value for bias	Slope	Std err. of slope
11	0.66***	0.663	0.138	0.966	0.966	0.485
12	0.52	0.702	0.156	0.013**	0.550	0.459
13	0.62**	0.728	0.143	0.090*	1.579	0.434
14	0.66***	0.768	0.133	0.126	0.133	0.517
15	0.58	0.711	0.128	0.074*	0.230	0.559
16	0.52	0.742	0.147	0.005***	-0.075	0.494
17	0.58	0.730	0.136	0.050*	-0.325	0.527
18	0.76***	0.738	0.141	0.726	0.393	0.437
19	0.54	0.733	0.142	0.014**	-0.332	0.508
20	0.58	0.692	0.148	0.114	0.698	0.476
21	0.62**	0.748	0.140	0.070*	0.517	0.499
22	0.72***	0.697	0.130	0.727	0.381	0.501
23	0.68***	0.717	0.144	0.567	0.955	0.454
24	0.70***	0.763	0.118	0.320	1.328	0.544
25	0.44	0.760	0.145	0.000***	-0.420	0.494
26	0.52	0.733	0.141	0.007***	-0.312	0.515
27	0.72***	0.738	0.155	0.761	1.114	0.392
Group Mean	0.613***	0.727	0.140	< 0.001***	0.434	0.488
Std. Dev.	0.089	0.028	0.010		0.615	0.043

Notes: Based on an initial voxel selection threshold of 0.01. Asterisks are used to denote statistical significance only in the columns for “success rate” (difference from uninformed benchmark of 0.50, binomial test for individual rates, 1-sided t-test for group mean rate) and “p-value for bias” (difference between success rate and mean predicted probability, two-sided t-test), as follows:

^*denotes p<0.1;

^**denotes p<0.05;

^***denotes p<0.01. “Success rate” is the frequency with which the item with highest predicted choice probability in each pair was actually chosen; “p-value for bias” refers to the test statistic for the hypothesis that the success rate equals the mean predicted probability, and “LPM” refers to a simple linear probability model relating a success indicator to the predicted probability. “Group Mean” is the mean of within-subject means, and “Std. Dev.” is the standard deviation of within-subject means.