Fig. 3 | Nature Communications

Fig. 3

From: Prefrontal mechanisms combining rewards and beliefs in human decision-making

Fig. 3

Model MIX fit to human performances. a Proportions of choosing true best bandits (maximising reward frequencies x proposed rewards) following reversals. b Proportions of choosing proposed rewards. In a and b, mean proportions over individuals (±s.e.m.) are shown in the congruent, neutral and incongruent condition for participants (blue), and fitted model MIX (purple) simulations. See Supplementary Fig. 1 for model RL fit. c Parameter estimates of proposed rewards biases in the congruent, neutral and incongruent conditions in model MIX fitting human choices, *p < 0.05 (T-tests). These estimates are similar to true values (±0.13, see Fig. 1b) indicating that according to model MIX, participants adequately inferred state beliefs. d Factorial analysis of choice log-odds over trials sorted according to inferred state beliefs (B1 > B2 vs. B1 < B2), normalised utilities (v1 > v2 vs. v1 < v2) and total proposed rewards (\(V_1^{{\mathrm{pr}}} + V_2^{{\mathrm{pr}}}\), median split: left vs. right panel) for model MIX simulations (purple) and human data (blue). Both model MIX and human data exhibited main effects of beliefs and utilities (all Fs(1,21) > 116.9, ps < 0.00001), no main effects of total proposed rewards (both Fs(1,21) < 1.2, ps > 0.47 and no interactions between these factors (all Fs(1,21) < 3.3, ps > 0.14). See Supplementary Table 2 for model best-fitting parameters. Error bars are s.e.m. over participants (N = 22)

Back to article page