Axiom | NAcc | AI | Caud | Put | MPFC | Am | PCC | Thal |
---|---|---|---|---|---|---|---|---|

1.1 + | + | = | + | + | + | + | + | + |

1.2 + | + | = | + | + | + | + | = | = |

1.3 + | + | − | + | + | + | + | + | = |

2.1 − | − | = | − | − | − | = | = | − |

2.2 − | − | + | − | − | + | − | + | = |

2.3 − | − | − | − | − | − | − | − | − |

2.4 − | − | − | − | = | − | = | − | − |

2.5 − | − | − | − | − | − | − | − | − |

2.6 − | − | = | − | − | − | − | − | − |

3 = | = | − | = | − | = | = | = | − |

Testing the three axioms of the axiomatic RPE model on our data requires 10 statistical tests. Wald tests of linear restriction were performed on parameter estimates computed with a baseline of TR 9–10 and an analysis window of TR 26–30 (parameter estimates for the nucleus accumbens and anterior insula are shown in Figure 4) with the sign of all significant tests indicated (

*p*< 0.05). We predicted that RPE signals would satisfy the axioms in the way indicated by the signs in the leftmost column. At*p*< 0.05, the nucleus accumbens and caudate each satisfy all three axioms. The anterior insula and thalamus falsify all three axioms. The amygdala and medial prefrontal cortex each satisfy two axioms and the putamen and posterior cingulate cortex each satisfy one axiom. Axiomatic statistical test 1.1: {+$5, 25% probability of winning $5} − {−$5, 25%}; 1.2: {+$5, 50%} − {−$5, 50%}; 1.3: {+$5, 75%} − {−$5, 75%}; 2.1: {+$5, 50%} − {+$5, 25%}; 2.2: {−$5, 50%} − {−$5, 25%}; 2.3: {+$5, 75%} − {+$5, 50%}; 2.4: {−$5, 75%} − {−$5, 50%}; 2.5: {+$5, 75%} − {+$5, 25%}; 2.6: {−$5, 75%} − {−$5, 25%}; 3: {+$5, 100%} − {−$5, 0%}. NAcc, Nucleus accumbens; AI, anterior insula; Caud, caudate; Put, putamen; MPFC, medial prefrontal cortex; Am, amygdala; PCC, posterior cingulate cortex; Thal, thalamus.