Fig. 8: The validations of the model prediction.
From: Thalamic regulation of reinforcement learning strategies across prefrontal-striatal networks
a The RSA results separated for the MF and MB Switch trials (SWMF and SWMF) for the PFC regions (dmPFC, dlPFC and OFC) in humans. The differences of the representational dissimilarity between Wrong Strategy and Correct Strategy model (upper left element- bottom right elements in Fig. 3a) were calculated. Less differences were found for the SWMF compared to SWMB (p = 2.5 × 10−4, p = 0.02, p = 0.009 for dmPFC, dlPFC and OFC, respectively, paired two-sample t-tests with two tails, n = 32 participants). b A box plot of value difference between correct strategy and wrong strategy before reversal (Steady State, SS) for model-based and model-free blocks (p = 1.14 × 10−26; two-sided rank sum test, MB n = 328 sessions vs. MF n = 172 sessions). c The different behaviors between MB and MF blocks before the reversals for both the human data (p = 0.018, n = 32 participants) and the model (p = 3.4 × 10−7, two-sided rank sum test, MB n = ⌊328/6⌋ pseudosubjects vs. MF n = ⌊172/6⌋ pseudosubjects). Box plots indicate the median (middle line), 25th, and 75th percentile (box), and the maximum and minimum (whiskers), as well as the outlier (red cross). SW Switch State, SS Steady State, MF model-free, MB model-based. (****p < 10−4, ***p < 0.001, **p < 0.01, and *p < 0.05). Source data are provided as a Source Data file.
