Table 1.

Reinforcement learning model fits

Model−LLBICp-R2αβθΦ
Joint355337340.440.53 ± 0.064.37 ± 0.480.46 ± 0.101.43 ± 0.30
Decomposed298131620.530.72 ± 0.154.43 ± 0.600.56 ± 0.110.12 ± 0.94
Random63786560
  • Shown are negative log-likelihood (−LL), BIC, pseudo-R2 (p-R2), and random-effects maximum-likelihood parameter estimates (mean ± SEM across subjects) for the joint and decomposed RL models. The bottom row shows summary statistics for the random (null choice) model in which all joint actions have equal probability.