Figure 2.

Computational model and model based results for PAL task. (a) Shows a schematic of the three level HGF model. See main text and methods for full model details. (b) Shows the results of the model comparison. (c) Chinese participants tend to have a higher ω₂ than British participants, but there are no cultural differences in ω₃ (d) While all participants increase their learning rates in response to volatility, there is no cultural difference in updating either α₂ or α₃ when switching from stable to volatile task phases. rw = Rescorla Wagner, sk = Sutton-K1, HGF 2 = two level HGF, HGF 3 = three level HGF. * denotes group differences with significance p < 0.05. Dotted lines show the linear increase in learning rate between stable and volatile task periods for each group.