Figure 4.

Posterior medians of the winning models’ group-level mean parameters, per age group. Error bars and shaded areas indicate 95% HDIs of the posterior distributions. This figure was created using RStudio 1.1.463, https://www.rstudio.com. Note ${\text{Q}}_1$ is expressed in cents. For the adult group, ${\text{Q}}_1$ was computed using the posterior median of the initial a parameter $\left({\text{a}}_1\right)$ of the winning Bayesian ideal-observer model, and trial-specific learning rates α₊ and α₋ were computed using the posterior medians of the update-rate parameters (π₊ and π₋) of the winning Bayesian ideal-observer model (Supplemental Text 1). The lines for α₊ and α₋ almost overlap.