²(Hari et al., 2010; Botvinick, 2012); in fact, RPE’s seem to be coded even in the cerebellum (Heffley and Hull, 2019; Kawato et al., 2021). Using RPE instead of reward loss simplifies the situation to some extent, since RPE considers the total future reward and is thus less dependent on the definition of the time scale, as explained in footnote 21 in Chapter 5.