³⁸(Hari et al., 2010; Botvinick, 2012). 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 20 in Chapter 5.