Anne Fleur Kortekaas-Rijlaarsdam

109 REINFORCEMENT LEARNING IN ADHD: EFFECTS OF MPH 5 difficult for children in a pilot version of our task. The current findings are in line, however, with studies showing no differences between ADHD and TD groups in improvements in accuracy over the course of the task (Hauser et al., 2014; Luman et al., 2015). With MPH, learning task performance improved significantly, suggesting earlier and more accurate acquisition of stimulus-reward relations with MPH compared to placebo, in line with Frank et al. (2007) and Luman et al. (2015). Concerning reversal learning, our results oppose the findings from the studies by Itami and Uno (2002) and Alsop et al. (2016) but are in line with the behavioral findings from Hauser et al. (2014). We found no difference between ADHD children and controls in reversal learning accuracy, nor did we find an effect of MPH on reversal learning accuracy. However, the number of trials in the reversal phase was limited as all children learned the reversed stimulus-response trials in the minimal number of trials offered (10 trials). Also, we only included the AB and CD pairs in the reversal task as we found conditioning for these pairs to be strongest (i.e. accuracy for the EF pair was significantly lower than for the AB and CD pair). However, perhaps future research should also include the more difficult EF pairs, as it may be more difficult to reverse associations that are more difficult to acquire, resulting in more trials as well as more variation between trials. The studies by Itami and Uno (2002), Alsop et al. (2016) and Hauser et al. (2014) used more trials (30, 60 and 60 trials respectively). Lack of MPH-effects were not due to carry-over effects. There are a few limitations to note. First, the task was perceived as quite difficult by the children, resulting in a smaller sample size for the test phase as not everyone successfully completed the training phase. Second, mean IQ of our ADHD group was significantly lower than the IQ of our TD group. We decided not to correct for this (Dennis et al., 2011) but did a sensitivity analysis including IQ as a covariate: All results remained unchanged, except for the significant effect of group on generalization, which became a trend ( p =.067). Thirdly, in this study we used a very controlled and standardized computer task to measure feedback learning. Although this allowed us to measure the acquisition and generalization of stimulus-reward associations, this limits ecological validity as in daily life learning stimulus-response associations is often much more complicated (e.g. more distraction; more stimulus-response associations need to be learned at once; feedback is less consistent). An important strength of this study is the placebo-controlled crossover design, ensuring complete blinding of conditions. Together these results support the view that children with ADHD show intact instrumental learning, but do find it difficult to keep these associations active and apply