Anne Fleur Kortekaas-Rijlaarsdam

CHAPTER 5 108 D I SCUSS I ON We investigated reinforcement learning in children with ADHD and TD children and tested the effects of MPH on learning task performance, generalization of knowledge and reversal learning. In line with previous research on adults with ADHD (Frank et al., 2007), our study showed impaired generalization of knowledge in children with ADHD compared to TD children, despite intact acquisition of stimulus-response relations. Also in line with Frank et al. (Frank et al., 2007), our results suggest that MPH has the potential to improve this important feature of reinforcement learning as MPH appeared to normalize generalization of the acquired stimulus-reward associations to novel stimulus combinations. However, effect-size was small- to medium and future research is necessary to confirm this potential. Our results extend the thus far limited findings on this topic and implicate generalization of knowledge as being deficient in reinforcement learning in ADHD. An alternative interpretation of these findings is that stimulus- reward associations are more sensitive to extinction in children with ADHD than in TD controls, resulting in impaired performance when re-testing takes place. This latter interpretation supports the Dopamine Transfer Deficit theory (Tripp &Wickens, 2008), predicting more rapid extinction of stimulus-reward associations in children with ADHD compared to TD children, due to the absence of anticipatory dopamine under partial reinforcement. According to Frank et al. (2007) the larger extinction rates in ADHD result from impaired reinforcement learning in combination with impaired working memory gating in children with ADHD. In other words, stimulus-reward associations are less robustly maintained in children with ADHD, especially when distractors are presented that interfere with maintenance. Importantly, when generalization of knowledge (or stimulus-reward extinction, or exploring behavior) is affected in a lab context, these problems are likely to be even larger in daily life when situations are more complex (e.g. higher impact on memory and attention; more different rules apply at the same time). In contrast to our hypothesis and to other studies (Frank et al., 2007; Groen et al., 2008; Luman et al., 2009) initial acquisition of stimulus-reward associations was intact in children with ADHD. Although our task and setup is very similar to that of Frank et al. (2007), we used higher reward probabilities (100/0, 85/15 and 70/30 percent compared to 80/20, 70/30 and 60/40 percent). The lower reward probabilities and thus an increased task difficulty may explain the group differences observed by Frank et al. (2007), although the low probabilities used in the Frank et al. study appeared too