Our study provides further evidence that emotional states can facilitate complex learning processes; however, it is too early to draw general conclusions. As a limitation of the study, it should be noted that the second research question could not be answered by the means of between-group comparisons, as the objectively given differences in design and usability failed to affect the per- ceived aesthetics and usability (failed manipulation check) and also failed to internally induce the intended emotional states. Nev- ertheless, the applied regression analyses revealed that the per- ceived aesthetics and usability did affect the emotional states of the learners in the expected way. The resulting emotional states did then affect the learning outcomes – albeit in a very subtle way. Here, our research profited from taking up a more differenti- ated perspective on emotional diversity, as we considered positive and negative activation and additionally their valence as dimen- sions of learners’ emotional state. This allowed differentiated state- ments about the effects on learning: positive and negative activation but not valence slightly affected the learning outcomes. Positive activation subtly affected retention and comprehension, and negative activation slightly affected retention and transfer per- formance. The valence of the emotional state seems to play a minor role, as valence did not predict one of the learning outcome mea- sures. Taken together, our more specific hypotheses on the rela- tionship between positive and negative emotional states and the different learning outcomes measures are only partly supported: contrary to our expectations, we found that retention performance was facilitated by more positive and less negative activation. In line with our expectations, the results showed that a more positive emotional state fosters more complex learning goals (comprehen- sion and transfer). To explain these findings, the expected effects might be moderated by other variables. In our study, we focused on emotional factors and learning outcomes excluding cognitive factors such as cognitive load. Previous research, however, sug- gested that positive emotions can increase cognitive load in work- ing memory (e.g., Um et al., 2012). Further investigations should therefore consider these possible moderator variables to analyze their influence on the expected relationship. Bearing in mind that our results showed only weak direct effects of the emotional states on learning, this study still went on to examine the impact of emo- tional aspects on complex learning processes in more detail. In contrast to the existing studies on emotional design (e.g., Plass et al., 2014; Um et al., 2012), we were able to derive more differen- tiated hypotheses for the relation between learners’ emotional states and learning outcomes, as we did not only consider positive emotional states but also investigated the effect of negative emo- tional states. A crucial implication of this study is that learners’ emotional states do not generally promote the learning process; instead, the impact of positive and negative emotional states depends on the requirements and learning goals (e.g., analytical vs. holistic). Since the different empirical findings showed that emotional state has no straightforward effect on any reasonable learning outcomes measure (retention, comprehension, transfer), future studies should also apply a variety of performance measures in order to reveal the multi-level diversity of effects.