The evaluative priming paradigm aims to uncover the processes underlying evaluations. For this purpose, this paradigm presents a sequence of two or more stimuli varying on the valence dimension to which participants must provide a response. The “standard” evaluative priming effect is a relative facilitation of the required responses in congruent trials compared to incongruent trials. The following thesis argues that this evaluative priming effect depends on prime-target similarity, with higher similarity between prime and target leading to larger priming effects. (...) Part one of this thesis presents a meta-analysis of existing data, which tests evidence for the impact of similarity on evaluative priming effects. This reanalysis is based on the assumption that positive information is overall more similar to other positive information than negative information is to other negative information. Thus, this analysis compares effects of positive and negative prime-target pairs. The results confirm that positive prime-target pairs elicit stronger priming effects than negative prime-target pairs. This analysis involves a broad sample of stimuli and designs which supports the generalizability of this finding. However, the results are also in line with alternative interpretations attributing the valence asymmetries to other effects caused by valence. The following four experiments manipulate similarity either by selecting prime-target pairs based on pre-ratings or by presenting identical and non-identical prime-target pairs. All four experiments show that similar prime-target pairs create larger priming effects than dissimilar prime-target pairs. These findings have implications for our understanding of the evaluative priming paradigm, the use of evaluative priming as a measure of attitudes, and the conceptualization of the evaluative system. These implications will be discussed. (shrink)