This dissertation proposes a new approach to understanding the long-standing question regarding the similarity-attraction relationship. Conceptualizing interpersonal similarity as a form of social distance (in which distance is minimal), I draw on recent theorizing and research to suggest that similar individuals should be represented more in terms of their concrete and subordinate features (i.e., low-level construals) and less in terms of their abstract and superordinate features (i.e., high-level construals) compared to dissimilar individuals. Given this association between similarity and mental construal, I argue that attraction to similar compared to dissimilar others should reflect the valence of low-level features more than the valence of high-level features. The results of four experiments supported this prediction. Experiment 1 demonstrated that participants expected that attraction between similar (vs. dissimilar) individuals would be related increasingly more to positivity derived from low-level relative to high-level features. Experiments 2 and 3 showed that participants' own personal liking of a similar target was based relatively more on the positivity of the target's behaviors and secondary help (i.e., low-level construals) and relatively less on the positivity of the target's traits and primary help (i.e., high-level construals) compared to their liking of a dissimilar target. Experiment 4 showed that liking of a significant other was better predicted by high-level positivity when participants reflected on their dissimilarities rather than similarities to this person. Together, these studies support the proposed construal level analysis of the similarity-attraction relationship, showing that the weight given to the valence of subordinate and secondary features of information (i.e., low-level construals) becomes more prominent in liking of others as interpersonal similarity increases. Implications for research on interpersonal and inter-group relations are discussed.