People constantly form implicit evaluations of various aspects of the environment automatically, rapidly, unintentionally, and sometimes outside of conscious awareness. Despite abundant research interests in implicit evaluations, previous work has primarily used relative measures of good versus bad to assess them. As a result, the question of whether certain attitude objects are associated with implicit evaluations of both good and bad has remained largely unexplored. Recent work has suggested that a significant other (e.g., mom, romantic partner) elicits bivalent implicit evaluations (both good and bad), whereas common attitude objects (e.g., chocolate, exam) elicit univalent implicit evaluations (either good or bad). Building on this work, this dissertation presents research projects that aim to answer two interrelated questions regarding the phenomenon of implicit bivalence, namely, what attitude objects elicit implicit bivalence, and why such implicit bivalence emerges. Chapter 1 examines whether implicit bivalence is specific to significant others, with whom we have had rich interaction experiences, or could be extended to even novel persons we know little to no information about. Across three studies, the results suggest that a novel person may be associated with implicit bivalence, even when perceivers have no other information about the person beyond the name and even when perceivers have learned univalent behavioral information about the person. Findings of Chapter 1 suggest that perceivers create representations of novel others “on-the-fly” by bootstrapping from existing representations and adjusting based on new information. Chapter 2 further suggests that implicit bivalence can be elicited by even non-human minds. Across four studies, social robots perceived to have a mind elicit bivalent implicit evaluations, whereas ordinary toys elicit univalent implicit evaluations. This work supports the social readiness account, which suggests that implicit bivalence may arise when entities are perceived to have a mind, even if there is no direct positive or negative experiences with the entity. Chapter 3 is motivated by testing another viable account for why implicit bivalence emerges—the mixed learning experience account. The mixed learning experience account predicts that attitude objects that are explicitly evaluated as ambivalent—suggesting that they have been associated with both positive and negative learning experiences—would elicit implicit bivalence. Using an idiosyncratic approach to identify attitude objects that are explicitly ambivalent for each participant, we did not find evidence that such explicitly ambivalent attitude objects elicit bivalent-priming. In contrast, they elicited weak priming effects. Result of Chapter 3 therefore did not provide direct support for the mixed learning experience account. Future studies should use explicitly ambivalent attitude objects with greater personal significance and attitudinal strength to further examine the relationship between explicit ambivalence and implicit bivalence. In summary, the three projects included in the dissertation disentangle and simultaneously measure evaluations of good and bad associated with novel persons, non-human minds, and explicitly ambivalent attitude objects. By examining the cognitive structure of mental representations automatically and spontaneously triggered by these different categories of attitude objects, the current work provides implications for long-standing interests in how perceivers quickly and effortlessly make sense of the social environment.