We employ experimental methods to investigate misperceptions regarding non-selfish behavior, particularly examining the extent to which second-party punishment and rewarding are either underestimated or overestimated. Our study encompasses various games and treatments that span a wide array of decision environments, including different (i) trade-offs (e.g., Pareto vs. non-Pareto dominated, socially efficient vs. inefficient), (ii) motives for non-selfish behavior (e.g., inequity aversion, reciprocity), (iii) costs associated with punishing and rewarding, (iv) incentives for belief accuracy, and (v) the number of allocations in a given game. Our findings reveal a consistent underestimation of rewards at both the aggregate and individual levels. This suggests that individuals harbor pessimistic perceptions or "mental models" regarding human rewarding behavior, regardless of the decision environment and incentives at play. In contrast, beliefs concerning punishment tend to be more accurate, indicating that participants anticipate altruistic punishment within our experimental setup. Additionally, we observe that women, economists, and individuals with stronger religious affiliations tend to underestimate rewards.