The present study investigates the relative extent to which judgments of people’s behavior are influenced by "truth" (as measured by averaged observer-judgments) and by systematic bias (i.e., perceivers’ preexisting views of target persons). Using data from online questionnaires and laboratory sessions (N = 155), we demonstrate that self- and peer-judgments of people’s actual behavior in specific situations are somewhat accurate but are also affected by what perceivers thought of the targets before observing their behavior. The latter effect comprises a general evaluative component (generally positive or negative views of targets) and a content-specific component (views of targets in terms of specific characteristics, for example, "restrained"). We also found that friends, but not targets themselves, tend to judge targets’ behaviors more positively than unacquainted observers do. The relevance of these findings for person perception in everyday life and in research contexts is discussed.