This research examined moderators of naturally occurring self-fulfilling prophecies. The authors assessed whether positive or negative self-fulfilling prophecies were more powerful and whether some targets were more susceptible to self-fulfilling prophecies because of their self-concepts in a particular achievement domain and previous academic records. Participants were 98 teachers and 1,539 students in sixth-grade public school math classes. Results yielded a strong pattern showing that teacher perceptions predicted achievement more strongly for low achievers than for high achievers. Results also yielded a much weaker pattern showing that teacher overestimates predicted achievement more strongly than teacher underestimates. Implications for social perceptual accuracy, self-enhancement theory, and understanding when self-fulfilling prophecies are stronger are discussed.