The aims of this study were to examine self-efficacy and other motivation variables among high school science students (n = 502); to determine the degree to which each of the four hypothesized sources of self-efficacy makes an independent contribution to students' science self-efficacy beliefs; to examine possible differences between life, physical, and Earth science classes; and to investigate patterns of gender differences that may vary among the fields of science. In Earth science classes, girls earned higher grades and reported stronger science self-efficacy. In life science classes, girls earned higher grades but did not report stronger self-efficacy, and did report higher science anxiety. In physical science, there were no gender differences in grades or self-efficacy, but girls again reported higher levels of science anxiety. For boys across science fields, science self-efficacy significantly predicted course grades and mastery experiences was the only significant predictor of self-efficacy. For girls, self-efficacy was also the strongest predictor of science grade across fields. Mastery experiences significantly predicted self-efficacy in Earth science for girls, but social persuasions, vicarious experiences, and physiological states were better predictors of science self-efficacy in life and physical science classes. Results support (Bandura, A., 1997) hypothesized sources of self-efficacy, previous research findings on self-efficacy in the domain of science, and validate the suggestion made by Lau and Roeser (2002) to disaggregate data by science field. (Contains 5 tables.)