Although, the predictive values of high school achievement and SAT scores were supported by many studies, the predictive values of other variables that were theoretically linked to achievement such as goal orientation were not always supported across studies. One possible reason for this inconsistency could stem from the differences on selfesteem levels of the students. It is possible that the predictive power of the variables for achievement could change based on self-esteem levels in that students with high self-esteem could cope with the failure by lowering the value of the topic, while people with low self-esteem could be opt to use more self-handicapping to protect self-esteem. If so, the predictive power of task value for achievement would be higher for high esteem students while the predictive value of self-handicapping would be higher for low self-esteem students. To test this hypothesis, a total of 387 college students attending to Department of Education in a state University in Turkey were asked to respond a series of test materials. Based on previous studies, a series of predictive variables including task value, intrinsic and extrinsic motivation, test anxiety and control beliefs about learning were entered to regression. According to the results, for students who got above average score on self-esteem scale, the best predictor of academic achievement was task value, while the predictors of academic achievement of students who got below average on selfesteem scale were major and test anxiety. Also, while for both students who had above average and below average scores on self-esteem scale, students’ task value is predicted by intrinsic motivation, control beliefs about learning, major and attitudes toward major, self-handicapping was predictive of only those scored below average and academic achievement was predictive only for those scored above average