The relation between prior knowledge and learning has been investigated in many studies. However, a recent meta-analysis showed that most of these studies suffered from serious methodological shortcomings, as they failed to account for knowledge growth over time, possible ceiling effects for learners with high prior knowledge, moderating effects of learning environments, and possible confounds between intelligence or traits that refer to intellectual investment and prior knowledge. The present study avoided these limitations. A total of 922 students were tested with progressively more difficult Rasch-scaled knowledge tests at four measurement points from Grades 5 to 7. The relation between prior knowledge and knowledge growth in six learning environments (i.e., mathematics, German, and English in regular and gifted classes) was investigated using multigroup latent growth curve models with and without controlling for students' intelligence and the intellectual investment trait of need for cognition. In regular classes, prior knowledge negatively predicted knowledge gains in all subjects. In gifted classes, prior knowledge negatively predicted knowledge gains only in German and English, and need for cognition positively predicted knowledge gains in mathematics but not in German or English. The results indicate that school instruction is designed to help all students reach the predefined learning goals and pass criterion-oriented tests. As a result, students with low prior knowledge tend to learn more than high prior knowledge students. More research is needed on how these processes are modulated by learning environments and are intertwined with students' intelligence and their intellectual investment.