Autonomy in learning is an ability required by each individual to cope all life challenges. A person with higher learning autonomy could study better and is capable of observing, evaluating and managing his study effectively and efficiently, and saving the time to accomplish given tasks. However, students’ learning autonomy (SLA) has in fact not yet well-developed up to senior high school level. Generative learning (GL) is considered able to elicit the development of SLA. This research aims to examine students’ autonomy in learning through the implementation of GL. The type of this research is quasi experiment with pretest and posttest control group design. The research used a set of learning autonomy scales as its instrument. The research result shows that GL could better enhance SLA compared to conventional learning; perceived from a whole: 0.58 > 0.51, perceived from the school category: 0.65 > 0.58 (superior), 0.57 > 0.51 (moderate), 0.51 > 0.44 (low) and students’ early mathematics skill (EMS): 0.74 > 0.69 (high), 0.60 > 0.54 (medium), 0.35 > 0.31 (low). Both control and experimental classes have moderate learning autonomy improvement. Meanwhile, in terms of EMS, those obtaining GL treatment improved moderately, and the control class has a poor increase.