Educational Robotics (ER) has emerged as one of the tools to improve STEM learning in primary education if students are properly instructed. However, there is a lack of studies that guide teachers on which type of instruction should be used for ER in STEM between direct (DI) and indirect instruction (II). As a result, the present study aims to compare the two types of instructions in terms of their effect on learning outcomes, students' perceptions, and students' gender differences. We adopted a quasi-experiment comparative research design involving 100 ninth-grade students (13-14 years old). We collected data through achievement tests and perception questionnaires and analyzed them using Cochran's Q-test, Mann-Whitney U-test, and independent samples t-test. Results show that the group in which II was used performed better than those from the group where DI was used. Also, the results show that girls performed better with DI than boys. Furthermore, students perceived ER as useful for developing collaboration and interest in STEM. Therefore, teachers should be supported in learning how to use II and DI strategically in ER to enhance STEM learning.