This study compares direct instruction to two constructivist approaches, namely experimentation and problem-based learning, in fifth-grade science learning in Hong Kong. Constructivist instructional approaches, contrary to direct instruction, stress on students' active role in constructing knowledge in learning. Experimentation and problem-based learning are constructivist approaches commonly adopted in science classes. Research has shown that experimentation enhances cognitive and affective learning outcomes, but mixed evidence has been found in similar comparisons between problem-based learning and direct instruction. We recruited 380 fifth-grade students, each of whom participated in a single science lesson that involved one of the three distinct instructional approaches. Results showed that students in the direct instruction group outperformed both the experimentation group and the problem-based learning group in immediate learning outcome as measured by a short test, while the three groups did not differ in learning interest. Immediate learning outcome and learning interest were weakly correlated only in the problem-based learning group. Our findings may be explained by students' unaccustomedness to constructivist approaches arising from the unique cultural-educational environment in Hong Kong, where direct instruction is the norm and the curriculum prioritises memorisation and understanding of facts and concepts over discovery processes in learning. This inconsistency with the existing literature highlights the importance of considering cultural factors when evaluating the effectiveness of instructional approaches in primary school science education.