Concrete manipulatives and dynamic geometry software (DGS) are both commonly used for geometry education in elementary and middle schools. This study sets out to understand which of the two approaches was better in improving the conceptual understanding of quadrilaterals for fifth graders. A pre-/post-test design was conducted in which the same topic was taught by the same teacher to three groups of students, with each group receiving a different type of instruction, namely concrete manipulative based, DGS based and paper-and-pencil based. The initial levels of the students were almost equal as determined by pre-tests. Quantitative analysis of post-test scores revealed that the DGS based approach contributed most to students' geometrical understanding, followed by the concrete manipulative based approach. Interviews and qualitative analysis were also conducted to gain deeper insights into students' reasoning and the effect of different instruction types. The interview results revealed that more students in the DGS group could overcome the misconceptions and difficulties than the students in other groups.