There is mixed evidence as to whether concrete manipulatives (e.g., toy animals) are better than abstract manipulatives (e.g., counters) for teaching mathematical concepts to children. Concreteness is defined as the amount of extraneous information a manipulative provides, and in this study we aimed to unpick which dimensions of concreteness influence manipulative choice. Researchers, teachers, and parents completed a comparative judgment task comparing images of manipulatives varying in different dimensions of concreteness, selecting which they would choose to teach arithmetic to children. The findings indicated homogeneous, 3-dimensional manipulatives were the most preferred across all groups to teach arithmetic to children, regardless of more extraneous features. This contradicts research recommendations to minimize the use of concrete manipulatives due to their distractive qualities. Instead, it suggests that some concrete features may be preferred in more naturalistic contexts. More research is required to investigate how different dimensions of concreteness influence learning outcomes for children both in artificial research contexts and in practice.