Background and Context. The relationship between spatial skills and computing science success has been demonstrated at multiple institutions. ICER has reacted positively to two theories for why this relationship exists, by both Parkinson & Cutts and Margulieux. However, only limited work has been done to validate these theories, and more confirmatory research about the relationship between spatial skills and module grades in CS is necessary.
Objectives. We wish to validate two dimensions of existing theories for the relationship between spatial skills and CS: does CS learning improve spatial skills, as Parkinson & Cutts propose, and does the relationship with grades predominantly apply to students with less prior programming fluency when they begin their learning, as Margulieux proposes. We also wish to contribute more data to the existing set of correlations between spatial skills and measures of CS success.
Method. We conducted a multi-institutional, multi-national project to capture prior programming experience and module grades in CS at three institutions, as well as conducting spatial skills tests at three points during the academic year. We compare spatial skills results with module grades, we examine changes in spatial skills over a period of CS learning and we explore whether the correlations between spatial skills and module grades apply for students at all levels of prior programming fluency.
Findings. We found that spatial skills correlated with module grades at each institution, spatial skills improved over the first semester of teaching (though not the second semester, and at different rates in different institutions) and students with lower self-reported prior programming fluency exhibited much stronger correlations between spatial skills and grades than students with greater programming fluency.
Implications. This work provides additional evidence that spatial skills are correlated with introductory CS outcomes. It also takes steps towards validating existing theories for the relationship by demonstrating that spatial skills can be trained through CS learning and students with lower levels of prior programming fluency are more likely to rely on spatial skills in their CS learning.