Background: Notional machines appear to be an essential aspect of computing education, but there are surprisingly few papers that identify strengths and weaknesses of particular notional machine representations and their impact on student learning.

Purpose: This article strives to fill a gap in the notional machine literature by using a randomized controlled trial to compare the effectiveness of different notional machine representations.

Methods: A multinational randomized controlled trial was conducted using different notional machine representations for two hash table algorithms: chaining and open addressing. Students were randomly assigned a video sequence using either 2D or 3D representations. Hypotheses were tested about learning, perceptions of helpfulness, and engagement based upon the extensive research on representations in science and mathematics education as well as on the theory of cognitive load.

Findings: Our analysis revealed minimal differential effect of 2D vs 3D representational form on student learning, perceptions of helpfulness, and engagement.

Implications: In addition to addressing a lack of research evaluating notional machines, our paper provides an example of how educational research can inform the design and evaluation of notional machines.