Should Explanations of Program Code Use Audio, Text, or Both? A Replication Study

Abstract

Studies in educational psychology suggest that people learn better when visual learning materials are accompanied by audio explanations rather than textual ones. Research on how this modality effect applies to computing education is scarce and inconclusive. We explore whether modality of instruction affects learning from videos that use a series of example programs to explain how variables work in Python. Learners (n=186) were crowdsourced from the internet and randomized in three groups, who received explanations as audio, text, or both, respectively. We did not find significant differences between the groups in near transfer to code-tracing tasks or perceived cognitive load. The result affirms the need to further investigate instructional modalities in programming education. There are a number of theoretical, methodological, and instructional-design factors that may explain these and earlier findings; we trace out future research that could explore those factors. CCS CONCEPTS• Social and professional topics → Computing education.

Algorithm Visualization and the Elusive Modality Effect

“…We found no support for there being a modality effect or verbal redundancy effect when Dijkstra's algorithm is taught using visualizations accompanied by textual and/or auditory explanations. This result is analogous to the null results obtained previously by Morrison [44] and Zavgorodniaia et al [69] on program code accompanied by explanations in different modalities. Compared to findings outside of CER, our result contrasts with the studies that have found evidence of a modality effect but is far from being unique in finding no support for the effect (see, e.g., [24,53]).…”

“…Morrison [44] suggested that a possible explanation for her results was the learning materials' excessive complexity. Zavgorodniaia et al [69] replicated Morrison's study using a less challenging video tutorial and a larger and different cohort of learners. They crowdsourced 186 paid participants with little to no self-reported programming experience and randomly assigned each to Audio, Text, or Both.…”

“…As we directly build on and partially replicate two modality studies in CER [44,69], we will describe those study designs in some detail.…”

“…This evidence to date is fragmentary, however, and an overall picture has not yet formed of how "picture-like" or "language-like" programs are-and how that depends on notations and expertise. Highlighting some of these issues, Zavgorodniaia et al [69] prompted CER to investigate the modality effect on materials that are more unambiguously diagrammatic than program code is. We do that in an algorithm visualization context.…”

“…However, the modality and verbal redundancy effects have not been demonstrated in computing education research (CER). As far as we are aware, two studies have examined whether the effects manifest when explaining program code in text and/or audio [44,69]. Neither found evidence for the effects.…”