Program Comprehension Through Sonification

Abstract

Background: Comprehension of computer programs is daunting, thanks in part to clutter in the software developer's visual environment and the need for frequent visual context changes. Non-speech sound has been shown to be useful in understanding the behavior of a program as it is running.
Aims: This thesis explores whether using sound to help understand the static structure of programs is viable and advantageous.
Method: A novel concept for program sonification is introduced. Non-speech sounds indicate characteristics of and relationships among a Java program's classes, interfaces, and methods. A sound mapping is incorporated into a prototype tool consisting of an extension to the Eclipse integrated development environment communicating with the sound engine Csound. Developers examining source code can aurally explore entities outside of the visual context. A rich body of sound techniques provides expanded representational possibilities. Two studies were conducted. In the first, software professionals participated in exploratory sessions to informally validate the sound mapping concept. The second study was a human-subjects experiment to discover whether using the tool and sound mapping improve performance of software comprehension tasks. Twenty-four software professionals and students performed maintenance-oriented tasks on two Java programs with and without sound.
Results: Viability is strong for differentiation and characterization of software entities, less so for identification. The results show no overall advantage of using sound in terms of task duration at a 5% level of significance. The results do, however, suggest that sonification can be advantageous under certain conditions.
Conclusions: The use of sound in program comprehension shows sufficient promise for continued research. Limitations of the present research include restriction to particular types of comprehension tasks, a single sound mapping, a single programming language, and limited training time. Future work includes experiments and case studies employing a wider set of comprehension tasks, sound mappings in domains other than software, and adding navigational capability for use by the visually impaired.