An experimental investigation of the effect of program structure on program understanding

Abstract

A within-subjects experimental design was used to test the effect of two variables on program understanding. The independent variables were complexity of control flow and paragraphing of the source code. Understanding was measured by having the subjects memorize the code for a fixed time and reconstruct the code verbatim. Also some subjects were asked to describe the function of the program after completing their reconstruction. The two groups of subjects for the experiment were students from an introductory programming class and from a graduate class in programming languages. The major findings were that paragraphing of the source had no effect for either group of subjects but that programs with simplified control flow were easier for the computer science students to understand as measured by their ability to reconstruct the programs. The dependent variable, rated accuracy of their description of the programs functions, did not differ as a function of either independent variable. The paper is concluded with a description of the utility of this experimental approach relative to improving the reliability of software and a discussion of the importance of these findings.