The influence of structure on computer program comprehension

Previous empirical studies that have relied on debugging (Lucas & Kaplan, 1974) and memory (Sheppard, Borst, Curtis, & Love, 1978) as measures of computer program comprehension have led to inconsistent results and conclusions regarding the benefits of structure. The present experiment utilizes a new technique for measuring comprehension that is sensitive to contextual factors that help the programmer to anticipate upcoming lines of a program. Subjects in this experiment were told the general purpose of a program written in English and asked to anticipate the subsequent lines. After the subjects entered their anticipated line, the actual line of the previously written program was shown. Subjects used the lines of the program shown as a contextual guide for their next anticipation. The percentage of correctly anticipated lines was greater for structured programs than for unstructured programs. However, the advantage of structure was more pronounced for long, complex programs than for short, relatively straightforward programs.