Program maintenance is complex, costly and difficult to manage. Program maintenance has been thought to be simplified through the development of well-structured programs because structured program code is expected to increase programmer's understandings of the program, thereby reducing error rates and increasing programmer productivity.
To test these notions, this research is to investigate empirically the effect of program structure on the error rates. The 21 subjects coded three cycles of the original programs followed by program enhancement for a total of three monitored maintenance assignments. Program size, more than any other factors, was found related to the error rates in the revised program code. Program structure was lso slightly related to the error rates for program modifications. Although a significant statistical evidence was not found, the relationship of program structure to the error rates appears to be a U-shaped curve; that is, low program structure and high structures both relate to increased error rates. The implication is that there may be an optimal size of modules and the optimum level of program structure by which we can minimize errors. The program size, which may represent the overall complexity of the program, was found most highly related to program quality.