Predicting Software Comprehensibility.
GENERAL ELECTRIC CO ARLINGTON VA
Pagination or Media Count:
This report describes the first experiment in a program of research designed to identify characteristics of computer software which are related to its psychological complexity. This experiment evaluated the effect of three independent variables mnemonic variable names, complexity of control flow, and general type of program on a programmers understanding of a computer program. The contributions of several variables, including Halsteads software science metric and McCabes complexity metric, to the prediction of program understanding were also evaluated. In a pilot study significant results were achieved with materials and procedures employed here. Thirty-six experienced programmers were instructed to study a computer program for 20 minutes, and were then given 25 minutes to reconstruct a functionally equivalent one. Performance was measured by the percentage of functionally correct statements recalled. Results indicated that complexity of control flow affected program understanding, while no relationship was found for mnemonic variable names and general program type. The metrics of both Halstead and McCabe were related to program understanding when differences between subjects and specific programs were taken into consideration.
- Computer Programming and Software
- Human Factors Engineering and Man Machine Systems