Factors Affecting Programmer Performance in a Debugging Task.
Abstract
This report is the third in a series investigating characteristics of software which are related to its psychological complexity. Three independent variables, length of program, complexity of control flow, and type of error, were evaluated for three different Fortran programs in a debugging task. Fifty-four experienced programmers were asked to locate a single bug in each of three programs. Documentation consisted of input files, correct output, and erroneous output. Performance was measured by the time to locate and successfully correct the bug. Small but significant differences in time to locate the bug were related to differences among programs and presentation order. Although there was no main effect for type of bug, there was a large program by error interaction suggesting the existence of context effects. Among measures of software complexity, Halstead's 'E' proved to be the best predictor of performance followed by McCabe's 'v(G)' and the number of lines of code. Number of programming languages known and familiarity with certain programming concepts also predicted performance. As in the previous experiments, experiential factors were better predictors for those participants with three or fewer years experience programming in Fortran. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 19, 1979
- Accession Number
- ADA068196
Entities
People
- Bill Curtis
- Phil Milliman
- Sylvia B. Sheppard
Organizations
- General Electric