Testing Influences on Human Understanding of Software

Publisher Website
Google Scholar
Abstract
Eight experienced programmers were each given three FORTRAN programs to memorize and reproduce functionally, without notes. Three levels of complexity of control flow and three levels of mnemonic variable names were independently manipulated. The experimental design was an incomplete split-plot factorial where each programmer was given one version of each program and all levels of the two primary independent variables. The participants correctly recalled significantly more statements when the complexity of control flow was reduced. Differences in recall for the three levels of mnemonic variable names were not significant. A further analysis compared the percent of statements correctly recalled to Halstead's E, a measure of the effort required to code a program. The Pearson correlation coefficient was −0.81, over the 24 data points; thus indicating that Halstead's E is a powerful predictor of one's ability to understand a computer program.

This publication has 2 references indexed in Scilit: