The Personal Software Process as a Context for Empirical Studies

Author(s): Claes Wohlin
Venue: TCSE Software Process Newsletter
Date: Spring, 1998



This paper presents an argument of doing more empirical studies using the PSP. While it's general point is the utility of the PSP, it spends the bulk of the paper presenting and discussing an example empirical study. A nuance that can be initially overlooked here is the study is advocating using the PSP to conduct empirical studies, rather than doing an empirical study on PSP.

The study is conducted within a PSP course given by the Department of Communications Systems at Lund University, Sweden from 1996-97. They use the standard PSP, with slight additions in the form of a coding and line counting standard. The course was given in C, regardless of student background. The objectives of the experiment were: to evaluate the difference in fault density based on prior experience of the programming language, and to investigate the relationships between different performance measures.

Given these objectives, and the basics of the PSP, there were ready made metrics for measurement of each students individual efforts. Using these metrics, the study demonstrates its ability to empirically differentiate between quality in terms of faults, productivity, and predictability as a function of previous experience with a given language.