Relative beginners to the software development field should read this paper. Regression testing is still quite expensive in terms of resource use. The recent emphasis on test-driven development and agile methods makes it easier to run tests, given automatic execution, but those methods actually increase cost, due to the large number of test cases.

The authors have run extensive experiments to find out the effects of testing techniques at the regression level. They find that adjusting the granularity of test cases, using an ingenious method of obtaining them, is the most effective means of locating bugs and reducing costs.

I was sorry to note the absence of any mention of Markov modeling in the section on test reduction, since it is well known in the testing community for that purpose. On the other hand, the authors may have been working at too high a level of abstraction for named techniques.