As I understand the paper, the researchers tried both asking programmers to use certain techniques to debug a program with known bugs, then measured the number

As I understand the paper, the researchers tried both asking programmers to use certain techniques to debug a program with known bugs, then measured the number of bugs they caught against the known total. In addition they went to larger companies, checked which bug detection techniques they were using and how well each one detected bugs. I have the gated PDF, email me if you want a copy.