The Mutation Testing Coupling Hypothesis

The mutation testing coupling hypothesis DeMilloLS78 says if a test suite is good enough to detect all first order mutants, it will detect ``a large percentage'' of higher order mutants (Offutt92, page 6).

Demonstration of the Coupling Hypothesis

triangle.c 1st,2nd Order Comparison Mutants

16383 test suites were generated by selectively including parts of the triangle test suite. Test suites that kill more first order mutants tend to kill more higher order mutants.

Not only does the graph support the coupling hypothesis, it shows that there can be a strong (non-linear) relationship between how effective a test suite at detecting first order mutants and how effective it is against higher order mutants

See also Figure 17 (page 2434) and Table 3 (page 2421) in Efficient Multi Objective Higher Order Mutation Testing with Genetic Programming, W. B. Langdon and M. Harman and Yue Jia, Journal of Systems and Software, 83 (2010) 2416-2430.

W. B. Langdon
CREST
Department of Computer Science, 27 May 2010 (Last update 3 September 2013).