Benchmarking Effectiveness for Object-Oriented Unit Testing

We propose a benchmark for object-oriented unit testing, called the behavioural response. This is a normative set of state- and equivalence partition-based test cases. Metrics are then defined to measure the adequacy and effectiveness of a test set (with respect to the benchmark) and the efficiency of the testing method (with respect to the time invested). The metrics are applied to expert manual testing using JUnit, and semi-automated testing using JWalk, testing a standard suite of classes that mimic component evolution.