Thread by @michaelbolton

1) Heuristic: When X is a noun: "X testing" is "testing focused on X-related risk".

Heuristic: When Y is an adjective or adverb, "Y testing" is "testing in a Y-ish way".

Heuristic: X testing can be done in ways modified or not by Y; and Y testing may be focused on X or not.

2) So: "Performance testing" means "testing focused on risk related to performance". "Usability testing" means "test focused on usability-related risk". "Function testing": "testing focused on risk related to functions". "Unit testing": testing focused on problems in the units.

3) Now: let's look at "regression testing". Regression testing means "testing focused on risk related to regression". (Regression means "going backwards", presumably getting worse in some sense.") *Repetitive* is an adjective, modifying something; not really something in itself.

4) "Regression testing", therefore, is "testing focused on the risk that things have got worse". There are several implications here. One is that "regression testing" *doesn't* necessarily mean "repetitive testing". Another is that a repeated test isn't always a regression test.

5) A repeated test might provide evidence that things have got worse. But a new test can do that too. In fact, a new test can do something a repeated test might not do: a new test might provide evidence that things have been bad all along in ways the repeated tests didn't detect.

6) After a new test suggests that there might be a problem, what will we do? We'll probably repeat the test! We'll do that to try to refine our understanding of the potential problem—or perhaps to conclude that our test result was in error, due to bad procedures or assumptions

7) The point: a repeated test is not necessarily a regression test; a regression test is not necessarily a repeated test. This is important, because *repeated* automated checks, intended to catch regressions, often lose their power to do so; they're overfocused on the familiar.