Insightful post on Registered Reports by I like the "dangerous experiments" framing. If you really want to put your predictions to the test, are how to do it -- and you'll get published regardless of whether you were right