← Research ledger
Draft first · register before trial one
Make the test falsifiable.
Compare one controlled preparation variable. Registration freezes this plan and fingerprints it; a draft remains editable and cannot assign trials.
A / Bbalanced random assignment