Registered reports are increasingly being discussed as a means of helping improve ethical and research standards in science and of reducing publication bias – the tendency for those studies with "interesting", novel results, rather than those which are "just" methodologically sound, to find an outlet and be published.
Preregistration basically involves submitting a description of an empirical research study (including theory, methods, hypotheses, etc.) prior to actually starting the empirical research. For a journal, it is this description which then undergoes peer-review and the paper will normally be accepted (or of course rejected) before the results are obtained. Once the results and analysis are in, then these parts are added to the paper and it can be published (subject of course to a final peer review).
Registered reports are one aspect of the current movement towards open science, which can also involve submission and the subsequent publication of research data, either as electronic supplementary material (ESM) in the journal where the article is published or in other external repositories.