A preprint by Tom Hardwicke and John Ioannidis. Abstract: Selection pressures for significant results may infuse bias into the research process. We evaluated the implementation of one innovation designed to mitigate this bias, ‘Registered Reports’, where study protocols are peer-reviewed and granted in-principle acceptance (IPA) for publication before the study has been conducted. As of February 2018, 91 journals had adopted Registered Reports and 91 Final Reports had been published. Psychology journals are the principal adopters, but expansion has begun into medicine, social science, and other fields. Among 29 journals that responded to a survey, 334 protocols had been submitted to them, 87 had been granted IPA and 32 Final Reports had been published or were in press as of July 2017. We encountered several sub-optimal implementation practices, including non-availability of IPA protocols, and diverse approaches to protocol registration in the absence of a single central registry. Registered Reports should be iteratively evaluated and improved to ensure maximal benefits.