Last year BJOG set up facilities for clinical researchers to register their randomised controlled trial protocols for electronic publication early in the research process. The idea was to encourage good research practice and to make it more likely that the published results reflected the truth. We are now offering the same facility for studies of test accuracy. We believe that we are the first journal to offer this facility for primary research into test accuracy. Studies of test accuracy can go wrong in all sorts of ways. Authors may fail to define the population under investigation, the test, the cutoff values or their reference standard before they start data collection. Some patients may miss some tests, undergo different ones or not get the reference standard. Finally, the results may not be analysed and presented in ways that clinicians can understand and act on. If any of these things go wrong, or get changed halfway through, it makes it almost impossible to interpret the results. The requirements for reporting test accuracy studies are widely agreed but they are more likely to be met if they are considered and published before starting. Protocol registration makes the whole research process more transparent. If sample sizes, cutoff values or reference standard definitions are altered at any stage, referees can see what has happened and judge whether it may have affected the results. If protocols never get published because editors are biased towards selecting papers with positive results, this will also be clear. To register a test accuracy protocol (PT), visit our Web site at http://bjog.allentrack.net and follow instructions. On pages 847–848, and on the Web site, we publish suggested headings for such protocols which authors are free to follow, although we do not insist on this. Protocols will be sent to referees who will use these headings for evaluation. Accepted protocols will be published electronically on BJOG online and a summary printed in the journal. We hope that the main study publication would be submitted first to BJOG and we promise to send it to referees. If their reports are satisfactory, we would normally agree to publish the study, although we cannot guarantee this at the registration stage.