In July 2013, before the submission buttons for the 2014 research excellence framework had even been pressed, the first consultation suggesting that some outputs submitted to future REFs should be in open-access form appeared. The details have been much debated since then, and?now, with the publication of the draft guidance on submission, the full set of rules surrounding OA for the REF?2021 is in place.

Or would it be better to say as many rules as we are going to get? Even with the publication of the draft guidance, there are some concerns that the?brief?is not detailed enough. This nervousness has been increased with potentially stronger OA requirements in the future?as a consequence of Science?·¡³Ü°ù´Ç±è±ð¡¯²õ . But Plan?S won¡¯t affect REF 2021 policy.

My mantra?to those involved in a REF submission is: ¡°Read the guidance.¡± I repeat?this so often that?I should just get it printed on a T-shirt. But what?the material offers is the assessment framework and guidance, not a REF rule book. The guidance gives the general principles for REF submissions, but there is always some room for interpretation.

OA is not the only place where REF managers work on principles. Another is researcher independence. The 2014 REF guidance had one scant paragraph defining independence.?

É«ºÐÖ±²¥

The proposed 2021 guidance has slightly more information, but how could?it?cover every single scenario?in which evidence of independence might be found?

É«ºÐÖ±²¥

The same can be said about impact. There are so many different ways for impact to occur, and the panel have given some examples, but how long would the REF guidance be if they included every scenario?

Would I like more information? Of course! However, I?recognise that, as with impact and independence, publishing life is messy and there will always be some complex situation that we haven¡¯t?encountered?before.?

Also, like impact in 2014, there isn¡¯t any custom and practice for the REF to include open-access rules. As such, we need to focus on the principle, which is encouraging and supporting OA with the aim that outputs should be freely available in a repository, ideally shortly after they have been accepted for publication. There are exceptions to allow for cases where this legitimately could?not be achieved, but these often describe standard circumstances.?

There is also a 5?per cent tolerance band per unit of assessment ¨C allowing some non-compliant research outputs to be submitted. The key principle for me is that we have to provide evidence that we have made our best effort to make an output OA.

É«ºÐÖ±²¥

I do recognise that the audit requirements aren¡¯t available yet?¨C another area of concern, I know?¨C but that shouldn¡¯t stop us from applying the broad principles and encouraging OA where we can. However, the real problem will be when an output wasn¡¯t even added to the repository when published, when no attempt was made to make it OA and when there was no credible reason for this to have occurred. So, on the back of that T-shirt it needs to say: ¡°Paper accepted? Tell the library!¡±

Anna Grey is research strategy and policy manager at the University of York.