This is the fourth in a series of blog posts on the Case for Open Research, this time looking at issues with peer review. The previous three have looked at the mis-measurement problem, the authorship problem and the accuracy of the scientific record. This blog follows on from the last and asks – if peer review is working why are we facing issues like increased retractions and the inability to reproduce considerable proportion of the literature? (Spoiler alert – peer review only works sometimes.)

Again, there is an entire corpus of research behind peer review, this blog post merely scrapes the surface. As a small indicator, there has been a Peer Review Congress held every four years for the past thirty years (see here for an overview). Readers might also be interested in some work I did on this published as The peer review paradox – An Australian case study.

There is a second, related post published with this one today. Last year Cambridge University Press invited a group of researchers to discuss the topic of peer review – the write-up is here.

An explainer

What is peer review? Generally, peer review is the process by which research submitted for publication is overseen by colleagues who have expertise in the same or similar field before publication. Peer review is defined as having several purposes:

• Checking the work for ‘soundness’
• Checking the work for originality and significance
• Determining whether the work ‘fits’ the journal
• Improving the paper

Last year, during peer review week the Royal Society hosted a debate on whether peer review was fit for purpose. The debate found that in principle peer review is seen as a good thing, but the implementation is sometimes concerning. A major concern was the lack of evidence of the effectiveness of the various forms of peer review.

Robert Merton in his seminal 1942 work The Normative Structure of Science described four norms of science*. ‘Organised scepticism’ is the norm that scientific claims should be exposed to critical scrutiny before being accepted.  How this has manifested has changed over the years. Refereeing in its current form, as an activity that symbolises objective judgement of research is a relatively new phenomenon – something that has only taken hold since the 1960s.  Indeed, Nature was still publishing some unrefereed articles until 1973.

(*The other three norms are ‘Universalism’ – that anyone can participate, ‘Communism’ – that there is common ownership of research findings and ‘Disinterestedness’ – that research is done for the common good, not private benefit. These are an interesting framework with which to look at the Open Access debate, but that is another discussion.)

Crediting hidden work

The authorship blog in this series  looked at credit for contribution to a research project, but the academic community contributes to the scholarly ecosystem in many ways.  One of the criticisms of peer review is that it is ‘hidden’ work that researchers do. Most peer review is ‘double blind’ – where the reviewer does not know  the name of the author and the author does not know who is reviewing the work. This makes it very difficult to quantify who is doing this work.  Peer review and journal editing is a huge tranche of unpaid work that academics contributions to research.

One of the issues with peer review is the sheer volume of articles being submitted for publication each year. A 2008 study  ‘Activities, costs and funding flows in the scholarly communications system‘ estimated the global unpaid non-cash cost of peer review as £1.9 billion annually.

There has been some call to try and recognise peer review in some way as part of the academic workflow. In January 2015 a group of over 40 Australian Wiley editors sent an open letter Recognition for peer review and editing in Australia – and beyond?  to their universities, funders, and other research institutions and organisations in Australia, calling for a way to reward the work. In September that year in Australia,  Mark Robertson, publishing director for Wiley Research Asia-Pacific, said “there was a bit of a crisis” with peer reviewing, with new approaches needed to give peer reviewers appropriate recognition and encourage ­institutions to allow staff to put time aside to review.

There are some attempts to do something about this problem. A service called Publons is a way to ‘register’ the peer review a researcher is undertaking. There have also been calls for an ‘R index’ which would give citable recognition to reviewers. The idea is to improve the system by both encouraging more participation and providing higher quality, constructive input, without the need for a loss of anonymity.

Peer review fails

The secret nature of peer review means it is also potentially open to manipulation. An example of problematic practices is peer review fraud. A recurrent theme throughout discussions on peer review at this year’s Researcher 2 Reader conference (see the blog summary here) was that finding and retaining peer reviewers was a challenge that was getting worse. As the process of obtaining willing peer reviewers becomes more challenging, it is not uncommon for the journal to ask the author to nominate possible reviewers.  However  this can lead to peer review ‘fraud’ where the nominated reviewer is not who they are meant to be which means the articles make their way into the literature without actual review.

In August 2015 Springer was forced to retract 64 articles from 10 journals, ‘after editorial checks spotted fake email addresses, and subsequent internal investigations uncovered fabricated peer review reports’.  They concluded the peer review process had been ‘compromised’.

In November 2014, BioMed Central uncovered a scam where they were forced to retract close to 50 papers because of fake peer review issues. This prompted BioMed Central to produce the blog ‘Who reviews the reviewers?’ and Nature writing a story on Publishing: the peer review scam.

In May 2015 Science  retracted a paper because the supporting data was entirely fabricated. The paper got through peer review because it had a big name researcher on it. There is a lengthy (but worthwhile) discussion of the scandal here. The final clue was getting hold of a closed data set  that: ‘wasn’t a publicly accessible dataset, but Kalla had figured out a way to download a copy’. This is why we need open data, by the way …

But is peer review itself the problem here? Is this all not simply the result of the pressure on the research community to publish in high impact journals for their careers?

Conclusion

So at the end of all of this, is peer review ‘broken’? Yes according to a study of 270 scientists worldwide published last week. But in a considerably larger study published last year by Taylor and Francis showed an enthusiasm for peer review. The white paper Peer review in 2015: a global view,  which gathered “opinions from those who author research articles, those who review them, and the journal editors who oversee the process”. It found that researchers value the peer review process.  Most respondents agreed that peer review greatly helps scholarly communication by testing the academic rigour of outputs. The majority also reported that they felt the peer review process had improved the quality of their own most recent published article.

Peer review is the ‘least worst’ process we have for ensuring that work is sound. Generally the research community require some sort of review of research, but there are plenty of examples that our current peer review process is not delivering the consistent verification it should. This system is relatively new and it is perhaps time to look at shifting the nature of peer review once more. On option is to open up peer review, and this can take many forms. Identifying reviewers, publishing reviews with a DOI so they can be cited, publishing the original submitted article with all the reviews and the final work, allowing previous reviews to be attached to the resubmitted article are all possibilities.

Adopting  one or all of these practices benefits the reviewers because it exposes the hidden work involved in reviewing. It can also reduce the burden on reviewers by minimising the number of times a paper is re-reviewed (remember the rejection rate of some journals is up to 95% meaning papers can get cascaded and re-reviewed multiple times).

This is the last of the ‘issues’ blogs in the case for Open Research series. The series will turn its attention to some of the solutions now available.

Published 19 July 2016
Written by Dr Danny Kingsley