Emma Ganley is the Director of Strategic Initiatives at protocols.io. Emma was formerly Chief Editor of PLOS Biology and worked in an editorial capacity in scientific publishing for 15 years. She has also consulted for the preprint server medRxiv.Which would you trust more, a research article posted as a preprint, or one that has been published after peer review? The reality is that all science communicated via either mechanism should be read with a discerning and critical eye.Under the COVID-19 SpotlightIn recent months, the downsides of both preprints and published articles have been thrown into the public spotlight. For preprints, there have been accusations of insufficient screening and filtering — one example was a controversial preprint that inferred a laboratory origin of SARS-CoV-2. This was quickly withdrawn as soon as concerns were raised (1-2. Pradhan et al. posted on bioRxiv on Jan 31st, withdrawn on Feb 2nd). Tweeters have proven to be very successful reviewers and Twitter an efficient reviewer forum for preprints; and in this instance the concerns and comments from on-the-ball tweeters were immediate, public, and were quickly addressed. As for peer-reviewed publications, there have been some damaging retractions in high-profile medical journals. Very recently, two papers about the globally (and politically) contentious therapeutic approach of hydroxychloroquine were retracted (3-4. Mehra et al. in the New England Journal of Medicine, published on May 1st and retracted on June 4th, and 5-6. Mehra et al. published in the Lancet on May 22nd, retracted on June 5th). These retractions were responsible for the halting of some COVID-19 treatment clinical trials. The issues in these papers were not identified during peer review, but instead were flagged post-publication after media investigations identified anomalies in the data (7).The preprint and publications listed above were all problematic due to medical and/or political interpretations. However, one key difference sits in how results from published papers can sometimes be interpreted as truth. Preprints on the other hand are generally received more cautiously and tend to be subjected to greater scrutiny by the Twittersphere. What’s also interesting are the response times to identify and address concerns and ‘correct’ the problematic papers (see the publication and withdrawal or retraction dates listed above).Dynamism of Preprints and Formal PublicationsFrom a content perspective, the differences between a preprint and the final published version are often minimal. Posted preprints will have passed through a screening process — some authors have been astonished to have their submission returned to them as it didn’t pass the screening process. medRxiv has a rigorous multi-step screening process to try and avoid posting submissions that could be a danger to public health. Preprint servers can be damned if they do and damned if they don’t — there’s an expectation and desire that they not function as gatekeepers and that all submissions should be posted, but they’re vilified if contentious or potentially harmful content makes its way through the filters and is posted. And from preprint to published article, what transformation will have occurred and how? The published article will (or should) have had some number of peer reviewers (let’s say between two and four, on average), and at least one editor (professional or academic) read through with a critical eye and provide feedback—this feedback may or may not have been responded to in one or more revisions. Authors of both preprints and published papers may also have sought feedback from colleagues or peers before submission.There are other differences though, more to do with what you can easily do to correct or update the content after the preprint or article has been published. Once published in a journal, responding to criticism from readers becomes very complicated, difficult and, well actually a bit bureaucratic. A publisher can issue formal corrections or retractions, but these often require post-publication concerns being discussed with at least one academic expert. Corrections and retractions also have associated costs for the publisher, which could lead to recalcitrance and a mentality of only issuing these when absolutely necessary (the publisher generally foots this bill). Retractions are complicated and often only happen after a lengthy investigative process, and ideally, they require agreement from the authors. On the other hand, preprint servers tend to accommodate quick and easy posting of new, updated versions, which can resolve the issues. But it’s also possible to remove posted content quickly if a study is found to be flawed or overtly controversial and a notice would be left in its place. It’s generally easier and faster to address concerns that relate to preprints than for issues flagged about formally published articles.Gaining insight into these processes doesn’t really answer the question of which is more trustworthy. However, it does give us an insight into which of the two routes to publication is more dynamic, versatile, and provides an immediate platform to facilitate effectual peer review, versioning, or withdrawal as appropriate. The Self-Correction of ScienceScience as a discipline should be self-correcting. Researchers strive to understand and make new discoveries, and they won’t get everything correct the first time around. Making incremental advances and insights publicly available quickly can help accelerate science overall, but speedy dissemination without sufficient care and attention to detail can necessitate quick correction. Speed must be juxtaposed against the need to optimize for accuracy; researchers should undertake to maximize confidence in their results and conclusions before making an article public, ideally excluding alternative explanations or interpretations, and/or including a detailed insight into known caveats. Ultimately, readers should never interpret ‘publication’ as a synonym for correct or truth. Both published and preprint articles should always be read with a critical eye. Trustworthiness is not really the right way to be thinking about results presented in publications. Regardless of how many people have seen and read an article already does not mean that all potential issues or errors have been identified. Submitting an article as a preprint, as well as simultaneously to a journal, is the best current approach to ensure there are as many eyes on a piece of research as possible. Via preprint server visibility and journal consideration the research article will reap the benefits from the dynamism of preprint servers, and formal editorial peer review, maximizing the probability that any issues would be identified, flagged, and could be fixed before formal publication. But all authors should check, double check, and triple check all details, data, and analyses in their research articles before they submit. Whether to a preprint server or for peer review at a journal; publish in haste, repent at leisure!Referenced LinksPlos Biology. https://journals.plos.org/plosbiology/medRxiv. https://www.medrxiv.org/Pradhan et al., Jan 31, 2020. bioRxiv “Uncanny similarity of unique inserts in the 2019-nCoV spike protein to HIV-1 gp120 and Gag.” DOI: https://doi.org/10.1101/2020.01.30.927871 Pradhan et al., Feb 2nd, 2020. bioRxiv “WITHDRAWN: Uncanny similarity of unique inserts in the 2019-nCoV spike protein to HIV-1 gp120 and Gag.” DOI: https://doi.org/10.1101/2020.01.30.927871Mehra et al., May 1, 2020. N Engl J Med “Cardiovascular Disease, Drug Therapy, and Mortality in Covid-19.” DOI: https://doi.org/10.1056/NEJMoa2007621Mehra et al., June 4, 2020. N Engl J Med “Retraction: Cardiovascular Disease, Drug Therapy, and Mortality in Covid-19.” DOI: https://doi.org/10.1056/NEJMc2021225Mehra et al., May 22, 2020. The Lancet “RETRACTED: Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis” DOI: https://doi.org/10.1016/S0140-6736(20)31180-6Mehra et al., June 5, 2020. The Lancet “Retraction—Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis” DOI: https://doi.org/10.1016/S0140-6736(20)31324-6The Guardian. ‘Covid-19: Lancet retracts paper that halted hydroxychloroquine trials’. https://www.theguardian.com/world/2020/jun/04/covid-19-lancet-retracts-paper-that-halted-hydroxychloroquine-trials