Studies published in prestigious academic journals receive more attention from scholars and media outlets, shaping public discourse and potentially accelerating academic careers. While the path to publication is often murky, a new analysis of more than 110,000 papers submitted to Science and Science Advances—two leading journals published by the world’s largest general science association—reveals Western scientists, larger teams, top-ranked universities, and select regions have slight advantages in acceptance rates.
The new study by University of Vermont (UVM) and University of Colorado (CU Boulder) researchers is the first to plumb submissions data to top science journals to better understand editorial dynamics. Their analysis stems from an anonymized dataset the team spent years developing to provide a window into the peer review process. The data and analysis were published today in Science Advances.
“This is really a first look at this process that was previously opaque,” says lead author Sam Zhang, an assistant professor of mathematics and statistics at UVM. “With the rise of AI slop and the ability to mass produce junk science, the role of established gatekeepers like these elite journals becomes more important as these processes get put under strain. These types of open data analysis about what is going on at the journal might help the community prepare and understand for how editorial processes will need to change.”
Key findings
- Group size matters. Studies with 10 or more authors are three times as likely to be published than papers with 1-5 authors.
- Reputation plays an outsized role. Manuscripts with corresponding authors from the most prestigious institutions had a 11.6% acceptance rate, compared to 3.4% for papers from the least prestigious institutions.
- Geographic disparities exist. Papers with corresponding authors originally from China were over three times less likely to be published than papers with U.S. or Canada-based corresponding authors.
- Manuscript topic is key. When COVID-19 was a novel disease, studies related to immunology and RNA were nearly six times more likely to be published in Science than social science on the political and societal dimensions of COVID.
The American Association for the Advancement of Science publishes six journals including its flagship publication, Science, and Science Advances, a top-ranked open-access journal. The editors of these journals provided more than 110,000 articles submitted to the publications between 2015 and 2020 for the researchers to build an anonymized dataset and identify patterns within.
“Peer review by scientific experts remains the gold standard for critically assessing whether new scientific work meets or surpasses high standards for scientific discovery,” explains co-author Aaron Clauset, professor of computer science at CU Boulder. “… But the need to keep peer reviews confidential has made it difficult to assess how well these journals achieve the scientific community’s ideals, or to develop and test theories to improve it. Our new anonymized data set, which covers 5 years of detailed peer review data at two elite general science journals, offers a first ever look at how elite peer review works.”
The team found that about 80 percent of submissions to Science and over 70 percent of submissions to Science Advances were screened and rejected by editors prior to undergoing peer review.
“Editors do a lot of the heavy lifting and gatekeeping,” Zhang says. “When scientists submit to journals, they are often emotionally fixated on the peer reviews. But to even get to that stage at an elite journal already puts you in the minority of successful papers. The vast majority never see that stage.”
The number of submissions these two journals combined receive each year is enormous—upwards of 50,000 in 2025 alone, and the number of papers submitted to science journals continues to balloon as artificial intelligence accelerates the number of papers produced each year. For context, the acceptance rate for a paper submitted to premier academic journals is akin to being accepted to an Ivy League college—not everyone can get in. But understanding who may be left out is important, too.
Releasing the dataset of submissions to Science and Science Advances for more in-depth analysis was a goal from the outset and an effort to encourage science journals to be more transparent with their own data and experiment with more equitable submission practices.
“We want the entire community to be able to analyze the data,” Zhang says, “… Seeing it happen at such a standard bearer of science might make other journals less squeamish about pursuing a project like this themselves.”
The authors stress that caution is needed when interpreting the results, particularly for bias.
“A crucial point about our analysis is that none of our findings rise to the level of causality, and they should not be interpreted in that way,” Clauset says. “Rather, what we do find is a number of fascinating associations between certain attributes of manuscripts and their relative success in peer review at elite journals like Science and Science Advances. For instance, we find little overall association between author gender and success in peer review. What little association we do find is consistent with a seniority effect: more senior researchers are somewhat more likely to be men.”
While the scientists were able to track the stages of how a study moved from submission to publication, they could not assess the actual manuscripts for quality of the work accepted versus rejected—there may be justification for disparities in publication rates, too.
For instance, there are institutional and financial incentives in China to submit to these journals, Zhang says, which could increase the number of studies sent that may not be an ideal fit for a general science audience. And while authors from more elite institutions are more likely to get their paper in the door, they also experienced more rejections after being sent out for peer review.
“We don’t know how much of it is due to differences in the novelty and quality of the research versus being sensitive to the status signaling of the authors,” Zhang says. “Realistically, it is likely both. And one thing we would like to see are experiments in double blinding. As of right now, it is single blinded—the authors don’t see who the reviewers are but everyone else sees who the authors are. We really need experiments to really untangle those types of effects.”
In other words, by removing both the names and associations of all authors and editors, it allows scientists to test for bias.
Zhang directs the Science and Humanity lab at UVM’s Vermont Complex Systems Institute which examines the human side of science. Studying the science of science is necessary because science is performed by people who shape how fields evolve: People determine which projects receive funding. People conduct experiments and review each other’s work. And ultimately, people decide which findings should be broadly shared. This is why leading science journals should examine their practices to uncover potential blind spots.
“This is data journals don’t typically look at,” says Zhang. “This might be something an ambitious editor chooses to do voluntarily once every blue moon but making it more of a systemic practice might raise awareness about disparities that are leaking into editorial process that aren’t intentional or undesirable byproducts.”
Visit Complex Stories for additional analysis.