In the late summer of 1991, a physicist at Los Alamos National Laboratory named Paul Ginsparg set up an automated email server to share preprints among a few dozen colleagues in high-energy physics. The server, which later became arXiv, was meant to speed up the circulation of results before journal publication. It did that, but it also quietly started a transformation that few anticipated: it turned the physics community's preprint culture into the backbone of a global publishing economy. Three decades later, the same server that was built to bypass paywalls now sits at the center of a system where researchers pay thousands of dollars to have their work peer reviewed, and where the largest costs are borne not by readers but by the scientists themselves.

The Vault That Turned Peer Review Into a Paid Service

arXiv is often described as a preprint server, but it functions more like a vault. It holds over two million papers, many of them in physics, mathematics, and computer science. When a researcher uploads a paper, it is not formally peer reviewed; it is screened by moderators for relevance and basic quality, then posted for the world to see. This has accelerated scientific communication enormously. A particle physics result can be read by thousands of people within days of being written, rather than months after submission to a journal.

The success of arXiv created a curious double standard. In most fields, a preprint is a draft, and the version of record is the one published in a journal after peer review. But in physics, the preprint is often the version everyone reads and cites. Journals still exist, and they still charge for publication, but their role has shifted from gatekeeping to certifying. That certification is expensive, and the cost is now borne by the authors, not the readers.

The shift did not happen overnight. In the 1990s and early 2000s, most physics journals were subscription-based, and libraries paid the bills. But as digital distribution made it clear that the marginal cost of sharing a paper was near zero, a movement for open access gained traction. The idea was that publicly funded research should be freely available to the public. The practical question was how to pay for the editorial work, the typesetting, the hosting, and the peer review coordination. The answer, in most cases, was to charge the authors.

This is the inversion. Before, the reader's institution paid to access the literature. Now, the author's institution pays to publish in it. The physics vault, which was supposed to liberate science from paywalls, ended up enabling a system where the paywall was moved to the front of the process. The preprint is free, but the stamp of approval is not.

How Open Access Became a Business Model

The mechanism that made this work is the article processing charge, or APC. A journal that is fully open access charges authors a fee to cover its costs. APCs vary widely, but in many reputable journals they run in the thousands of dollars. Some journals charge more than $5,000 per paper, while others, particularly in less wealthy fields, charge a few hundred. The fees are often waived for authors from low-income countries, but the default is that someone has to pay.

Libraries, meanwhile, face a double burden. They still pay subscription fees for journals that are not fully open access, and they also pay APCs for their researchers in open access journals. This is called the double-dipping problem. A library might pay a publisher $2 million a year for a bundle of journals, and then pay another $200,000 in APCs for its faculty to publish in those same journals. The publisher collects both the subscription and the fee. Some publishers have introduced offsetting agreements, but they are complex and not universal.

For individual researchers, the cost is felt directly in their grants. A typical grant from a national funding agency might include a line item for publication costs, but it is often a small fraction of what is actually needed. In fields like physics, where a single collaboration can produce dozens of papers a year, the APCs can quickly eat into the travel budget or the equipment budget. Researchers have to choose between publishing in a high-prestige journal and buying a new detector component.

The economics are even more skewed for early-career researchers. A postdoc or a graduate student does not usually have their own grant money to pay an APC. They have to ask their supervisor, who has to ask the department, which has to ask the dean. This creates a system where the ability to publish in the best journals depends on institutional wealth. A researcher at a well-funded university can publish anywhere; a researcher at a less wealthy institution, even a brilliant one, is effectively priced out of the top tier.

The Worked Example: A Particle Physics Collaboration

To see how this plays out in practice, consider a large particle physics collaboration, such as those working at the Large Hadron Collider at CERN. These collaborations often have thousands of members, and they publish hundreds of papers per year. Each paper lists every member as a co-author, so a single paper might have 3,000 authors. The APC is charged per paper, not per author, so the cost per paper is the same as for a single-author paper. But the collaboration publishes so many papers that the total annual cost is staggering.

Some estimates put the annual APC expenditure for a major LHC experiment in the range of several hundred thousand to over a million dollars. That money comes from the collaboration's budget, which is ultimately funded by national agencies. In other words, the public pays for the research, then pays again to have the results published, and then pays a third time when libraries subscribe to the journals that contain those results. This is not a hypothetical; it is the routine operation of the system.

The APCs vary by journal, and collaborations often choose to publish in high-impact journals like Physical Review Letters or the Journal of High Energy Physics. These journals have APCs that can be in the thousands of dollars. For a collaboration that publishes 200 papers a year, even an average APC of $3,000 adds up to $600,000. That is a significant line item in any budget, and it is money that could have been spent on data analysis, computing, or training the next generation of physicists.

The irony is that the preprint, which is the version most physicists actually read, costs nothing. The peer-reviewed version, which is the one that gets cited in tenure files and grant applications, is the one that carries the price tag. This has led some physicists to question whether the journal is worth the cost. But the incentive structure is such that they cannot simply skip the journal step; hiring committees and funding panels still expect to see publications in recognized venues.

The Evidence That the Model Distorts Science

The shift to a pay-to-publish model has consequences beyond the financial. A growing body of evidence suggests that the current system distorts what gets studied, what gets published, and how quickly. One of the most cited effects is the pressure to produce positive results. Journals prefer papers with novel, positive findings, and this pushes researchers to either avoid negative results or to frame them as secondary. In physics, negative results are often published as null measurements, but they are harder to get into high-prestige journals.

The replication crisis, most famous in psychology and biology, also has a physics dimension. For instance, in 2014, the BICEP2 collaboration announced the detection of primordial gravitational waves, but the result was later retracted after further analysis revealed that dust in the Milky Way could explain the signal. Similarly, in 2011, the OPERA collaboration reported neutrinos traveling faster than light, but the finding was retracted after a faulty fiber optic cable was identified. These examples illustrate how the rush to publish and the lack of thorough verification can lead to errors that are costly in time and reputation.

There is also evidence that the APC model creates a perverse incentive for journals to accept more papers. Because the journal's revenue depends on the number of papers it publishes, there is a temptation to lower the bar for acceptance. This is not true of all journals, and many maintain rigorous standards, but the structural incentive is there. A journal that rejects too many papers is leaving money on the table.

Some studies have looked at the relationship between APC and editorial standards. While it is difficult to measure quality directly, there are indicators such as the rate of corrections or retractions. One analysis by Kyle Siler and colleagues in 2018 examined the relationship between journal prestige and retraction rates, finding that higher-prestige journals had lower retraction rates, but the study did not focus on APCs specifically. Another study by Vincent Larivière and colleagues in 2015 found that the share of open access articles has grown, but the quality varies. The evidence is suggestive, not conclusive, but it is enough to worry researchers.

Who Actually Pays? The Redistribution of Cost

The burden of the APC model falls unevenly across the global research community. Researchers in high-income countries are more likely to have their APCs covered by their institutions or grants. Researchers in middle- and low-income countries often have to rely on waivers, but waivers are not always available, and the process of applying for one can be daunting. A study by Juan Pablo Alperin and colleagues in 2016 found that many eligible authors do not even know the waivers exist.

Within a single country, the burden also falls on the early-career researchers who are least able to absorb it. A graduate student who needs to publish to get their degree may have to choose between a paywalled journal that is free to publish in and an open access journal that charges a fee. If their supervisor cannot pay, they may have to go with the paywalled journal, which means their work is less accessible, which means it is cited less often, which hurts their career prospects. This point resonates with a broader pattern: the APC model exacerbates existing inequalities in science, favoring those with institutional wealth.

The redistribution also affects the kind of research that gets done. Fields with large, well-funded collaborations can absorb the costs, but smaller fields, such as certain areas of pure mathematics or theoretical physics, may struggle. A mathematician who writes a single-author paper in a top journal might pay $3,000, which is a significant portion of their discretionary research budget. In fields where grants are smaller, the APC can be a deterrent to publishing in the best venues.

There is also a global inequity in access to the published literature. Even with open access, the preprint is often the only version that scientists in developing countries can read, because they cannot afford the subscription fees. And while the preprint is free, it is not always the final version, and it may lack the formatting and editorial polish of the published paper. This creates a two-tier system where the richest institutions get everything, and the poorest get only the drafts.

The Counter-Argument: What APCs Buy for Readers in Developing Countries

It would be unfair to paint the APC model as an unmitigated disaster. For readers in developing countries, the shift to open access has been a net positive. Before the open access movement, a researcher in a low-income country often had no access to subscription journals at all. The cost of a single journal subscription could exceed the entire library budget of a small university. Open access journals, funded by APCs, have made a vast amount of research freely available to anyone with an internet connection. This has democratized access to knowledge in ways that the subscription model never could.

Consider a researcher in sub-Saharan Africa. In the 1990s, they might have had to rely on interlibrary loans or personal contacts to obtain papers. Today, they can download the latest findings from open access journals without paying a cent. This is a genuine benefit, and it should not be dismissed. The APC model, for all its flaws, has helped to level the playing field for readers, even if it has tilted it for authors.

The tension is that the same model that benefits readers in developing countries can hurt authors in those same countries. A researcher in Nigeria or Vietnam may struggle to pay an APC, and if they cannot, they may be excluded from publishing in the most visible venues. Some publishers offer waivers, but they are not always sufficient or well-publicized. This creates a paradox: the system that opens access to readers can close it to authors.

This trade-off is not easily resolved. It suggests that the solution is not simply to abandon APCs, but to design funding mechanisms that support authors from all backgrounds. Plan S, for example, includes provisions for waivers and for funding APCs centrally, but implementation varies. The debate is not between open access and subscription; it is about who bears the cost and how to make that cost equitable.

Practical Takeaways for Researchers and Funders

For individual researchers, there are steps to mitigate the cost. First, negotiate APC waivers. Many publishers offer discounts for authors from low-income countries, and some journals will waive the fee if you ask, especially if your institution has an offsetting agreement. It is worth checking the journal's policy before you submit, rather than after acceptance.

Second, use preprint servers like arXiv for early feedback. This is already common in physics, but it can be extended to other fields. Posting a preprint before submission to a journal can help you catch errors and improve the paper, and it ensures that your work is accessible even if the journal is paywalled. This is not a substitute for peer review, but it is a complement.

Third, push for systemic changes. Plan S, an initiative by a coalition of European funders, requires that research funded by their grants be published in fully open access journals or platforms. This has forced many publishers to change their business models, and it has created new funding mechanisms that cover APCs centrally. Similar initiatives are being discussed in other countries, and researchers can support them by signing open letters and advocating for open mandates.

Fourth, support community-led publishing initiatives. There are examples of journals that are owned by scholars and operate on a non-profit basis, such as the Journal of Machine Learning Research or the Open Library of Humanities. These journals often have much lower APCs or no APCs at all, and they are run by academics who are committed to keeping the costs down. By submitting your best work to these journals, you help them gain prestige and influence.

None of these steps is a complete solution, and the system is unlikely to change overnight. The physics vault that started as a tool for sharing preprints has become a fixture of the research landscape, and its role in the publishing economy is complex. The next decade will likely see continued tension between the ideals of open access and the realities of funding. But by understanding how the costs are distributed, and by making deliberate choices about where to publish and how to advocate, researchers can help steer the system in a direction that serves science rather than the other way around.