Brazil built much of its open-science capacity from the ground up, through public infrastructure, scholarly communities, and institutional experimentation. Its achievements—and its unfinished transition beyond open access—show why a national policy matters, but cannot do the work alone.
Long before Brazil had a comprehensive national open-science policy, public and academic institutions were already building systems that made research easier to find and read. Yet Brazil also shows the limits of infrastructure without aligned incentives, training and enforceable standards. The country is strong in open access, but adoption of open data, code sharing, preregistration, protocol sharing and replication remains uneven. The first two sections draw on my personal experience, while the rest of the article examines open science in Brazil from a broader perspective.
Tales from the Crypt: science inside and outside an Institution
During my career in psychology, I have experienced Brazilian research from both sides of an institutional boundary. When I was working within a laboratory and formally affiliated with a university, access to scientific information was considerably easier. Through the publicly funded CAPES Journal Portal, researchers, students, and staff at participating institutions can consult subscription content licensed by the Brazilian government. From inside the university, many paywalls became almost invisible. I could follow a reference, download an article, and continue working without confronting its individual price.
However, the access was not a fully functioning endeavor. Not every journal, database, or publication year was covered, and availability could vary according to the institution and its agreements. More importantly, the apparent openness depended on my institutional credentials. When I worked outside a formal institution, or between affiliations, the same literature often became practically inaccessible. An article that I could download with one click from inside a laboratory was suddenly placed behind a paywall. Conducting a literature review could require paying out of pocket, restricting the review to open-access publications, searching repositories and preprint servers, requesting copies from authors, or relying on informal sharing networks.
That experience taught me that institutional access and open access are not the same. The CAPES Portal is a major public achievement and makes research possible for a large part of the Brazilian academic community. Yet the knowledge it licenses remains available primarily to people who are formally connected to participating institutions. Independent researchers, practitioners, former students, journalists, patients, and members of the public may help finance research through taxation while remaining unable to read its results. Science is not fully open if the reader can enter only while an institutional badge is active.
The institutional divide also shapes who can produce knowledge. In my experience, prestigious universities and established laboratories are better positioned to attract research funding because they already possess the infrastructure, administrative support, senior networks and publication records that funding competitions reward. New grants then strengthen those advantages, making the same institutions even more competitive in the next round. Universities and research groups with fewer resources must compete from a weaker starting point: they have less equipment, less technical support, fewer research assistants and fewer opportunities to produce the outputs needed to demonstrate “excellence.”
The result is a self-reinforcing cycle. Prestige attracts resources; resources generate publications and visibility; and those outputs are then used to justify directing further resources toward the same institutions. Meanwhile, researchers in less-resourced settings are expected to meet similar standards without comparable conditions. When they struggle, the outcome may be interpreted as a lack of excellence rather than a lack of opportunity.
This is why open science became personal to me. Instead of focusing on licenses, repositories, or solely on journal policies. It is about whether the ability to learn from science—and to contribute to it—depends on a university login, a laboratory affiliation, or the prestige of the institution beside one’s name. Publicly funded access can reduce inequality enormously, but genuine openness requires making knowledge available beyond institutional borders while ensuring that the capacity to produce it is not concentrated only where resources already exist.
Weird Science: I learned the norms of publication before I learned the practices of openness
My own scientific training illustrates why changing policies is not enough. I was taught research methods, statistics, and academic writing, but I received almost no systematic training in open-science practices. I was not taught how to preregister a study, create a data-management plan, organize a reproducible workflow, document code for reuse, choose an appropriate license, or distinguish between data that should be open and data that must remain protected. These practices were not presented as an ordinary part of becoming a researcher.
At the same time, I received advice that I would later recognize as encouraging questionable research practices. I do not mean that I was formally instructed to falsify evidence. The process was more subtle—and, for that reason, perhaps more influential. Decisions about which analyses to retain, which findings were worth reporting, and how to transform complicated results into a clean, publishable story were often presented as ordinary parts of research. The priority was not always to document transparently how the evidence had been produced, but to make the final paper more persuasive and easier to publish. HARKing and p-hacking were the most common thing on Earth, for example.
This is how a questionable practice becomes a social norm. It is rarely introduced as questionable. It is learned through comments from supervisors, examples from published articles, conversations with colleagues, and the outcomes that journals reward. When senior researchers recommend a practice, peers reproduce it and successful papers appear to confirm it, an early-career researcher may interpret that practice as evidence of professional competence. What I was learning was not only how to analyze data; I was learning what researchers around me considered normal. My experience should not be treated as evidence about every Brazilian laboratory. Yet it is consistent with a 2019 survey of 232 Brazilian psychologists, in which 85% reported having used at least one questionable research practice.
Open science entered my training much later and without the same social reinforcement. I could develop a favorable personal attitude toward transparency, but that did not mean I knew how to put it into practice—or that the people and institutions around me expected or supported me to do so. Institutional knowledge remains highly uneven. In some universities, there may be no clear guidance on preregistration, repositories, reusable code, data licensing, or responsible sharing. Supervisors may want to support open science while lacking the training needed to guide their students. Evidence from the State University of Maringá captures part of this gap. In a survey answered by 90 graduate supervisors from 892 invited, 70% said they would adopt open-science practices, but 90% had never participated in an open-science activity and 97.8% had received no institutional training. Respondents also reported receiving no institutional incentives or technical support in the areas examined. Although this single-university study cannot represent the entire country, its results resemble what I experienced: researchers may be receptive to openness while their institutions remain unprepared to teach or sustain it.
Under these conditions, open science often reaches researchers as an external expectation rather than an internalized professional value. A journal requests a data-availability statement. A funder adds a data-management requirement. A conference promotes preregistration. These measures can change the perceived social norm—this is something researchers are now expected to do—without necessarily changing personal attitudes—this makes my research better—or perceived behavioral control—I know how to do this and have the resources to do it properly.
This distinction matters. Social norms can initiate change, but adoption based only on external expectations is fragile. If a journal does not verify the requirement, if a supervisor dismisses it, or if the researcher is under time pressure, the older and more familiar norm may prevail. Open science then becomes a compliance exercise: uploading an undocumented spreadsheet, writing that data are “available upon request” or preregistering a vague plan after important decisions have already been made. The appearance of openness changes while the research process remains largely the same.
Sustainable reform therefore requires more than making open science the new thing researchers are expected to say. It requires changing what they believe, what they observe others doing and what they feel capable of doing themselves. Researchers need practical training, examples from their own disciplines, supportive supervisors, appropriate infrastructure and career incentives that reward transparency even when the results are null, complicated or difficult to publish.
Training should also prepare researchers to resist familiar rationalizations: “everyone does it,” “the journal expects a clean story,” “reporting all the analyses will only confuse readers,” or “being completely transparent will make the paper impossible to publish.” Researchers should encounter these arguments early, learn why they are persuasive and rehearse responsible alternatives before facing them under real publication pressure.
I had to learn open science partly by unlearning what had previously been presented as normal research. That experience changed how I understand reform. Policies can establish expectations, but laboratories and institutions teach researchers what those expectations mean in everyday practice. If openness is to become more than another rule, it must be learned as a skill, modeled as a shared value and reinforced as the way credible research is actually done.
Brazil’s Structure and first lesson: build the commons
Brazil’s best-known contribution is SciELO, created in 1997 and launched in 1998—four years before the Budapest Open Access Initiative gave the modern movement much of its vocabulary. SciELO developed as a decentralized, publicly supported publishing infrastructure and now connects national collections across 16 countries.
The 2025 volume Open Science in Brazil: Achievements and Challenges describes the scale reached by 2023: about 1,200 active journals across the SciELO Network, including 318 in the Brazilian collection. It also describes the platform’s evolution beyond journal articles through SciELO Preprints and SciELO Data.
SciELO is only part of the ecosystem. By the end of 2025, the Brazilian Institute of Information in Science and Technology reported that Oasisbr had surpassed 6 million documents from more than 2,000 sources, while the Brazilian Digital Library of Theses and Dissertations had reached 1 million records. These are public-interest infrastructure: standards, metadata, preservation, interoperability, software and human expertise accumulated over decades.
For other countries, the lesson is not simply “launch a repository” like Zenodo, ArXiv or the like. It is to treat scholarly communication as a durable public good. Infrastructure needs recurring finance, professional staff, open standards and governance that is answerable to research communities. Thus, commitment to open science should be a value researchers have built before or during their academic career.
Brazil also offers a warning about how open access is financed. A country can remove barriers for readers only to create new barriers for authors if subscriptions are replaced by article-processing charges. The Brazilian and wider Latin American tradition—largely public, university-based and noncommercial, although not universally fee-free—offers an alternative worth protecting.
Open access is a foundation, not the finished building
Brazil’s strength in open access can obscure a harder question: How open is the research process itself?
A Brazilian-led taxonomy, developed with input from specialists across the Americas, organizes open science into 10 major facets and 96 labels. Open access is only one. Open data, reproducible research, research assessment, infrastructure, education, citizen science and dialogue with other knowledge systems are also part of the picture.
Empirical studies show why this distinction matters. An analysis of Brazilian-authored articles indexed in Web of Science from 2015 to 2018 found that 39% were openly accessible, while identifiable open-data practices remained incipient. A later study covering 2013–2022 estimated that Brazilian authors published a substantially larger share of articles openly than authors worldwide—about 52% versus 28%—with much of Brazil’s open publishing concentrated in Global South journals. Yet open access did not, by itself, erase the citation and prestige advantages associated with highly ranked or Global North journals.
That result should not be misread as evidence that open access lacks value. Citation indicators are not the same as research quality or public benefit. The more useful conclusion is that access alone does not dismantle the hierarchies built into research evaluation.
The same pattern appears in journal policies. A 2024 study of 40 high-performing journals and 400 articles from Brazil, Mexico, Portugal and Spain found Brazilian journals comparatively strong in several measures. But across the full sample, only one journal encouraged preregistration; none encouraged replication studies; and none had implemented open peer review. Identifiers and disclosure statements were more common than deeper practices such as sharing data and materials.
Brazil, in other words, has opened the door to the article more successfully than it has opened every room in the research process.
Policy works—but only where it reaches
It would be wrong to conclude that policy does not matter. The evidence suggests something more precise: well-designed rules can change the behavior they explicitly target, but their effects do not automatically spill over into other practices.
An exploratory analysis of more than 142,000 articles in the PLOS Open Science Indicators dataset, including 3,790 with Brazilian affiliations, illustrates the point. Among PLOS articles, Brazilian data-sharing rates were comparable to the rest of the world—very likely because PLOS requires data availability. But Brazilian rates were generally lower for preprints and code sharing, while preregistration and protocol sharing remained uncommon in both Brazil and the rest of the sample. Because the indicators were generated automatically and the non-PLOS Brazilian comparison group was small, these results are preliminary. Still, the pattern is telling: a data-sharing mandate was associated with stronger data-sharing performance, not with uniformly open science.
SciELO’s own guide based on the Transparency and Openness Promotion guidelines makes policy design unusually concrete. It distinguishes between three levels: disclose a practice, require it, and verify compliance. A policy that praises transparency is not equivalent to one that makes a data statement a condition of publication, and neither is equivalent to checking whether the files are complete and usable.
Brazilian funders show the same mixture of policy effects and policy gaps. A 2024 review of the websites of all 27 state research foundations and their national council found open-science-related actions in only 12 of the 28 organizations examined. Most concerned government or administrative data and compliance with existing access-to-information and data-protection laws.
FAPESP stood out for measures more directly connected to research, including data-management plans and an open-access policy. The study was limited to publicly available website content, so it may have missed provisions contained only in calls for proposals. Even with that caveat, it shows that law can drive action while leaving large parts of the research cycle untouched.
Brazil is now trying to connect these dispersed efforts. Its 2023–2027 Open Government action plan includes a national open-science policy, an implementation plan, monitoring tools, research-assessment reform, educational resources and incentives. But the March 2026 monitoring record described policy development as slow and compromised by its dependence on the still-pending national science, technology and innovation strategy; the implementation plan, in turn, depends on the policy.
This is precisely why policy is necessary and why announcing one is not the same as implementing it.
The missing layer is behavior
Open science asks researchers to do additional, unfamiliar, and sometimes “risky” work: document decisions, clean and describe data, write reusable code, select licenses, protect sensitive information, preregister plans, respond to scrutiny, and maintain files after a project ends. Good intentions are not enough if the surrounding system makes those actions costly or professionally irrational.
A useful behavior-change lens asks three questions. Do researchers believe the practice is worthwhile? Do the norms and rewards around them support it? And do they have the time, skills, and infrastructure to carry it out?
Evidence from Brazil suggests that the first condition is often stronger than the other two. In a survey at the State University of Maringá, 90 graduate supervisors responded from 892 invited. Seventy percent said they would adopt open-science practices, yet 90% had never participated in an open-science activity, and 97.8% reported receiving no university training. The most frequently selected barriers were inadequate infrastructure, funding constraints, and the absence of clear steps. This is a small, single-university survey with a low response rate, not a national estimate. But it captures the implementation gap in miniature: willingness without capability.
The incentive gap is equally clear. In a 2024 survey completed by 355 of 2,179 Brazilian dental researchers invited, 96.1% considered the existing evaluation system flawed. Respondents viewed nontraditional activities as more important for scientific progress and social impact, while traditional outputs—especially publishing many papers in recognized journals—were perceived as more important for career advancement. The field-specific sample and 16% response rate limit generalization, but the contradiction it identifies deserves attention.
Researchers respond rationally to that contradiction. If hiring, promotion, and funding reward publication volume and journal prestige, a policy asking for careful data stewardship or replication adds work without changing the payoff.
In an eLife reviewed preprint from the Brazilian Reproducibility Initiative, 56 laboratories completed 143 replications of experiments using three common biomedical methods. After independent validation, 90 replications of 45 experiments remained; replication rates ranged from 20% to 44%, depending on the prespecified criterion. The study cannot be generalized to all Brazilian biomedicine or used to rank Brazil against other countries. It does show why access to a paper is not enough: clear methods, realistic estimates of variability, usable protocols, infrastructure and incentives for reliable rather than merely positive results all matter.
Nor should such findings be framed as a Brazilian defect. In that field-specific 2019 survey of 232 Brazilian psychologists, 85% reported having used at least one questionable research practice—but similarly high rates appeared in the US and Italian comparison samples. The authors also cautioned about self-report and cross-country comparability. The relevant target is the research system, not national character.
Communities turn rules into social norms
One of Brazil’s most transferable innovations may therefore be organizational rather than regulatory.
Created in 2023, the Brazilian Reproducibility Network brings together institutions, research groups, societies, journals and individuals across disciplines and regions. It works through four connected areas: community, advocacy, education and research, including metascience. Instead of treating researchers as the passive targets of reform, the network makes them co-designers, trainers and evaluators of it.
This is the social infrastructure that policy documents often omit. Policies can authorize change. Funders can finance it. Journals can require it. But communities make new practices understandable, credible and normal.
Borrowing cautiously from inoculation research on self-efficacy—which was not conducted on open science—training should anticipate the arguments researchers will encounter: “I will be scooped,” “my data are too messy,” “nobody in my field does this,” “sharing creates legal risk,” or “there is no time.” Preparing practical answers in advance can strengthen confidence and uptake. But communication must never become a substitute for structural reform. Researchers cannot be persuaded out of missing storage, understaffed libraries, insecure contracts or evaluation systems that reward the opposite behavior.
Seven lessons other countries can take from Brazil
1. Build infrastructure with policy, not after it
Mandates without repositories, data stewards, preservation and technical support are unfunded obligations. Finance open infrastructure as a continuing public service, with interoperable standards and community governance.
2. Do not use “open access” as shorthand for “open science”
Track access to articles separately from data, code, materials, protocols, preregistration, peer review, citizen participation, and responsible research assessment. Different disciplines will need different combinations and justified exceptions.
3. Move deliberately from encouragement to requirement to verification
State exactly what must be shared, where, in which format, under what license, and with which exceptions. Then verify compliance in substance, not just on paper: a broken link or an undocumented spreadsheet may satisfy the letter of openness while defeating its purpose. Some of this review can be made more scalable with tools such as regcheck or metacheck, which flags potential reporting problems and areas for improvement in manuscripts. Such tools should support expert review, however—not replace it or be used to generate automated judgments of research quality.
4. Change rewards at the same time as rules
Give credit for reusable datasets, documented software, replication, negative results, open educational resources, mentorship and public engagement. Reduce reliance on publication counts and journal brands. Otherwise, openness will remain extra labor performed disproportionately by people with secure jobs and institutional support.
5. Design for attitudes, norms and perceived control—not compliance alone
The Theory of Planned Behavior offers a useful framework for understanding why policy does not automatically change research practice. Researchers are more likely to adopt open-science practices when they see them as beneficial to their work (attitudes), believe that supervisors, colleagues, journals and funders genuinely expect and model them (subjective norms), and feel that they have the skills, time, infrastructure and autonomy needed to act (perceived behavioral control).
Brazil’s experience shows what happens when these conditions diverge. Researchers may support openness in principle while receiving little practical training, observing questionable research practices as the stronger local norm, and working without the resources needed to behave differently. A mandate can change what researchers think they are supposed to do, but by itself it may produce only symbolic or minimal compliance.
Countries should therefore pair policy with credible demonstrations of value, visible role models, hands-on training, protected time, suitable infrastructure and continuing technical support. They should also evaluate actual research behavior—not merely awareness, approval or stated intentions. Open science becomes sustainable when researchers understand its value, see it practiced around them and possess a realistic capacity to participate.
6. Invest in people and communities
Fund librarians, data stewards, research-software engineers, statisticians, and methods training. Support peer networks that can translate broad principles into disciplinary practice. Adoption depends on competence and social norms as much as compliance.
7. Protect equity and responsible limits to openness
Do not replace reader paywalls with unaffordable author fees. Support local-language publishing and noncommercial platforms. Build governance for personal, sensitive, Indigenous and strategically important data. The responsible principle is not “open everything,” but “as open as possible and as restricted as necessary.”
The real Brazilian lesson
Brazil’s decentralized strength also produces fragmentation, regional inequality and uneven institutional capacity. Its long-running effort to create a national framework was still unfinished in the March 2026 monitoring record. And some practices that are easy to endorse remain difficult to perform.
But Brazil changes the question other countries should ask.
The Brazilian question is:
- Where will research outputs live?
- Who will maintain them?
- Who will train and support researchers?
- What will hiring and funding reward?
- How will compliance be verified?
- Which disciplines and communities may be disadvantaged?
A policy without infrastructure is a promise without a delivery system. Infrastructure without incentives is likely to be underused. Requirements without verification become declarations. And openness without equity can reproduce the very exclusions it was meant to remove.
Brazil’s central lesson is therefore not that policy is unimportant. It is that open science is not a document. It is an environment—and environments are built collectively, financed continuously, and sustained through everyday practice.