Language Barriers in Science and Research: The Knowledge We're Missing

Science aspires to universality. Its lingua franca is anything but. With 98% of the most-cited research published in English, the global scientific enterprise systematically disadvantages non-English-speaking researchers, buries knowledge produced in other languages, and creates blind spots in our understanding of the world — particularly in ecology, medicine, and conservation where local knowledge is irreplaceable.

How English Conquered Science

Scientific publishing wasn't always in English. Before World War II, German was arguably the dominant language of science — particularly in chemistry, physics, and medicine. French and Latin held historical prominence. The shift to English dominance was accelerated by two world wars that scattered German-speaking scientists to English-speaking universities, the post-war economic and institutional dominance of the United States, and the emergence of Science and Nature as the two most prestigious journals in the world.

By the 1980s, the shift was effectively complete in the natural sciences. By the 1990s, the social sciences had followed. Non-English journals survived in some fields, particularly regional medicine and local ecology, but their citation counts and perceived prestige lagged so far behind English-language publications that career incentives for non-English researchers increasingly pointed toward English regardless of native language.

Today, researchers whose native language is Spanish, Mandarin, Japanese, Hindi, Arabic, Portuguese, or any of thousands of other languages face a structural choice: publish in their native language to reach local audiences and practitioners, or publish in English to access global academic prestige and citation metrics that determine funding, hiring, and career progression. For most researchers at institutions that track international impact, this isn't much of a choice.

The Costs of the Monoculture

Research That Disappears

Perhaps the most consequential documented impact of scientific language barriers is the research that effectively ceases to exist when published only in non-English languages. A landmark 2021 study in PLOS Biology, examining conservation biology specifically, found that approximately 36% of research relevant to conservation of threatened species exists only in non-English languages — Spanish, Portuguese, Mandarin, French, and others — and is effectively invisible to the international scientific community that makes policy and management decisions.

The implications are stark. Conservation priorities are set based on synthesized scientific evidence. If more than a third of that evidence isn't accessible to the synthesizers — not because it's hidden, but because it's in a language they don't read — conservation decisions are being made on a systematically incomplete evidence base. The researchers who study the local ecosystems in the languages of the people who live in them are producing knowledge that doesn't reach the people who need it.

"We discovered a Brazilian study that had documented exactly the intervention we were proposing, demonstrated it failed, and explained why. It was published in Portuguese in 2014. Nobody on our international team read Portuguese. We learned about it when a Brazilian colleague reviewed our preprint. Years of work had nearly repeated a documented failure — because of a language barrier in the literature."

— Conservation biologist, cited in 2022 PLOS ONE correspondence

This pattern extends beyond conservation. Traditional and indigenous knowledge systems — medicine, agriculture, ecology, materials science — are often documented in local languages and have limited entry points into English-language scientific literature. The ethnobotanical knowledge of Amazon indigenous communities, the agricultural practices documented by local agricultural extension services in non-English languages, the traditional medical practices studied by local researchers in dozens of countries — this knowledge exists, is documented, and is largely inaccessible to the global scientific community that might validate, extend, or apply it.

The Productivity Tax on Non-English Researchers

For researchers whose native language is not English, publishing in English imposes a measurable productivity cost. Multiple studies across disciplines have estimated that non-native English-speaking researchers spend 15-30% more time preparing manuscripts for English-language journals than their native-English-speaking counterparts — time spent on translation and editing rather than additional research.

This time cost is directly subtracted from research productivity. Researchers who could spend that time running additional experiments, analyzing additional data, or developing additional hypotheses are instead polishing prose in a language that isn't their own. Over a career, the compounded productivity loss for non-native English researchers is substantial — and the broader scientific community loses the additional discoveries that time might have produced.

The manuscript time tax is compounded by rejection rates. Studies of peer review have found that papers from non-native English-speaking authors are rejected at higher rates at English-language journals — and while quality differences may account for some of this disparity, analysis of blinded reviews suggests that assessable language quality affects reviewer ratings even when content is controlled. Reviewers struggle to separate "the writing is non-native" from "the science is unclear," and the conflation disadvantages non-English speakers regardless of their research quality.

The Conference Problem

Scientific conferences are where knowledge is shared informally, collaborations are initiated, and reputations are built. International conferences in most scientific fields operate almost entirely in English — presentations, poster sessions, hallway conversations, and social events. For researchers whose English is limited, international conferences are professionally hostile environments despite their nominal internationalism.

The consequences compound in multiple directions. Non-English researchers present less frequently, are less likely to be invited as keynote speakers, and have fewer opportunities to build the informal relationships that lead to co-authorship and collaboration. Their absence from conference visibility reinforces their absence from editorial boards and grant review panels, which reinforces their absence from the power structures that determine funding priorities and publication standards. The pipeline from non-English to underrepresented in scientific power structures is well documented if underacknowledged.

The Peer Review Dimension

Peer review is the primary quality control mechanism in science. It is also a profoundly language-dependent process — and one where language barriers systematically disadvantage non-English researchers in ways that the scientific community has been slow to address.

Reviewers for English-language journals are typically English-speaking researchers. When they encounter manuscripts from non-native English authors, they must make a judgment: is the language quality insufficient to assess the science? Is this unclear writing or unclear thinking? Even well-intentioned reviewers struggle to maintain clean separation between these questions, and research on reviewer behavior suggests that non-native language features — the minor grammatical patterns that characterize second-language writing, even at high proficiency levels — negatively influence scientific quality assessments independent of actual scientific merit.

The result is that English-language journals receive poorer reviews for identical science when it's written with non-native language features. This doesn't mean all rejections of non-native writing are biased — genuine quality differences exist. But the conflation of language and science quality represents a systematic bias in science's primary quality-control mechanism that is not adequately accounted for in how we think about peer review's validity.

The Ecology of Lost Knowledge

Regional and Local Science

Some science is inherently local. The ecology of a specific watershed, the disease patterns of a specific regional population, the agricultural challenges of a specific climate zone — this research is most likely to be conducted by researchers with direct access to the system being studied, who are often researchers embedded in local institutions that publish in local languages.

The global scientific community's English dominance creates a structural devaluation of this local knowledge. A paper characterizing the ecology of a Peruvian mountain ecosystem, published in a Spanish-language Peruvian journal, produces less career benefit for its authors than a paper characterizing the ecology of a well-studied English-language system published in a top English-language journal — even if the Peruvian ecology paper documents something genuinely new and important. The incentive structure steers researchers toward English regardless of where the important knowledge is.

Indigenous Knowledge Systems

Indigenous communities across the world have developed sophisticated knowledge systems over generations — agricultural practices adapted to local climates, medical uses of local plants, ecological understanding of specific habitats. This knowledge often exists in indigenous languages and in oral tradition rather than written form, and its interface with formal science is complicated by language barriers on both sides.

When ethnobotanists, ecologists, or medical researchers want to engage with indigenous knowledge, they face the double barrier of language (indigenous language → researcher's language → English publication) and epistemological translation (oral, contextual, relational knowledge → formal, written, decontextualized scientific literature). Much is lost at each step. What survives to English-language publication is a thin extract of what actually existed in the indigenous knowledge system.

"The scientific literature on traditional plant medicine from our region is almost entirely written by outsiders who spent weeks here and published in English. The practitioners themselves — the people who know — have never published in any scientific journal. Their knowledge is documented in their language, in their tradition. When they die, what we lose isn't in any database."

— Ethnobotanist working in Central Africa, 2021 interview

What Science Is Doing About It

Open Access and Translation

The open access movement has significantly increased the visibility of research by removing paywalls, but has done relatively little to address language barriers. A paper freely available in English remains inaccessible to Chinese-speaking practitioners who can't read it. Open access and language access are distinct problems, and the scientific community has invested much more in the former than the latter.

Machine translation of scientific literature is improving — Google Scholar increasingly offers translation options, and specialized scientific translation tools are advancing in accuracy. But technical vocabulary, specialized notation, and the precise language of scientific claims remain significantly more error-prone in machine translation than general text. A mistranslated statistical term, a wrongly interpreted confidence interval description, a misrendered molecular structure name — these errors in scientific communication can have substantial downstream consequences.

Journal Policies

Some journals have begun explicitly acknowledging the language barrier problem. A small number of high-profile journals now offer professional editing support to authors who identify as non-native English speakers, recognizing that manuscript rejection based on language quality is a poor proxy for scientific quality. The journals that offer this are still a small minority, and the availability of editing support doesn't eliminate the underlying burden on non-native authors of needing to write in a second language.

A few journals have experimented with accepting submissions in multiple languages with parallel English abstracts. These experiments have been limited and mixed in their outcomes — the citation economy still strongly rewards English-language publication, and journals that publish non-English content face challenges in peer review (where do you find reviewers for papers in 20 languages?) that are genuinely difficult to solve at scale.

The Multilingual Preprint Experiment

Preprint servers — repositories where researchers post papers before peer review — have seen experiments in multilingual posting. Some researchers now post both their preprint and a version in their native language, reaching local practitioners who need the research but can't access English. These are grassroots experiments rather than systematic solutions, but they represent a growing recognition in the research community that the English monoculture is a problem with real costs.

The Future of Multilingual Science

AI-powered translation is advancing rapidly and its implications for scientific language barriers are significant. Neural translation systems trained on scientific literature are becoming able to handle technical vocabulary with substantially higher accuracy than general translation models. If the quality of machine translation of scientific papers reaches a point where practitioners can trust it for professional decisions — a threshold we haven't quite reached but are approaching — the access problem changes substantially.

What machine translation doesn't solve is the production side: the structural incentives that push researchers toward English regardless of native language, the citation economy that rewards English-language publication, the conference networks that favor English speakers. Even perfect translation into every language wouldn't address the fact that non-English-speaking researchers produce less English-language research than their scientific capacity would suggest, because the costs of doing so are systematically borne by them and not by the system.

The structural problem requires structural solutions: journals that explicitly value and facilitate non-English contribution, funding bodies that don't penalize researchers whose citation counts are lower because they publish in local-language venues, conference organizers who provide simultaneous translation rather than English-only programming. These aren't impossible — they're choices the scientific community hasn't yet made at scale.