In a world where most science writing is in English, a new study invites readers to listen to the chorus of languages that still carry knowledge across borders. The paper asks not just who publishes, but in what languages, and when readers cite those languages, too.
The team behind the work is based at Université de Montréal, with lead author Carolina Pradier and senior author Vincent Larivière, together with Lucía Céspedes, all centered at the École de bibliothéconomie et des sciences de l’information, and linked to the Consortium Érudit and other research partners. They pulled data from two massive bibliometric databases—OpenAlex and Dimensions—to map the language of publication and the language of cited references across 1990–2023, covering tens of millions of articles and hundreds of billions of citation links. The big takeaway is sobering and hopeful at once: English remains the dominant language of science, but non-English publishing and local-language citation circuits are surging in important regions and fields, strengthening the global conversation rather than drowning it in English alone.
Two languages, one planet, many voices
English still dwarfs everything else, but the pace of change is notable. In Dimensions’ dataset, English language publications fell from 93.86% of articles in 1990 to 85.51% in 2023. The absolute number of English-language papers rose dramatically—from about 877,000 in 1990 to nearly 5 million in 2023—but the share of non-English work climbed as more journals and articles in Indonesian, Portuguese, and Spanish entered indexers and gained visibility.
Indonesian is the standout surprise. It went from nearly invisible to accounting for 2.69% of publications by 2023. This wasn’t arbitrary: in 2014 the Indonesian government mandated open publication of scientific outputs, a policy move that rippled through universities and journals. Meanwhile, Latin American languages—Portuguese in Brazil and Spanish across the region—also rose, sometimes surpassing several Central European languages in share of publication by the end of the period. French suffered the sharpest decline among major languages, while German’s share ebbed in the last decade.
On the inputs side, the language story runs a notch deeper: 98.89% of cited references are in English. That reflects how many sources simply exist in English, but the pattern hides the fact that many non-English literatures circulate in robust regional circuits too. The paper introduces the idea of own-language preference—the tendency to cite literature written in one’s own language—and its variation by field and by language community. For Indonesian, more than a quarter of references from Indonesian-language articles point to Indonesian sources; for Latin American languages, the own-language effect is real but less pronounced. Taken together, these patterns connect multilingualism to bibliodiversity—the idea that science thrives when a diversity of languages and knowledge systems circulates side by side.
Discipline matters too. In the natural sciences and engineering, almost all publications are in English, while in the social sciences and humanities the English share, while still large, is lower, leaving more room for non-English journals and local conversations. That suggests that the power of a language in science isn’t just about who uses it, but what kind of science is being done in which places.
Journals as gateways and guardians
Journals sit at the intersection of communication and prestige. The study finds that non-English journals account for roughly a fifth to two-fifths of journals in MED and NSE, while SSH hosts a somewhat larger non-English presence. But when you look at what gets published in multilingual journals, most of it—almost half in some cases—still ends up in English, underscoring a tension between broad accessibility and the drift toward English as the lingua franca of indexing and evaluation.
Latin American and Indonesian publishing ecosystems have built durable regional infrastructures that support local-language output. Spanish and Portuguese journals have grown hand in hand with Latindex, Scielo, Redalyc, and LaReferencia in Latin America, while Indonesia has become a world leader in open access publishing via the Open Journal System, helping editors publish in Indonesian and in multiple languages. These moves aren’t ornamental; they create real channels for local scientists to circulate findings and for the public to access science in languages they trust.
Yet the study also notes a practical reality: even when journals accept submissions in several languages, many articles in multilingual venues are still in English. And English-dominated journals tend to cite English-language sources almost exclusively. That shows how the prestige economy of science—where journals and publishers shape what counts as authoritative—can reinforce a single language even as multilingual niches survive. It’s a reminder that the road to bibliodiversity isn’t just about opening doors; it’s about rethinking what counts as evidence and value in different languages.
What this means for researchers, funders, and the public
The authors argue that recognizing the value of national-language publishing matters for equity and for knowledge justice. Policies that encourage publishing in local languages alongside English have tangible effects on how knowledge circulates, who can participate, and how science speaks to local communities. In other words, language policies are science policy—shaping who gets heard and how widely.
There are caveats worth holding in mind. The data come from large bibliometric databases that, while broad, have biases—non-English items are sometimes misclassified as English, and citation networks are sparser for non-English work. The authors caution that their findings are lower-bound estimates of multilingualism, which means the actual linguistic life of science could be even richer than the maps suggest.
What does this mean for researchers halfway through their careers and those just starting out? It’s about choices. Publishing in one’s national language can sharpen local impact, while English can unlock international reach. The study suggests a deliberate strategy: treat multilingualism as a resource rather than a hurdle; cultivate local journals, participate in regional networks, translate and summarize work for broader audiences, and value citation of non-English sources as legitimate scholarly work. For funders and policymakers, the message is practical: support journals and platforms that maintain bibliodiversity, invest in language-appropriate metrics, and resist the overemphasis on English-only prestige as the sole measure of impact.
And for readers, the takeaway is hopeful. The science we rely on is not a single voice but a chorus in many tongues that reflects diverse ways of knowing. When Indonesian researchers publish in Indonesian, when Brazilian scholars publish in Portuguese, and when Spanish-speaking scientists publish in Spanish, science broadens its base. It becomes more comprehensible, more participatory, and—crucially—more just. The rising currents of multilingualism in science remind us that knowledge is not the monopoly of a single language but a shared ecosystem that benefits from many languages talking to one another.
