Why science settled on English, and what it costs the people it was not written for

The short answer

Scientific papers are written in English because the twentieth century's two wars, one boycott and one migration moved the centre of research from a three-language Europe to the English-speaking world, and because a shared language, once it exists, rewards everyone who joins it and punishes everyone who does not. There was nothing inevitable about English. Around 1900 a chemist who could not read German was unemployable, and a mathematician who could not read French was missing half the field. Within three generations that world was gone.

The result is a system that works remarkably well for people who grew up in English and quietly taxes everyone else in time, rejections and silence. This post is about how that happened and what the tax looks like from the inside, because the people paying it rarely see the bill itemised.

Before English: a science in three languages, and before that in one

For most of the scientific revolution the language of learning was Latin. Newton published the Principia in Latin in 1687 and Opticks in English in 1704, and the choice was a statement about audience: Latin for the learned of Europe, English for the local reader. A Latin translation of Opticks followed in 1706, for the readers the English edition had left out.

By the nineteenth century Latin had faded and no single language had replaced it. Science was conducted in roughly three: German, French and English, with each holding a home advantage in particular fields. Chemistry and much of physics ran through German journals such as the Annalen der Physik and the Berichte der Deutschen Chemischen Gesellschaft; Einstein's 1905 papers appeared in the Annalen, in German. Mathematics and parts of medicine leaned French. English held ground in geology, biology and engineering. A working scientist read all three or accepted being a generation behind.

The historian Michael Gordin, whose *Scientific Babel* (University of Chicago Press, 2015) is the fullest account of this story, describes a long stretch from about 1880 to 1910 in which German, French and English carried roughly equal numbers of publications, with German pulling ahead by 1910. English was one of three, and not the leading one. That is the baseline to hold in mind, because everything that follows is a story of how a balanced system tipped.

How the shift happened

There is no single cause, and the honest version is a sequence.

The First World War and the boycott. The International Research Council, founded in 1919 by the academies of the victorious Allies, was built on the principle that the defeated Central Powers had no place in it. Scientists from Germany and Austria were excluded from its congresses, and the German language was barred from them. The boycott held until June 1926, when Germany joined the League of Nations. Seven years is not long in a career, but it was long enough: conferences, journals and standards bodies re-formed without the country that had led chemistry and physics, and they re-formed in French and English.

The war at home in the United States. Anti-German feeling during and after the war also reached classrooms. Several US states restricted or banned the teaching of German. Nebraska's law of April 1919 forbade teaching any subject in any language other than English below the ninth grade, and a teacher was convicted for reading a Bible story in German to a ten-year-old. The Supreme Court struck the law down in Meyer v. Nebraska, 262 U.S. 390 (1923), but by then the generation that would have learned German as the language of chemistry had already been steered away from it. American science, growing fast, grew up monolingual.

The 1930s. The Law for the Restoration of the Professional Civil Service, enacted on 7 April 1933, dismissed academics who had a Jewish grandparent or the wrong politics. It took about a fifth of the country's university faculty, and it took the best of them: one study of the physicists expelled found that the 15 percent who lost their posts accounted for 64 percent of German physics citations. Most went to Britain and the United States and published, from then on, in English. The centre of physics moved with them.

After 1945. The United States emerged with the largest research budget in the world, new funding agencies, and journals that everyone wanted to be in. The Soviet Union built a large scientific establishment in Russian, and for a while the world ran on two languages; the American answer was to fund translation rather than to learn Russian. The National Science Foundation paid for cover-to-cover translations of entire Soviet journals: the American Institute of Physics began publishing Soviet Physics JETP in 1955 on an NSF grant, and by the end of the decade one publisher alone was producing fourteen translated journals. It is worth sitting with what that choice says. Faced with a large body of science in another language, the richer system translated it wholesale rather than read it, which tells you which way the current was already flowing. When the Soviet system collapsed, its scientists also moved to English.

The compounding phase. By the last quarter of the century the shift no longer needed a push. Citation databases indexed English journals; funders and hiring committees counted publications in indexed journals; researchers everywhere concluded, correctly, that a paper in German or French would be read by fewer people and count for less. Every individual decision to publish in English was rational, and together they made the next decision more so. Ulrich Ammon's figures put English at about 90 percent of publications in the natural sciences in a selection of international journals by the mid-1990s, and about 82 percent in the social sciences and humanities.

The share has actually slipped since. A 2026 study of the major databases by Carolina Pradier, Lucía Céspedes and Vincent Larivière at the Université de Montréal found English falling from 94 percent of indexed articles in 1990 to 85 percent in 2023. That sounds like the tide turning, and it is not. Over the same period the number of English articles grew from roughly 877,000 a year to nearly five million, and 98.9 percent of the references those articles cite are in English. More science is published in other languages than before, and almost none of it is read. The dominance moved from the writing to the reading, where it is harder to see and harder to escape.

Note what is missing from this sequence: any claim that English is easier, clearer or more suited to science. It is not. It won a war of attrition, and the prize was the network.

What it costs today

A shared language is a genuine good. Researchers in Lisbon and Lagos can read the same paper, cite each other and argue in public. Nobody sensible wants to return to three languages. But the good is not free, and it is not evenly paid for.

The clearest recent evidence is Amano et al., "The manifold costs of being a non-native English speaker in science", published in PLOS Biology in 2023, which surveyed 908 environmental scientists across eight countries with very different relationships to English, from Bangladesh and Bolivia to Japan, Nigeria, Spain, Ukraine and the United Kingdom.

Its findings are the bill itemised. Among early-career researchers with one publication, those from countries with low English proficiency spent a median 90.8 percent more time reading a paper in English than native speakers did, and those from moderate-proficiency countries about 46.6 percent more. Writing cost 29.8 to 50.6 percent more time. Between 35.9 and 38.1 percent of non-native speakers had a paper rejected on the strength of its English, against 14.4 percent of native speakers, and about 42.5 percent were frequently asked to improve the language during revision, against 3.4 percent. Preparing a talk took 38 to 94 percent longer. Around 30 percent of early-career non-native speakers reported avoiding conferences altogether, and roughly half avoided giving oral presentations.

Read those numbers as the study frames them. They are medians for particular groups, sharpest at the start of a career and in countries where English is least present in daily life, which is to say the tax lands hardest on the people with the least slack to absorb it.

There is an argument on the other side, and it deserves a hearing. In "Academic publishing and the myth of linguistic injustice" (Journal of Second Language Writing, 2016), Ken Hyland argued that the framing is overstated: that academic English is nobody's first language, that native speakers also struggle with the register, and that emphasising the disadvantage risks discouraging the very people it describes while distracting from larger inequities in access to research funding and resources. Politzer-Ahles and colleagues replied in the same journal that year, under the title "Is linguistic injustice a myth?", and the exchange is worth reading in full. He is right that the register is learned by everyone. He is wrong, on the evidence above, that it is learned at the same price. Both things can be true: nobody is born writing a methods section, and some people start further from it.

What the cost looks like from the inside

Statistics describe the population. Here is what the tax feels like for one person, because that is where it is paid.

Reading. The paper takes an afternoon instead of an hour, not because the science is hard but because every hedged sentence has to be unpacked twice: once for the meaning, once to work out how strongly the author is actually claiming it.

Writing. The first draft is written in the mother tongue and translated, which produces sentences that are correct and wrong at the same time. Or it is written directly in English at a third of the speed, with the same idea circled three times because the one verb that would pin it down is not there when needed.

Review. The comment says "unclear" or "the argument is hard to follow". The author reads it as a verdict on the science. It is usually a verdict on the paragraph, and specifically on hedging and claiming: "will deliver" where the author wrote "could potentially", "we show" where the author wrote "it may be seen that". Reviewers rarely say this in those words, so the author fixes the wrong thing.

Speaking. The talk goes fine, because it was rehearsed. The question afterwards does not, because it was not. Silence in the questions session is read as having nothing to say, when it is the two-second lag of assembling a sentence under pressure in a second language. Skipping the conference next year removes the risk and the visibility with it.

Time. Every one of these adds hours, and hours are the currency of an early career. The same experiment, the same result, one fewer paper a year for a decade.

What actually helps, and what does not

Some of the advice given to non-native writers is useless and some of it is expensive. In rough order of what works:

Write in English from the first line. Translation doubles the work and produces a hybrid nobody speaks. Thinking in the target register from the start is slower for a month and faster forever after.

Learn the moves, not the words. Academic writing in any field runs on a small number of rhetorical moves. John Swales set them out in Genre Analysis: English in Academic and Research Settings (Cambridge University Press, 1990) as the "create a research space" model: establish a territory, establish a niche, occupy the niche. Three moves, described in 1990, still visible in the introduction of nearly every paper published this morning. A researcher who can name the move a paragraph is making can fix the paragraph. Vocabulary lists do not give you that.

Collect hedges and claims from your own subfield. Ten published papers in your exact area contain every sentence pattern you will ever need for a methods section, a limitations paragraph and a rebuttal. They are free, and they are already in your register.

Practise the question, not the talk. The talk is rehearsed; the question is where the tax is paid. Practise answering hostile questions out loud, under interruption, until the two-second lag closes. This is the one skill where solitary study does not work and something that pushes back is needed.

Treat reviewer comments as language signals. "Unclear" nearly always means "I had to reread". Find the sentence, find the hedge, decide what you are actually claiming, and say it once.

Professional editing. It helps the paper in hand, teaches nothing for the next one, and costs a tax on top of the tax. Use it for the submission that matters; do not mistake it for training.

Generic Business English or academic English courses. They teach the average of every field, which is nobody's field. A course on "writing for publication" that has never seen your subfield's papers will polish the paragraphs that were already fine.

Where practice fits

The through-line of everything above is that the cost is specific. It is not "English"; it is the English of your own subfield, under the specific pressures of review and the questions session, and it is closed by practice in that register, sustained long enough to change how you write on a Tuesday afternoon when nobody is watching.

That is the gap we built nicheAce: language to close. It starts with a conversation about your field and where you get stuck, and generates a fifteen-week journey from it: the vocabulary of your own area first, then the careful language of hedging and claiming, then the situations that decide things, such as answering a reviewer who doubts your method or defending a choice on a consortium call, with a role-play partner who does not accept "it just is". It is five days a week, five exercises a day, about fifteen minutes each, and it is honest about that. I built it because I paid this tax myself, in a different language, and found that the only thing that reduced it was practising in the register I actually needed rather than the one the courses offered.

None of that changes the history. English won, and it is not going to lose in any career that starts today. What can change is how much of your working life goes to paying for a decision that was made in 1919, in a room you were not in.