The manifold costs of being a non-native English speaker in science

Picture a researcher in Kathmandu or Bogotá, sitting down to write a paper they already know how to write. The ideas are there. The data is there. The analysis is done. But now they have to put it all into a language that isn't theirs. Amano and colleagues actually measured what that costs in hours, and the number is larger than most people in science want to admit. The study surveyed nine hundred and eight researchers working in environmental sciences, recruited through seventy-one universities, twelve institutes, twenty-three societies, and direct contact with nearly five hundred individual researchers across eight countries. The sample skewed toward active, publishing scientists — a limitation the authors are upfront about — which means the true picture is likely worse than what the data show. That's because the people most severely affected by language barriers may have already left science entirely before anyone could survey them. The core question was simple and rarely asked: how much longer does it take to do the same scientific work when you have to do it in English rather than your first language? The answer depends on your country's English proficiency level and how early you are in your career, but the gaps are substantial at every stage. Start with reading. For a researcher who has published just one English-language paper — a proxy for being early in their career — Amano and colleagues estimate that someone from a low English-proficiency country needs, on average, about forty extra minutes to read and understand a single English-language article compared to a native English speaker. For researchers from moderate-proficiency countries, that gap is about twenty-one minutes per paper. Those sound like manageable numbers until you scale them. If a researcher reads two hundred articles in a year — not an unusual workload — the low-proficiency penalty adds up to roughly nineteen extra working days annually. The moderate-proficiency penalty is about ten extra days. That's two to four weeks of labor every year just to read the literature that native speakers absorb at baseline speed. Writing carries its own burden. Early-career researchers from moderate-proficiency countries spend a median of about fifty-one percent more time writing a first draft in English than native speakers do. For low-proficiency countries, the figure is roughly thirty percent more. These gaps narrow as researchers publish more, but they don't disappear. Then there's the gatekeeping that happens in peer review. Among researchers with only one English-language publication, about thirty-eight percent from both moderate- and low-proficiency countries report having had a paper rejected specifically because of English writing. For native English speakers, that figure is fourteen percent — a two-and-a-half-fold difference. And when it comes to being asked to improve English during revision, the gap is even starker: around forty-two percent of non-native speakers from both proficiency groups report this happening often or almost always, versus just three point four percent of native speakers. That's twelve times higher. The science isn't being evaluated on its merits alone. Now move from the desk to the conference hall, because that's where careers are built in ways that don't show up in citation counts. Amano and colleagues found that roughly thirty percent of early-career researchers from non-native English backgrounds in high-income countries — their sample combined Japanese and Spanish nationals — often or always decide not to attend an English-language conference because they lack confidence in their English. About half often or always avoid giving oral presentations for the same reason. Even when non-native speakers do present, the preparation cost is enormous. Researchers from moderate-proficiency countries spend a median of nearly ninety-four percent more time preparing and practicing an oral presentation in English compared to native speakers. That's almost double the time. For low-proficiency countries, the figure is thirty-eight percent more. And unlike the reading and writing gaps, this presentation burden does not shrink with career stage. It persists. More than sixty-five percent of early-career researchers from low-proficiency countries report that they often or always find it difficult to explain their work confidently during a presentation in English. These are people who understand their research as well as anyone. The language itself becomes an obstacle to being heard. Amano and colleagues are explicit about the downstream consequences. Missing conferences means missing the networks that generate collaborations, recommendations, and invitations. Being rejected more often in peer review means a slower publication record, which means fewer grants, fewer promotions, and less visibility. The authors note that some apparent narrowing of the gap late in careers may reflect survivorship bias — the researchers who made it through are precisely those who managed to overcome these barriers — which means the population who didn't make it simply isn't in the data. A word on the study's methods, because they matter for how much weight to give these numbers. The team used separate statistical models suited to each type of outcome: negative binomial generalized linear models for time measurements, binomial models for yes or no outcomes, and cumulative link models for ordered frequency responses. They tested interactions between English proficiency level and publication count to check whether gaps close at higher career stages, and included publication count as a covariate after finding that age, years in research, and gender all correlated too tightly to separate cleanly. The survey itself was deployed in multiple languages — Japanese, Nepali, Portuguese, Spanish, and Ukrainian among them. A monolingual survey about language barriers would have been self-defeating. The main limitation the authors flag is recruitment bias. The sample draws from active researchers who are already engaged enough to respond, excluding people who never published a first-authored English paper and people who left research entirely. The magnitude of the barriers measured here is probably a floor, not a ceiling. So what can actually be done? Amano and colleagues propose interventions at every level of the research system, and the proposals are specific. Supervisors and institutions should acknowledge extra time needs, provide training, establish buddy systems, and fund professional English editing and translation — especially for researchers from lower-income regions and early-career scientists. Journals should promote non-English abstracts, strengthen double-blind review to reduce language-based bias, and consider explicitly allowing AI-assisted proofreading. Funders should create grant schemes that cover editing and translation costs and reward plans to disseminate research in multiple languages. Conferences should offer editing support for presentations and encourage multilingual formats. The paper names ChatGPT and DeepL specifically as tools that could reduce the time burden on non-native speakers, and notes that machine translation is currently underused — suggesting that making these tools free or affordable and explicitly permitted could close part of the gap without requiring structural change. But AI tools are a supplement, not a solution. The structural changes — funding, review processes, conference culture — are where the leverage is. Non-native English speakers represent roughly ninety-five percent of the world's population. The extra effort Amano and colleagues quantify is not a small inefficiency at the margins of science. It is a systematic tax levied most heavily on the people with the least slack in their careers, in the countries with the fewest resources to absorb it. What the survey captures is the effort that went into translation instead of discovery — the hours spent rewording instead of thinking, the conferences not attended, the presentations not given, the papers not written because the effort of writing them in English made the cost too high. That's the untapped potential the authors describe. Not an abstraction — a countable number of working days, a measurable gap in presentation rates, a documented difference in rejection frequencies. The data are there. The question is whether the institutions that run science are willing to read them. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.