> This point is much easier to sustain if the speaker grew up speaking English, but the majority of scientists working today are actually not native English spe

> This point is much easier to sustain if the speaker grew up speaking English, but the majority of scientists working today are actually not native English speakers. When you consider the time spent by them on language-learning, the English-language conquest is not more efficient than polyglot science – it is just differently inefficient. There’s still a lot of language‑learning and translation going on, it’s just not happening in the United Kingdom, or Australia, or the United States. The bump under the rug has been moved, not smoothed out. That's not true, and this is why you see hub-and-spoke models everywhere. Even if no one spoke the language natively (like, oh say, Latin?), it would still be efficient to have a single standard language so you only have to learn 1 rather than _n_. In the polyglot scenario, you still have to learn (at least) another language, but now each additional language only unlocks 1/n% instead of 'all of it'...