HOW would you rank “important” languages? If asked to rattle them off, many people start with English, but after that are reluctant to go further. Important how, they ask. One approach would be to look at people and money: surely a language is important if it is spoken by lots of people, in countries with great wealth (and presumably, therefore, power).

But in December came a new approach. A group of scholars* approached the task by first looking at how languages are connected to one another, rather than viewing them in isolation. They then decided to see if this was a good predictor of how many famous people spoke a given language. If a language is well connected to others (a “hub” language with many bilinguals), its speakers will tend to be famous. And the names of the connected languages turns out to be rather interesting.

To find links between languages, the researchers created a “global language network” (GLN) three different ways (see graphic). One was Wikipedia editors: a bilingual Wikipedian who edits articles in both Arabic and English counts as strengthening the bond between Arabic and English. The second was Twitter: users who had tweeted at least six full sentences in a second language were treated as strengthening the bond between those two languages. The third was a more formal, old-fashioned metric: book translation. UNESCO, the United Nations’ cultural organisation, keeps a database of translated books, and each of the 2.2m translations was counted as a bond strengthening the two languages. (These bonds, of course, are asymmetrical: some languages have more books translated out of them than into them and vice-versa.)

View full image here

The resulting networks are striking in many ways. English is central to all of them. But with many other languages, their connectivity has little to do with their home country’s modern power. Take the network implied by book-translation. The data come from 1979-2011, and so Russian is an important node in the network. Not only were books translated between Russian and other languages of the former Soviet Union (Armenian, Kirgiz and Latvian, say), but Russian is significantly connected to languages from South and South-East Asia and the Middle East. The contrast with Wikipedia and Twitter, which skew much more modern, is striking: Russian suddenly becomes a peripheral node.

Chinese, too, is peripheral in the authors’ networks. In the book-translation network, the world’s most spoken language is isolated, connected mainly to other Chinese languages plus a few in South-East Asia, notably Vietnamese. This may make sense given the time-frame of the book-translation database, skewed to decades before China’s spectacular rise. But Chinese is also a bit-player on Twitter, as a result of the popularity of Sina Weibo, a competing Twitter-like service, in China. The same is true of Wikipedia: here Chinese is somewhat better connected, but it is still much less than its size or GDP would predict, possibly thanks to the existence of a Wikipedia-like Baidu Baike collaborative encyclopedia.

The upshot is clear: big languages are not necessarily global, and vice-versa. Arabic and Hindi—two other languages with hundreds of millions of speakers—are as peripheral as Chinese and Russian. The big nodes in the networks besides English are predictable: French, Spanish and German, especially. The first two were successfully flung far and wide by colonialism. German has centuries of prestige in science, philosophy and literature, despite the failures of Germany’s colonial efforts.

But these results must be handled with care, the authors note. The paper says nothing about the inherent qualities of any language, or the cleverness of its speakers. César Hidalgo, one of the authors, notes that the paper is really about elites. Bilinguals with time to edit Wikipedia are not typical people, nor are book translators (or even bilingual Twitter users). But they do play an outsized role in the transmission of culture across borders. The main finding of the paper is that people are more likely to become globally famous (as measured, in part, among people with Wikipedia entries in at least 25 languages) if they speak one of the most networked languages. The world's most brilliant person may be a speaker of Hmong or Nahuatl, but the road to fame leads through other languages.

There is a “what else did you expect?” nature to the outsized role of Western languages in three Western-born products: the mass book-publishing industry, Wikipedia and Twitter. But the results are still meaningful: Twitter really is globally important (just ask the Iranian mullahs or the former president of Tunisia). Sina Weibo and VK (Russia’s Facebook) are not. Their homemade nature may be a point of pride—it certainly keeps censors happy—but it also means cutting down a country’s cultural influence.

For the language learner, the networked nature of languages poses an interesting dilemma: should you learn a language popular among global elites? The tradition of learning French still looks like a good bet here. Or should you learn a language whose number of globetrotting bilinguals is small relative to its importance? Supply and demand says that this will be valuable. This is the real case for learning Chinese: not because it is the rising global language, but because it isn't—at least certainly not yet. Given China’s huge role in the global economy, the number of outsiders fluent in Chinese is still far too few.

“Links that speak: The global language network and its association with global fame,” by Shahar Ronen, Bruno Gonçalves, Kevin Z. Hu, Allesandro Vespignani, Steven Pinker and César Hidalgo, Proceedings of the National Academy of Sciences, December 2014