Think you could safely tell the difference between Russian, Hindi, Spanish and Telugu? The Great Language Game is the place to test that theory.
The online guessing game has yielded a ton of data for academics to sift through, offering insight into just how good we are at telling Icelandic from Tongan.
In the game, developed by programmer Lars Yencken in 2013, players are asked to listen to an audio clip and select its language match. Hedvig Skirgård, a PhD student at the ARC Centre of Excellence for the Dynamics of Language in Australia, has poured over 15 million guesses from the game, culled from 400 audio recordings of 78 languages.
She wanted to discover whether there was any pattern to incorrect guesses. The results have now been published in the academic journal, PLOS.
As it turns out, the game yielded some unexpected results, not least that geographic proximity between languages was often more important to an incorrect guess than a shared genealogy.
Take Romanian and Bulgarian, which were sometimes confused by players. The two countries share a border, but their languages are not from the same linguistic branch. "Romanian is actually a romance language, so it's correlated to Latin and Italian," she explained. "While Bulgarian is a Slavic language, so it's more related to Russian and Ukrainian and Slovenian."
All up, the two tongues most likely to be confused were the Indian languages Punjabi and Kannada, while the two least likely were French and Vietnamese.
The paper's authors also examined the impact of a language's fame and economic power using how often it appeared in Google Books and the gross domestic product of its home country, respectively. Ultimately, a language with "global fame" in the game was more likely to be recognised accurately.
"The frequency in Google Books was the stronger correlation," she explained. "In the case of the Slavic languages, Russian is more well known. People may not know that Slovenian is a Slavic language."
In other words, players hearing something Slavic often just guessed Russian, likely because the country is more talked and written about (and lately, more often in the news).
Of course, Skirgård emphasised that as a linguistic experiment, the game was far from perfect.
Ultimately, a language with "global fame" in the game was most likely to be recognised accurately.
Audio recording quality could have influenced people's guesses. Not to mention, there are up to 7,000 languages in the world and the team had results for only 78.
These issues might also affect country accuracy rankings. Australia ranked 35th out of all countries that had at least 5,000 guesses each, while the U.S. came in at 61st.
To give Americans a break, language selection in the game may have had an impact. "There are languages like Tongan, Maori and Samoan, where I expect those immigrant groups are not as visible in the U.S. as they are in Australia and New Zealand," she suggested.
It must also be remembered that people playing the game may be self-selecting language fans. "All we know about players is they know enough English to navigate the game, where their IP address is, and that they hang around probably the nerdier parts of the internet," Skirgård laughed. "There is a bit of a selection bias."
Still, the game yielded some bizarre mixups. Danish, Hungarian and Turkish, for instance, were between five and six times more likely to be confused for Vietnamese than the other way around.
Skirgård said she couldn't explain it. "Either it's a fluke of players, we don't know why, or there is something with those audio clips. That one was very surprising," she added. "It may be they share something we linguists have not discovered yet."
To further test language selection, Skirgård and her team have developed a new game, Linquest. If you fancy yourself a bit of a polymath, log on and have a go.
