back to top

The Evolution Of Language Involves A Lot Of Random Shit, Says This Study

But we may have started saying "dove" instead of "dived" because of the rise of the car, so there's that.

Posted on
Cherissa Dukelow / CC

The grammar of negating a sentence: "Ic ne secge" (Beowulf, c. 900); "Ic ne sege noht" (the Ormulum, c. 1100); "I seye not" (Chaucer, c. 1400); "I doe not say" (Shakespeare, c. 1600); "I don't say" (Virginia Woolf, c. 1900)

Languages, like living things, evolve. And a new study has found that, just like biological evolution, linguistic evolution involves a lot of total fluke.

Researchers at the University of Pennsylvania, including both linguists and biologists, have studied a huge body of written English going back to the 11th century, and found evidence for a process analogous to "genetic drift". Genetic drift, in biological evolution, is when traits arise and disappear not because they're evolutionarily helpful, but because of blind chance. The study is published in the journal Nature.

The researchers looked at the rise of the "periphrastic do" – the use of "do" as an auxiliary verb, in sentences such as "Do you think this?" or "Don't say that," instead of the older form "Think you this?" or "Say that not," over the past 1,000 years. They also examined more recent changes in certain verbs with both "regular" and "irregular" past-tense forms – that is, verbs such as "dive", which have both a normal past tense with an -ed on the end ("dived"), and an irregular one which doesn't ("dove").

Richard Blythe, a professor of physics at the University of Edinburgh who does statistical modelling of language change and did not work on this study, told BuzzFeed News that he was glad the study had been published. "Lots of people have talked about the possibility of different evolutionary processes in language change," he said, but few have actually looked at it with lots of data. "So this study gives us some new quantitative information to get our teeth into."

They wanted to see whether these changes were driven by some force, or whether they were random. To see that, they looked at how the use of the words changed on a decade-by-decade basis. Steady movement from one form to another – say, from regular to irregular – was evidence that something was driving the change; a meandering walk which shifted back and forth was evidence that it might be random. "We did a statistical test," said Joshua Plotkin, a professor of biology at the University of Pennsylvania and the lead author on the study. "What is the likelihood of seeing this change if it's completely random? And if the pattern is really unlikely, then we say it must have been selection."

Of the 36 irregular-regular verb pairs, just six were found to show clear evidence of selection: lighted/lit, waked/woke, sneaked/snuck, dived/dove, wove/weaved, and smelt-smelled.

The first four of those six – lighted/lit, waked/woke, sneaked/snuck, and dived/dove – were found to be changing from regular to irregular. Plotkin said that that is not what traditional linguistic theory would predict.

It's easy to explain why verbs, especially rare verbs, might become more regular – if you don't hear it very often, you might not learn the irregular form. Small children do this regularly, and will say "I drinked the drink" or "I holded your hand", because they know the rule but not the exceptions. But it is harder, says Plotkin, to explain change away from the regular form. "It's a bit of a surprise," he said. "Normally you'd expect change toward regularisation."

In the case of "dove", though, they think they have an explanation. They noticed that the irregular verb rose alongside the use of the verb "to drive", and its past tense "drove". That happened at the same time as the car started to become common in the early 20th century. "The rhyming pattern '-ive/-ove' became more common in the literature with the emergence of cars," said Plotkin.

More generally, they found that irregular forms were more likely to be favoured if other, similar-sounding irregular verbs were also becoming common. For example, "quitted" lost out to "quit" at the same time as "split" became more common in the last century.

The story of the rise of the "periphrastic 'do'" is a bit more complicated, says Plotkin. There are three main areas it's used in: questions ("Do you think this?" as opposed to "Think you this?"), declarative statements ("I don't believe it," as opposed to "I believe it not"), and imperative statements ("Don't say that!" as opposed to "Say that not!").

"We found that in the question sentences, its rise was attributable to chance alone," he said. "There are a lot of theories about how it came into English, but once it got there its rise can be explained by random chance.

"But once it became common in question sentences, after that it rose very quickly in declarative or imperative statements. So the story is that a random process helped the auxiliary verb 'do' drift in, but then set the stage for selection."

Tom Chivers is a science writer for BuzzFeed and is based in London.

Contact Tom Chivers at

Got a confidential tip? Submit it here.