Read our COVID-19 research and news.

William Shakespeare’s first folio at the Bodleian Library at the University of Oxford in the United Kingdom. Many forms of Shakespearean English are used today—but many have changed. 

Ben Sutherland/Flickr

How the English language has evolved like a living creature

Some linguists think of language as a living thing: It grows and changes, and every time a child learns it, the language reproduces itself. Now, a team of researchers is using the analogy of evolution to explain language change, arguing that key factors in biological evolution—like natural selection and genetic drift—have parallels in how languages change over time. And it turns out that the random changes, known as “drift” in biology, may have played an outsized role in the evolution of the English language.

Historians of English have long acknowledged that social and cognitive factors shape language over time. For example, languages lose irregular verb conjugations or other word forms that are hard to remember. And certain words or pronunciations get used because they are associated with people who have status and power—think about how new arrivals adopt the local accent in order to fit in. These pressures on language are based on concrete factors, similar to the biological pressures of natural selection.

But that explanation didn’t satisfy University of Pennsylvania (UPenn) evolutionary biologist Joshua Plotkin. He was puzzled by oddities such as a growing preference for the word “clarity” over its synonym “clearness.” According to standard linguistic theory, “clearness” should be more common because adding “-ness” is an easy-to-remember rule for making a noun out of an adjective. But that’s not what happened in English. “As an outsider,” Plotkin says, “this increase seemed at odds with the notion that language … regularize[s] over time.” So he decided to roll up his sleeves and apply some theories from evolutionary biology.

With another evolutionary biologist and two linguists from UPenn, he analyzed three databases of historical English together containing more than 400 million words and ranging from 1100 C.E. to the 21st century. The researchers used statistical methods from population genetics to analyze three well-known changes in the English language: how past-tense verbs in American English have taken the “-ed” ending, (as when “spilt” became “spilled”), how the word “do” became an auxiliary verb in Early Modern English (as in “Did you sing?”), and how negative sentences were made in Old to Early Modern English.

They found that selection was the likely cause of how negative sentence structures changed over time (like how the Old English “Ic ne secge” became the Early Modern English “I say not”). But the two other changes were likely the results of random drift, they write today in a letter published in Nature. That’s because, rather than having an even rate of change, the frequencies of alternative forms changed in fits and starts—jagged fluctuations that were obvious in the data set. When it came to the verbs, they found that drift’s influence was stronger when the verb was less frequent. Only six past tense changes in their data set, such as “lighted” to “lit,” were deemed to have changed for purposeful reasons, such as being easier to learn and use.

The paper is “extremely exciting,” says Erez Lieberman Aiden, a geneticist at Baylor College of Medicine and a computer scientist at Rice University, both in Houston, Texas, who has also applied quantitative methods to massive language data sets. “Any sophisticated model of evolutionary change needs to deal with the balance of selection and drift, and this one is sophisticated because it tries to dissect the principles that drive the change.”

But to other researchers, the role of randomness in language is intuitively obvious. “Every single speaker on Earth will have their own specific linguistic variants,” says Andreea Calude, a linguist at the University of Waikato in Hamilton, New Zealand. “This variation is sometimes driven by selection, but at other times, we like to choose our own options from the linguistic buffet available to us.”

Plotkin hopes that linguists will find the notion of drift and his statistical tests to be useful, because they allow researchers to study the patterns and timing of change in a single language rather than having to compare languages. “We’re not saying that pure drift is the only thing happening,” he says, “but rather that drift is often involved, and we shouldn’t rule it out.”