From ‘pop’ to ‘soda’, language may obey the law of physics

Words do not spread at random. They cluster, hold their ground, and sometimes surge across a country with surprising speed.

That is the idea behind a new model from the University of Portsmouth. Statistical physicist James Burridge has built a system that tries to predict how language changes over time. His work treats speech a bit like a physical system. In fact, he borrows tools used to study magnets, fluids, and other large collections of interacting parts. Instead of particles, the moving pieces are people. The things shifting across space are words, pronunciations, and dialect forms.

“Just as meteorologists use mathematical models to forecast tomorrow’s weather, the same kind of thinking can be applied to language,” Burridge said.

The comparison is not perfect. In his telling, language behaves less like the atmosphere and more like matter organizing itself into patterns.

“Where you are affects how you speak and if you map how people use certain words, you see clear geographic patterns, just like a weather map. However, the physics of language is closer to crystals and magnets than the atmosphere.”

The research aims to help explain why some local expressions remain trapped in one area. Others push outward and become dominant across a much larger region.

To test the model, Burridge used large dialect surveys from the Cambridge Online Survey of World Englishes. This survey was created by Bert Vaux, a linguist at the University of Cambridge. The dataset includes close to 100,000 responses and tracks where speakers grew up. It also records what words they use and when they were born.

That let Burridge look at changes unfolding across decades in the United States.

One example involves the names Americans use for a woodlouse, the small crustacean that curls into a ball when touched. In 1950, “roly-poly” was mostly limited to a relatively small group of speakers in the South. By 1995, that same term had spread across much of the country and became nearly universal in many places.

Another case looked at soft drinks, tracing where “soda” gained ground against other terms such as “pop.”

The new system was able to reproduce sweeping changes like these by treating language as a dynamic field spread across space and time. It assumes speakers influence one another locally. Furthermore, popular forms can gain an advantage through conformity. Outside pressures can also push one variant ahead of another. Burridge also included what he calls a bias field. This is a way to capture forces that are harder to pin down directly, such as prestige, standardization, or other historical pressures that favor one form.

The result is a model that does more than describe where words are. It tries to show how they got there, and where they may go next.

The research also offers a way to explain stubborn holdouts.

In one of Burridge’s earlier studies, the word “splinter” spread north from southern England and became standard through most of the country. But it did not fully displace “spelk” around Newcastle in the far northeast.

“Splinter is used across almost all of England, except around Newcastle, where people still say spelk,” Burridge said. “Although Newcastle itself is densely populated, it is surrounded by more sparsely populated areas, which helps the local form hold its ground and prevents splinter from taking over.”

That pattern matters because it points to something larger than individual choice. Dense populations, local isolation, regional clustering, and social conformity can all shape whether a word keeps spreading or stalls at an invisible border.

The study argues that those patterns can be understood with some of the same mathematics used to describe physical boundaries. In the model, linguistic borders can behave a little like the surfaces of bubbles, smoothing and shifting over time. Population gradients may also help explain why some pockets of speech persist, even when they are surrounded by competing forms.

For Burridge, that is where the work becomes especially interesting.

“Language change can seem mysterious, but my research argues that as well as being driven by individual human behaviour it may also obey some of the same broad rules that govern physical systems like magnets, bubbles, and fluids.”

He added: “My research suggests that language may be much more law-like than it first appears. Beneath the creativity and messiness of human speech, there may be hidden statistical forces shaping how we all end up talking.”

The model does not promise crystal-ball certainty.

A central result is that language appears to have a natural prediction horizon, a period during which forecasts remain useful before uncertainty grows too large. Burridge compares that limit to weather prediction. This becomes less reliable the farther out it tries to look. In his analysis, the useful half-life of a bias trend was about 17 years.

Using that framework, the model projected that “roly-poly” would continue expanding in the United States and could become dominant across nearly the whole country. This would occur aside from a few holdout regions, by around 2040. The system also generated future maps for “soda.”

Still, the study builds in caution. Burridge relies on the “apparent time” principle, which uses the speech of adults as an approximate record of the language community they grew up in. The research notes that this has limits, because adult speech is not perfectly fixed. The model also works with binary choices, such as soda versus not-soda. Even though real language often involves several competing forms at once.

The analysis further acknowledges that language does not change in a vacuum. Politics, technology, culture, and social life can all push speech in new directions. Some of those outside forces may be hard to anticipate. The current version also assumes a static population distribution and mostly short-range geographic interactions.

Even with those caveats, Burridge sees the work as an important step.

“For physicists like me, this is particularly exciting, as it suggests that the elegant tools of statistical field theory may help explain not just the natural world, but patterns in human communication as well.”

This work could give linguists a stronger way to connect large survey datasets with models that actually forecast change, rather than only describe it after the fact. It may also help sociologists study how culture moves through regions and generations.

Beyond academia, systems that depend on tracking human language, including technologies built around speech and text, could benefit from better ways to anticipate how usage shifts over time.

The larger lesson is simpler: the words people choose may feel personal and unpredictable, yet across whole populations they can still follow patterns that mathematics can detect.

Research findings are available online in the journal Physical Review E.

The original story "From 'pop' to 'soda', language may obey the law of physics" is published in The Brighter Side of News.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.