AI analyzed 7.9 million speeches and discovered something unexpected about humans

A new study looked at how word meanings shift and asked who actually changes their language when that happens. It found that adults of all ages tend to move together, with only a small delay between younger and older speakers.

The work centers on semantic change, which is the way a word’s meaning rises or falls in everyday use.

The headline result is simple and a bit surprising: older adults usually adopt new meanings within a few years of younger adults.

Human language and word meanings

Lead author Gaurav Kamath of McGill University (MU) and his team wanted to test a classic claim in linguistics.

Do older speakers keep speaking the way they learned in youth, or do they update their usage as the language around them changes.

The team focused on meaning, not sound or grammar. They asked whether shifts in the dominant sense of a word spread through all age groups in real time.

They also checked whether age differences seen at a single moment really reflect deep generational divides. Or if those differences fade as people of different ages respond to the same cultural moment.

Their question matters because many studies use cross-sectional snapshots. If adults keep changing within their own lifetimes, those snapshots can mislead researchers about how fast change is unfolding.

Testing a theory

The researchers analyzed 7.9 million U.S. Congressional speeches delivered from 1873 to 2010. They pulled the text from a public dataset that spans more than a century of congressional debate.

They matched each speech to the speaker’s age at the time of delivery using an open database. That join let them track how usage patterns varied across both year and age.

The team identified about 100 words that likely changed meaning over the 20th century. They then asked how the probability of using each sense rose or fell across time.

This setup provided an unusually strong test. It kept the venue constant, tracked thousands of named speakers, and covered nearly 140 years of public language.

What the models did

To detect which sense a word carried in context, the team used a masked language model (MLM) to predict context-appropriate replacements. They clustered those predictions into interpretable senses for each word.

They modeled how each sense rose or fell over calendar years, while also accounting for the age of the speaker. The core question was whether age added a consistent lag in adopting a newer sense.

They used generalized additive mixed models (GAMM) to estimate smooth trends over time while absorbing person-level variation. They then ran a Bayesian meta-analysis across all words to summarize the average age effect.

This approach let them separate two things. One is the overall tide of the times. The other is any extra delay tied to a speaker’s age.

What the results show

The main pattern was a shared shift across years, not a hard handoff from one generation to another. Older adults tended to lag younger adults by roughly two to three years on average.

In some cases, older speakers even led the change. For example, the geopolitical sense of satellite climbed during the Cold War and older members were often early adopters.

The authors also tracked prolific individuals who used the same word many times over their careers. Those speakers changed their usage within their own lifetimes, which matches the population pattern.

“In a nutshell, older people DO pick up new meanings of words,” said Kamath.

Why word meanings matter

The results challenge a long-standing tool known as apparent time, which infers change by comparing young and old speakers at one moment.

If older speakers keep updating their usage, apparent time can underestimate how much change is already underway.

The findings fit a zeitgeist perspective, where speakers of all ages respond to the social moment, not just to their birth cohort.

That contrast, discussed in detail by Fruehwald, helps frame when to expect generational replacement and when to expect shared movement across ages.

They also point to a practical takeaway for researchers. When possible, use historical data to test whether today’s age differences are stable or already closing.

For educators and communicators, there is another angle. Adults keep learning meanings throughout life, so terminologies in public life can shift quickly and broadly.

Limits and next steps

The dataset covers adult professionals in a formal setting. That leaves out teenagers and very young adults who often start changes in casual speech.

Congress is also not a social mirror of the country, especially in earlier decades. Women and minorities were underrepresented for long stretches of the record.

The method relies on models that guess a word’s sense from context, and such models can make errors in rare or ambiguous cases. That noise does not erase the main trend but it can blur fine details.

“The main result, that older speakers are highly adaptable to new word meanings, was itself a surprise,” said Kamath.

The team notes that expanding beyond North American English and adding adolescent speech are clear next steps.

The study is published in Proceedings of the National Academy of Sciences.

—–

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates. 

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–