Who drives change in languages? The young or the old? Most people would probably say that it’s the former who lead the way, while the latter stay put. However, a new large-scale study suggests the picture is more complicated, especially when it comes to changes in word meaning.
The study
Researchers Kamath, Yang, Reddy, Sonderegger & Card analysed over 7.9 million speeches from the US Congressional Record (1873–2010). Crucially, they knew the age of every speaker, allowing them to look at how different age groups used the same words across nearly 140 years.
From an initial set of 100+ words likely to have undergone semantic change, they used [1] masked language model–based word sense induction to determine which meanings were present at different times. For example, “articles” once meant physical objects far more often, before the journalistic sense took over. They tracked shifts like this for 163 word-senses.
Finally, they modelled meaning change using [2] generalised additive mixed models, estimating how much a speaker’s age influenced the likelihood of using a newer sense.
The findings
Age matters… but only a little. The average age effect means an older speaker lags behind a younger one by ‘only’ about 3 years in adopting a new sense. Older adults do catch up—very quickly.
Changes are rarely generational. Meaning shifts didn’t wait for older speakers to “die out”. Instead, adults of all ages tended to adopt new meanings within a few years. Word meaning change is best described as a zeitgeist process, not a generation-by-generation handover.
Sometimes older speakers even led the change. A minority of cases (such as the geopolitical sense of satellite) showed older speakers adopting the new sense slightly earlier.
Individuals changed within their own lifetimes. Speakers who appeared repeatedly in the corpus altered their usage as meanings shifted around them.
This aligns with research showing that adults retain flexibility in vocabulary acquisition throughout the lifespan, even if phonology & morpho-syntax are harder to change past adolescence. This echoes arguments by Nevalainen, Raumolin-Brunberg & Manninen (2011) and by Sankoff (2006), who note that lexical & syntactic features tend to show more intra-speaker flexibility than phonological ones, partly because they carry less social evaluation.
If you think of contemporary English: when cloud gained a computing sense, or viral shifted from pathology to social media, adults didn’t wait for Gen Z to replace them. They adapted as the world changed.
Teacher Takeaways?
- When teaching vocabulary, don’t assume adults are “set in their ways”. They can adapt to new meanings as readily as younger learners.
- Highlight semantic flexibility: activities that compare older & newer senses (e.g. stream, follow, tablet) can build learners’ awareness of real-world change.
- Provide rich contextual examples—semantic shifts are often driven by cultural innovation, so situating words in changing social landscapes speeds comprehension.
How do you help learners stay attuned to shifting word meanings in English?
[1] A masked language model (like BERT) predicts a missing word in a sentence. If you mask a target word in many different contexts, the model proposes likely replacements based on how it interprets each context.
By clustering these predicted replacements, you can reveal the different senses the word takes.
For example, if bank is masked, contexts about money lead the model to predict credit union or lender, while river contexts lead it to predict shore or riverbank. Grouping these predictions lets researchers infer distinct word meanings without manually annotating them.
[2] These are statistical models that let researchers examine how an outcome changes smoothly over time (or other continuous variables) while also accounting for differences between individuals or groups.
Leave a Reply