"Immediately after an event, writers often use elaborate descriptions, such as ‘a plane crashed into a building.’ Over time, this becomes shorter and more implicit, and the reader only needs to read ‘9/11’ to understand,” explains linguist Levi Remijnse. “I wanted to understand which factors determine how that framing evolves.”
Real-world events
The motivation for Remijnse’s research was that existing linguistic models often focus solely on words or meanings, rather than the actual events being written about. “I combined those elements by studying how shared knowledge (common ground) and time influence the way we describe events. I demonstrate that language shifts in step with what readers already know: the more familiar an event is, the more concise and self-evident the description becomes,” says Remijnse.
Shaping the narrative
The study also shows that writers shape the narrative through their word choices. They highlight certain aspects and omit others. “My main conclusion is that language does not merely reflect facts, it also helps shape our collective perspective on them. In other words, language doesn’t just tell the story; it helps build it.”
Insight
Language therefore evolves alongside our knowledge and beliefs. An insight that, according to Remijnse, is important for journalists, communication specialists, and researchers. “They can better understand how words subtly influence how people experience events. Think, for example, of the difference in impact between ‘climate change’ and ‘climate crisis’. By using framing more consciously, media can report more balanced stories and policymakers can better anticipate how their messages are received by the public.”
A sense of context
Artificial intelligence, including language models, can also benefit from these insights — enabling them to interpret or generate texts with greater sensitivity to context. “The relevance is immediate: in a time where news, social media, and AI language bots all help shape our worldview, this research helps us understand how language constructs narratives and how we can steer those narratives more mindfully,” Remijnse notes.
Terrorist attacks and disasters
Remijnse’s research combines linguistic analysis, computational modelling, and human interpretation. It is based on a large collection of news articles about real events, such as terrorist attacks and disasters — for instance, the MH17 crash or the COVID curfew riots. “Using computer programmes, I analysed how these events were framed over time: which words and sentence structures writers used, and how these changed as the news aged.”
Special system
Together with a team of linguists and students, Remijnse developed a specialised system to annotate which words referred to which events, and what kind of frame or perspective they evoked. “This allowed me to trace step by step how language shifts from detailed description to shorter, more conventional references and how these descriptions vary in the frames they call to mind.”