New Study Reveals How Language Meaning Is Shaped by Power, Danger and Structure

Language influences how people think, feel and relate to the world around them. For decades, scientists thought they had a reliable method for measuring the meaning of words.

That approach informed work in psychology, linguistics and artificial intelligence. However, emerging evidence indicates that this widely accepted model may not describe how meaning truly operates.

A recent study in Science Advances presents a different view. It argues that the organisation of meaning differs from what researchers had previously believed.

Rather than the old model, the findings point towards a deeper, more intuitive system that mirrors the way people genuinely use language.

The early model of meaning

During the middle of the 20th century, researchers sought to chart meaning through straightforward dimensions.

Charles Osgood and colleagues asked participants to assess words using opposing pairs, including good versus bad and strong versus weak. From this analysis, they identified three central dimensions: evaluation, potency and activity.

These concepts later developed into the valence arousal dominance model. Valence assessed whether a word seemed positive or negative, while arousal measured whether it appeared exciting or calm.

Dominance represented authority or power. The framework subsequently became common across a wide range of disciplines.

How small datasets shaped ideas

The initial studies drew on extremely limited datasets. Some examined only a few dozen words, with ratings supplied by university students.

Even subsequent research increased the scope only modestly, covering roughly a thousand words.

At the time, these early constraints were not viewed as problematic.

Researchers presumed that the model would remain valid at greater scale, but lacked the data needed to test that belief thoroughly.

Reassessing the structure of meaning

The situation changed in recent years. The large NRC VAD lexicon contained ratings for more than 20,000 English words, enabling a much more precise test of the established model.

Peter Sheridan Dodds and his University of Vermont team used the dataset to revisit the structure of meaning.

They anticipated that the conventional framework would be confirmed. What they discovered instead was unexpected.

The dimensions are not independent

The three dimensions were intended to function independently. Yet the analysis identified strong links between them.

Words judged to be positive frequently also seemed powerful. Arousal and dominance likewise changed together in ways that should not have occurred if the model was accurate.

This showed that the model's basis was unstable. Its axes were not as distinctly separate as researchers had once thought.

Visual maps reveal patterns

To investigate the data more closely, the team produced visual maps known as ousiograms. The maps located real words in a space determined by their ratings.

Distinct clusters emerged. Terms such as success and greatness appeared together, whereas toxic and earthquake occupied separate clusters.

The overall form did not match the predicted axes. Instead, the patterns indicated that meaning had a different organisation.

A new framework for meaning

The researchers then performed a new analysis without depending on the former assumptions. It uncovered three alternative dimensions.

They labelled them goodness, aggression and structure.

Goodness distinguished positive terms from negative ones. Aggression represented calm ideas versus intense or violent ones, while structure captured order in contrast with unpredictability.

However, the team then identified something more striking still.

A rotated view makes meaning clearer

Rotating the primary plane formed by these dimensions produced a more understandable picture. Two intuitive axes became apparent: power and danger. Structure remained the third dimension.

This produced a new model, the power danger structure framework. It provided a more natural means of interpreting meaning.

Study co-author Julia Zimmerman explained that power, danger and structure are relevant to every person that’s ever lived.

Real-world language reveals a bias

The team next applied the model to real-world writing. Their material included books, news reports, social media posts and other texts.

One result was consistent across every source: people showed a strong preference for words associated with safety rather than danger.

“"This is a big observation that comes out of this work," said Dodds. "Expressions of safety are crucial to all language."”

The same tendency was present in each dataset, irrespective of its period or style.

Safety influences positivity

Previous research has found that language generally leans positive, a tendency called the Pollyanna principle. The new study offers a more fundamental account.

This apparent positivity may stem from a stronger preference for safety. Words perceived as safe also tend to seem positive and calm, giving the impression that separate biases are at work.

Instead, both may be driven by a single underlying factor.

Unlike danger, the power dimension displayed no pronounced bias. Terms were distributed evenly across weak and powerful meanings.

The difference underlines safety's distinctive place in language. Rather than being simply one influence among many, it affects communication at a fundamental level.

Characters follow the same pattern

The researchers also examined fictional characters, analysing the traits of two thousand characters from numerous stories.

The same three dimensions emerged once more. Characters extended from weak to powerful, safe to dangerous, and structured to unpredictable.

This consistency between distinct sources of data indicates that the model may reflect something basic in human thought.

Tracing stories through meaning

The team created a tool named an ousiometer, which follows a narrative's movement through power, danger and structure.

Used on Les Misérables, it revealed clear changes in tension and safety across the narrative.

High points of danger corresponded with dramatic events, while safer periods identified quieter passages.

The method creates new opportunities for examining storytelling.

Researchers challenge the old model

The results call a long-established scientific assumption into question. Although the earlier model was useful, it does not completely fit data at large scale.

The new framework provides a more lucid perspective and is more closely connected to how people employ language in everyday life.

It further indicates that meaning is formed by fundamental human concerns, particularly the requirement for safety.

Meaning continues to evolve

Science often advances by revisiting established beliefs. This research demonstrates how improved data can transform understanding.

Meaning's structure is not fixed; it shifts as researchers acquire new tools and insights.

By examining language at scale, the work exposes a simpler reality: people describe the world through power, danger and structure. Underlying all of this, a persistent pull towards safety directs the words they select.