Some Laugh and Others Don't: How The Brain Interprets Sarcasm, Tone, and Hidden Meanings

Humans talk all the time, and understanding what has been said is not just about understanding words, but also about perceiving intentions, emotions, and social rules. This study showed that this ability is not a single area, but rather formed by three main components: understanding social norms and irony, capturing emotions through tone of voice, and using prior knowledge to draw conclusions. This discovery helps explain communication difficulties in some people and may contribute to the improvement of artificial intelligence.

Human language goes far beyond the literal meaning of words. In everyday life, in different cultures, people rarely say exactly what they mean. To understand a real conversation, it is necessary to interpret intentions, read the social context, perceive tone of voice, see facial expressions, and use prior knowledge about the world. When someone says "It's cold here," for example, they may not just be commenting on the temperature, but implicitly asking for a window to be closed. Autistic individuals are also known to interpret everything literally, without sarcasm, irony, or facial expressions.

This ability to understand what lies behind words is called the pragmatic use of language. It involves complex cognitive skills that allow us to place meanings that are not explicitly stated, understand irony, metaphors, jokes, and social intentions. Although essential for human communication, it is still not entirely clear how these skills are organized in the brain and mind.

Researchers have long debated whether understanding non-literal language depends on a single general skill or whether it is composed of several distinct skills that work together. Some people, for example, find it easy to understand irony but have difficulty perceiving emotions through tone of voice.

Some people are even able to interpret these emotional cues, but they also get confused by implicit and unspoken social norms. These observations suggest that the pragmatic and direct use of language may not be a single skill, but a set of different areas and pathways. To investigate this issue systematically, researchers conducted a long-term study with hundreds of participants, using an approach based on people's actual behavior when performing daily linguistic tasks.

In this study, a large group of participants performed a battery of tasks that assessed different aspects of understanding something said non-literally. Each participant spent about eight hours performing varied activities, including interpreting indirect phrases, understanding irony and sarcasm, perceiving emotions conveyed by voice intonation, and drawing conclusions from situations described only verbally.

It is important to emphasize that these tasks were chosen to reflect real-life communication situations, similar to those that occur in everyday life. Instead of analyzing each task in isolation, the researchers observed how people's performance related across different tasks; that is, whether those who performed well on one type of task also tended to perform well on another.

To analyze this complex data, the researchers used a statistical method that allowed the identification of hidden patterns, observing which tasks tend to "run together."

The central idea is that if several tasks require the same mental skill, people who perform well in one will likely also perform well in the others. In this way, it is possible to identify the existence of components associated with that task.

After ensuring that the results were not simply explained by the fact that more intelligent people perform better in everything, the analysis revealed that the tasks were organized into three large groups. The first component is related to understanding social conventions.

This involves knowing what is appropriate to say in different social contexts, understanding indirect requests, grasping hidden implications in a conversation, and recognizing irony. For example, understanding that someone who says "How punctual you are" might be being ironic depends heavily on this type of skill.

The second identified component is linked to an individual's ability to extract meaning from speech intonation. Intonation includes variations in tone of voice that indicate emotions, intentions, or contrasts in meaning. The same sentence can express joy, anger, doubt, or sarcasm, depending on how it is said. Therefore, text messages that appear offensive and incisive do not always reflect the reality that the sender intended to convey.

This component allows people to understand emotions and intentions even when words, in isolation, do not make them clear. This ability is particularly important in oral communication and can function relatively independently of social norms or logical reasoning.

The third component is related to causal reasoning based on world knowledge. It involves the ability to use general information about how the world works to interpret what someone means. For example, when listening to an incomplete story, a person can infer plausible causes and consequences, even if the ending of the story has not been explicitly mentioned.

This type of inference depends less on language itself and more on accumulated knowledge about everyday situations, cause-and-effect relationships, and expectations about human behavior.

To ensure that these results were not specific to a single group of people, the researchers repeated the study with a new, large, and independent sample. The same three components emerged again, with very similar patterns, reinforcing the conclusion that this structure reflects something fundamental about the organization of human communication.

Furthermore, the results proved robust, even using different methods, indicating that this is not a statistical artifact, but rather a consistent characteristic of cognitive functioning.

These findings have important implications. They suggest that the pragmatic use of language is cognitively simple in terms of structure, as it can be described by a few main components, but at the same time, it is functionally diverse. Each component may be associated with different areas of the brain, may develop at different rates during childhood and adolescence, and may be affected differently by neurological or genetic conditions, such as autism.

This helps explain why some people experience specific communication difficulties, as occurs in certain developmental disorders or after brain injuries, without other linguistic abilities necessarily being compromised.

Furthermore, these results are relevant to the development of artificial intelligence systems. Machines may be good at interpreting linguistic rules or statistical patterns, but they often fail to understand social intentions, irony, or emotions conveyed by voice.

The fact that the pragmatic use of language is composed of different components suggests that artificial systems need to address each of them separately to achieve communication closer to human communication.

This study, therefore, offers an initial map of the cognitive abilities that underpin human communication in the real world.

By identifying its main components, it paves the way for new behavioral research, neuroimaging studies, and computational models that can deepen our understanding of how we speak, understand, and infer meaning beyond words.

READ MORE:

Three distinct components of pragmatic language use: Social conventions, intonation, and world knowledge-based causal reasoning

Sammy Floyd, Olessia Jouravlev, Moshe Poliak, Zachary Mineroff, 

Edward Gibson, and Evelina Fedorenko 

PNAS. December 9, 2025. 122 (50) e2424400122

https://doi.org/10.1073/pnas.2424400122

Abstract:

Successful communication requires frequent inferences. Such inferences span a multitude of phenomena: from understanding metaphors, to detecting irony and getting jokes, to interpreting intonation patterns. Do all these inferences draw on a single underlying cognitive ability, or does our capacity for nonliteral language comprehension fractionate into dissociable components? Using an approach that has successfully uncovered structure in other domains of cognition, we examined covariation in behavioral performance on diverse nonliteral comprehension tasks across two large samples to search for shared and distinct components of pragmatic language use. In Experiment 1, n = 376 participants each completed an 8 h battery of 20 critical tasks. Controlling for general cognitive ability, an exploratory factor analysis revealed three clusters, which can be post hoc interpreted as corresponding to i) understanding social conventions (critical for phenomena such as indirect requests, conversational implicatures, and irony), ii) interpreting contrastive and emotional intonation patterns, and iii) making causal inferences based on world knowledge. This structure largely replicated in a new sample of n = 400 participants (Experiment 2, preregistered) and was robust to analytic choices. This research uncovers structure in the human communication toolkit and can inform our understanding of pragmatic difficulties in individuals with brain disorders. The hypotheses put forward here about the underlying cognitive abilities can now be evaluated in new behavioral studies, as well as using brain imaging and computational modeling, to continue deciphering the ontology of the component pieces of linguistic and nonverbal communication.