Recent research published in journals psychophysiology We provide evidence that the human brain processes taboo words in a completely unique way compared to regular negative or neutral language. This study suggests that the unique brain patterns triggered by these socially inappropriate words remain detectable even when people actively try to regulate their emotional responses. These findings help explain how social rules and emotional meaning are deeply intertwined in our neural wiring.
Language serves as a primary tool for communicating emotional experiences in everyday life. Emotional words tend to evoke strong reactions, and taboo words represent a very special category of such words. People frequently use swear words to express frustration, relieve pain, or increase the impact of a message. The authors of the new study wanted to see if specific neural patterns could reveal how people process and manage emotional information.
“One of the main goals of affective neuroscience is to understand whether it is possible to objectively identify emotional states from brain activity,” explains Parisa Ahmadi Gomroodi, a researcher at the University of Trento in Italy, and Professor Alessandro Grecucci, director of the Laboratory of Clinical Affective Neuroscience at the University of Bari in Italy. “Language provides an ideal model because it can reliably evoke a variety of emotional responses while remaining highly experimentally controlled.”
Scientists were interested in testing whether modern machine learning approaches could detect subtle differences in how the brain processes emotion-related language. They focused especially on taboo words that have a special place in the language. The researchers also wanted to see whether these neural signatures could provide a window into the broader mechanisms underlying emotional experience and regulation.
“At the neural level, emotional word categories were associated with distinct electrophysiological features across multiple stages of processing,” the authors noted. “Differences appeared in early perceptual attentional components, including the P200, and extended to later components such as the late positive potential (LPP), which is classically associated with sustained emotional appraisal and motivational associations.”
To understand this, it helps to know that the P200 is a positive spike in electrical activity in the brain that occurs approximately 200 milliseconds after viewing a stimulus. This acts as an automatic marker for early visual attention. Late positive potentials are similar but delayed brain waves that occur about 0.5 seconds later and reflect the moment when the mind deeply evaluates the word and pays sustained attention.
To answer their research questions, the scientists recruited 40 native Italian speakers. All participants were right-handed and had no reported history of neurological or psychiatric disorders. Before the main experiment, researchers removed data from five participants due to technical issues or excessive noise in the recordings, leaving a final sample of 35 youth.
The experimental task involved reading a total of 240 words displayed on a computer screen. These words were evenly divided into neutral, negative, and taboo categories. The selected words were matched based on their length, how often they occur in everyday language, and how familiar they are to the average speaker.
Participants wore a hat with 64 electrodes attached to measure their brain wave activity. Electrical sensors in the scalp detect small voltage changes that occur as the brain processes information. The experiment was divided into two different blocks to test different mental states.
In the first block, called the look condition, participants passively observed the words as they appeared. Participants then rated how pleasant and emotional the words were on a 9-point scale. In the second block, called the acceptance condition, participants read a different set of words but applied emotion regulation strategies. They were instructed to notice their emotional reactions without judgment and to let those feelings pass naturally.
To analyze the vast amounts of brainwave data, the researchers used a machine learning technique known as support vector machines. This type of artificial intelligence algorithm is trained to recognize complex hidden patterns within large datasets. The algorithm evaluated specific voltage spikes associated with the exact moment the word appeared.
“Machine learning analysis further demonstrated that distributed spatiotemporal brainwave patterns can reliably distinguish between neutral, negative, and taboo words,” Ghomroudi and Grecucci told PsyPost. “In particular, taboo words produced the most distinctive neural signatures, suggesting enhanced allocation of attentional and emotional processing resources to stimuli with both emotional and sociocultural significance.”
A machine learning algorithm was able to distinguish between three word categories based solely on electrical activity in the brain. In the passive viewing condition, the model was very accurate in telling the difference between neutral and taboo words. This distinction was primarily driven by brain activity occurring between 637 and 878 milliseconds after the word appeared.
“One of the most interesting findings was how reliably the algorithm could distinguish between different emotion categories from EEG activity alone,” the researchers said. “While some degree of separation between categories was expected, the fact that the neural patterns were sufficiently distinct to allow predictions beyond chance suggests that emotional language leaves a stronger and more structured neural footprint than previously assumed.”
Although the performance of this algorithm decreased slightly during the acceptance condition, it still correctly identified word categories. This suggests that even when a person takes a nonjudgmental stance, the brain continues to register the specific emotional and social weight of words. When we try to accept the emotion, the neural response softens, but the core characteristics of the taboo and negative language remain.
“Our study suggests that emotional states leave measurable traces in brain activity that can be detected using artificial intelligence,” the authors explained. “Although we are still far from being able to read thoughts, these results show that the brain responds in systematically different ways to different types of emotionally meaningful information.”
“More broadly, this study contributes to the long-term goal of understanding how emotions are expressed in the brain and how that expression varies between individuals,” the researchers added. “We also found evidence of neural signatures that control the emotional content conveyed by words.”
There are some limitations to consider, such as the fact that the study relies entirely on the written Italian language. This means that the findings may not fully translate to other cultures, as what makes a word taboo is largely determined by cultural norms. The researchers also emphasized that their algorithm is not a mind-reading device.
“This study cannot accurately determine what people are thinking, nor does it provide tools to read the contents of a person’s private mind,” the researchers warned. “We identify broad patterns related to categories of emotional processing under controlled laboratory conditions. Much more research is needed before similar approaches can be applied to real-world or clinical settings.”
The researchers noted that the experiment could serve as a starting point for more complex investigations. EEG recordings are good at tracking the exact milliseconds in which mental processes occur, but they are bad at pinpointing the exact physical location within the brain. Because machine learning models look at broad patterns across the scalp, the exact brain regions involved in producing these taboo responses remain somewhat unclear.
“This is primarily a proof-of-concept study, not a clinical application,” Ghomroudi and Grecucci clarified. “The importance of this discovery lies less in absolute classification accuracy and more in demonstrating that emotional states can be predicted from non-invasive brain recordings. Establishing this principle is an important step towards future efforts aimed at identifying neural biomarkers of emotional function and dysfunction.”
In the future, the researchers hope to expand their focus to include individuals who struggle to control their emotions. “Our long-term goal is to develop objective brain-based markers of emotional processing and emotion regulation,” the authors state. “One particularly promising direction is the study of clinical populations characterized by abnormal emotional responses, such as anxiety disorders, depression, borderline personality disorder, and post-traumatic stress disorder.”
“If we can identify reliable neural signatures of healthy emotional processing, we may eventually be able to detect when these mechanisms are dysregulated and use this information to improve diagnosis, prognosis, and individualization of treatment,” the researchers continued.
Despite its limitations, this study highlights the powerful ways in which social context shapes our fundamental biology. A word is not just a collection of letters. The human brain treats language as a complex social event that requires continuous moral and emotional evaluation.
“We believe that one of the most exciting aspects of this work is that it shows how neuroscience and artificial intelligence can complement each other,” the scientists concluded. “While neuroscience helps us understand how emotions are expressed in the brain, machine learning provides powerful tools to detect patterns that are difficult to observe with traditional approaches. Combining these methods may ultimately lead to a better understanding of individual differences in emotional functioning and vulnerability to mental health disorders.”
The study, “EEG-based decoding of taboo word perception and regulation,” was authored by Parisa Ahmadi Ghomroudi, Michele Scaltritti, Bianca Monachesi, Atefeh Jalali, Peera Wongupparaj, Remo Job, and Alessandro Grecucci.
