Fake Words With Real Feelings: Giant Study Shows Emotion Sneaks Into the Brain’s Word Detector

Imagine staring at a string of letters that does not exist in any dictionary—something like “rapidly” scrambled into a plausible-looking but entirely invented word. Your brain has to decide, within a fraction of a second, whether it is looking at a real word or an impostor. Psychologists have long used this task, known as the lexical decision task, to probe how we recognize words. But a new and unusually large study suggests something surprising: even when a letter string is completely meaningless, it is not necessarily emotionless. Fake words that carry emotional echoes—through their roots, their suffixes, or the very sounds they contain—trip up the brain’s word detector, making people slower and more error-prone when trying to reject them.

The research, published in the journal Behavior Research Methods, was conducted by Celia Martínez-Tomás of the Universidad Complutense de Madrid and colleagues, including Marc Guasch and Pilar Ferré of the Universitat Rovira i Virgili. It belongs to a class of experiments known as megastudies: instead of testing a few dozen carefully chosen items on a small group of volunteers, megastudies throw thousands of stimuli at huge samples of participants and let statistical models tease apart the factors that shape performance. The sheer scale makes it possible to detect effects that smaller experiments might miss, and to examine how different properties of words—frequency, emotional tone, structure—interact with one another in a single dataset.

In this case, the scale was extraordinary. A total of 1,025 native Spanish speakers made lexical decisions on 1,200 real experimental words and 4,800 experimental pseudowords, along with 1,200 filler words and 1,200 filler pseudowords. The pseudowords were not random gibberish. They were systematically constructed from emotional and neutral base words, varying in their morphological structure: some contained both a real root and a real suffix, some contained just one real morphological component, and some contained neither. Every participant thus faced letter strings that ranged from fully word-like assemblies to wholly invented forms, allowing the researchers to ask precisely how much each ingredient of wordiness influences the decision to accept or reject an item.

The results, analyzed with linear and generalized linear mixed-effects models—statistical techniques that simultaneously account for variability across participants and across items—revealed a clear hierarchy. Pseudowords containing both a real root and a real suffix were the hardest to reject: they produced the slowest response times and the highest error rates. Pseudowords with only one real morphological component, such as a real root paired with a non-suffix or a non-root paired with a real suffix, fell in between. And pseudowords lacking any identifiable morphological parts were rejected fastest and most accurately. In other words, the more the fake word resembled a genuine morphological assembly, the more the brain hesitated before declaring it a fake.

This pattern fits with decades of work on morphological processing. Skilled readers do not simply match letter strings against a mental dictionary; they also decompose words into their component parts. Classic experiments have shown that readers segment words along morpho-orthographic lines almost automatically, even when the segmentation leads to odd results—famously, the string “brother” surfacing inside “brothel.” When a pseudoword contains a real stem and a real affix, that decomposition machinery fires anyway, activating traces of real words and creating a lingering sense of familiarity. The new megastudy extends this logic to Spanish and demonstrates, with unprecedented statistical power, that morphological composition is a major determinant of how confidently we reject nonwords.

But the emotional findings are what make the study stand out. The researchers found that pseudowords with higher emotional valence—whether strongly positive or strongly negative, as rated for their base words—and higher arousal, meaning greater emotional intensity, elicited more errors and slower rejection times. Emotionally charged fake words, in short, were more likely to be mistaken for real ones. This suggests that affective information is not a bonus feature that only activates once a word has been recognized; it can be accessed even in the complete absence of a lexical representation. The brain appears to extract emotional meaning from strings that it has never seen before and that, by definition, have no stored meaning at all.

How can a meaningless string carry emotional weight? The authors propose two complementary routes. The first is via activation of existing representations: a pseudoword built from the root of “death” or “love” partially activates those real words and, through them, their emotional associations. The second route is more radical and touches on a lively debate in psycholinguistics: systematic links between certain sublexical features—particular sounds, letter sequences, articulatory gestures—and emotional meaning. Research on affective iconicity has shown, for example, that words involving the lips versus the tongue can differ systematically in pleasantness, and that the sound of a word alone can evoke affective brain responses. Previous studies have even found that people can learn emotional connotations for pseudowords through conditioning, and that statistical regularities between phonological form and valence exist in the lexicons of English, Spanish, and German.

Crucially, the new study found that the emotional effects interacted with morphological structure, indicating that affective cues and sublexical, structural cues do not operate in isolation but contribute jointly to pseudoword processing. A fake word’s emotional pull depends partly on how word-like its architecture is. This joint contribution complicates simple models of lexical decision, in which a noisy evidence-accumulation process compares the input against stored word knowledge. It suggests that multiple parallel streams of information—orthographic, morphological, and affective—all feed into the decision, and that emotion can bias the process even when the lexical stream comes up empty.

The findings also speak to a broader theoretical question: where does emotional meaning come from in the first place? One influential view holds that emotional content is grounded in embodied experience—words for disgust, say, inherit their affect from bodily reactions. The pseudoword results add a twist: if emotional information can be gleaned from form alone, then the mapping between sound, structure, and feeling may be more systematic and more deeply woven into language than previously assumed. For models of word recognition, this means that affective dimensions such as valence and arousal cannot be treated as late-arriving semantic add-ons; they may need to be built into the earliest stages of processing.

Beyond its theoretical implications, the study delivers a substantial practical resource. The researchers have released an open database, hosted on the Open Science Framework, containing reaction time and accuracy data for 2,400 Spanish words and 6,000 Spanish pseudowords that vary in emotional and morphological composition. Materials and analysis code are also publicly available, although the study itself was not preregistered. For researchers designing experiments on emotional language, the database offers a way to select stimuli with known processing profiles; for modelers, it provides a rich benchmark against which computational accounts of lexical decision can be tested. The work was approved by the Psychological Ethics Committee of the Universidad Complutense of Madrid and funded by Spanish and European research grants.

There are, of course, limits to what any single dataset can settle. The study was conducted in Spanish with native speakers, and morphological structure is far richer in Spanish than in some other languages, so cross-linguistic replication will be needed to establish generality. The authors also note that the emotional properties of the pseudowords were inherited from their base words and from sublexical statistics, rather than being directly felt by participants—future work combining lexical decisions with explicit ratings, electrophysiology, or learning paradigms could trace exactly when and how affective activation arises. Still, the headline conclusion is striking: the brain’s word detector does not merely check whether a string exists in the mental dictionary. It also listens for echoes of feeling. Even a word that means nothing can, in a very real sense, mean something emotionally—and that hidden emotional life is enough to fool us, if only for a few hundred milliseconds.

Subject of Research: Effects of morphological structure and emotional content on pseudoword recognition in a large-scale Spanish lexical decision megastudy

Article Title: Meaningless, but not emotionless: Evidence from a megastudy of pseudoword lexical decisions

Article References: Martínez-Tomás, C., Guasch, M., Lázaro, M., Baciero, A., Ferré, P., & Hinojosa, J. A. (2026). Meaningless, but not emotionless: Evidence from a megastudy of pseudoword lexical decisions. Behavior Research Methods, 58(11), Article 314. https://doi.org/10.3758/s13428-026-03186-8

Image Credits: AI Generated

DOI: 10.3758/s13428-026-03186-8

Keywords: pseudowords, lexical decision task, megastudy, emotion, valence, arousal, morphology, psycholinguistics, word recognition, affective iconicity, Spanish language, cognitive psychology