Everyone has been told, at some point, to talk about it. Keep a diary, say it out loud, tell someone how you feel — the advice is old enough to be furniture, older than psychotherapy, older than the word for it. For most of that history the claim rested on the fact that it seemed to help, which is the weakest kind of evidence there is. Then in 2007 a lab in Los Angeles put people in a scanner, showed them upsetting photographs, and asked a plain question: what happens in the brain when you give the feeling a name? The answer was specific enough to build on, and it arrives with two catches. One is that almost everyone predicts the opposite of what actually happens. The other is a boundary that showed up fourteen years later, in a result that quietly undercuts the most common way this trick gets used.
🧠 The Scanner and the Word
Matthew Lieberman and colleagues ran the study at UCLA in 2007, and the design turns on a problem the earlier work had never solved cleanly. Everyone in the scanner saw the same set of negative images. What varied was the job they had to do with them. In one condition they chose an emotion word that matched what the picture made them feel — afraid, angry, disgusted. In the comparison conditions they had to encode the same pictures some other way, without naming the emotion. Same face, same lighting, same photograph. Only the instruction differed.
The result, published in Psychological Science under the title “Putting feelings into words,” was a double dissociation. Choosing the feeling word produced less activity in the amygdala and other limbic regions than the other encoding tasks. At the same time, it produced more activity in a single region: the right ventrolateral prefrontal cortex. And the two moved against each other — the more that language region lit up, the quieter the amygdala went. The link between them ran through the medial prefrontal cortex, which gave the authors a physically plausible route for the signal: right ventrolateral prefrontal cortex, through medial prefrontal cortex, down to the amygdala.
The paper called itself the first unambiguous evidence of the effect, and it was standing on two earlier results. Hariri and colleagues in 2000, and Lieberman’s own group in 2005, had both found that processing the emotional content of a face linguistically rather than perceptually produced less amygdala activity and more ventrolateral activity. There is supporting evidence from damage, too: people with lesions in the right ventrolateral prefrontal cortex are worse at identifying a film character’s emotional state as the story unfolds. Take the region out and the labelling gets harder, which is what you would expect if it is doing the work.
⚙️ How It Works
The mechanism is easier to follow if you start with what a feeling is before you have a word for it, because that version is genuinely worse to be inside.
Step one: the body gets there first, and it arrives unlabelled. Something tightens in the chest, the heart rate climbs, attention narrows, and the whole system mobilises before you know what the emergency is. The amygdala is the region doing a lot of that mobilising, and it responds to threat-relevant input that is uncertain by spending generously — it does not know yet whether this is a small thing or a large one, so it funds it like a large one. The vagueness is not a side effect of the distress. It is a large part of what makes the distress expensive.
Step two: naming force-converts a continuous bodily state into a discrete symbol. “I am humiliated” rather than “something is wrong.” That is not free, and it is not automatic — it requires taking a mixed, continuous, whole-body signal and cutting one piece out of it that fits a category you already own. This is the step that everyone imagines is the dangerous one, and it is the one the scanner says is doing the opposite of harm.
Step three: the language region picks up the job. Converting the state into a symbol recruits the ventrolateral prefrontal cortex, a region associated with symbolic processing of emotion and with top-down inhibition. In the scanner it is the one area that goes up while everything around it goes down.
Step four: the inhibition travels. The signal moves from the ventrolateral region through the medial prefrontal cortex to the amygdala, and the two ends are negatively correlated. Dynamic causal modelling on this pattern has been taken to show that the direction runs from the prefrontal region to the amygdala, not the reverse — the naming is causing the quiet, not the quiet allowing the naming.
Step five: what has changed is not the situation. Nothing about the image, the news, or the argument has been reinterpreted. What changed is the classification of the signal — from an unclassified alarm, which commands resources on the scale of its uncertainty, into a known object with a name on it. You did not solve anything. You moved it from the emergency channel to the ordinary one.
Three notes on what this mechanism is not. It is not reappraisal: you are not changing the meaning of the event, only its format, which is why the two have overlapping but not identical brain signatures — reappraisal additionally recruits the anterior cingulate, the supplementary motor area, and dorsolateral prefrontal cortex, and produces a larger drop in felt distress. It does not require intent: you can label without meaning to regulate anything, which is why researchers file it as implicit emotion regulation, distinct from the deliberate kind. And it may not be about emotion at all: classifying images into non-emotional categories such as “human” or “landscape” produces the same pattern of more ventrolateral and less amygdala activity, which suggests the driver could be the act of turning a stimulus into language, full stop.
That last point also weakens the simplest explanation, which is distraction. A pure attention-turning task — deciding whether the face in the picture is male or female — does not produce the same reduction. And in the exposure-therapy work below, labelling kept beating distraction on the measures that lasted.
🤔 Where the Story Gets Complicated
The prediction is backwards, and experience does not fix it. In 2011, Lieberman, Inagaki, Tabbinia and Crockett ran four studies on the subjective side of this effect, published in Emotion. In the first, self-reported distress was lower during labelling than during passive viewing. The second and third added reappraisal and distraction as comparisons. In all three, participants were asked ahead of time what they thought labelling would do to their distress, and they said — reliably, across studies — that it would make them feel worse. Then it didn’t. The fourth study flipped the pictures from negative to positive, and here something stranger surfaced: labelling also reduced self-reported pleasure. The overall pattern is not “words soothe pain.” It is that putting a feeling into words turns the volume down on affect generally, good and bad alike. The authors’ own read on why nobody expects this is unflattering: people’s theories about how emotion regulation ought to work may be largely immune to their own contradictory experience.
The boundary: naming first can make the next step harder. This is the finding that should change how you use the trick. In 2021, researchers at Harvard, Northeastern and Columbia published two experiments in Affective Science (N = 80 and N = 60) that tested an assumption so ordinary nobody had bothered to check it — that naming your feelings should help you manage them. Participants looked at aversive images and were assigned to passively observe, to name the emotions the images provoked, to regulate by reappraising, or to do both.
Naming made the reappraisal worse. People who named their emotions before reappraising reported feeling worse than people who reappraised without naming — replicated in a second within-participants study, which also showed the same interference with mindful acceptance, a strategy that never asks you to argue with anything. And in those studies, naming on its own was not better than just looking: the difference between “look” and “name” was not significant. The authors’ explanation is the memorable part. Giving a feeling a specific category may crystallise it — fix it into a form that is harder to modify afterwards. The comforting word that lowers the temperature is also, it seems, a setting of the mould.
Mind the channel you are measuring. In 2012, Katharina Kircanski, Lieberman and Michelle Craske took spider-fearful participants and exposed them to a live tarantula over two days. Four conditions: affect labelling, reappraisal, distraction, and exposure alone. At a one-week follow-up, the labelling group showed the largest drop in skin conductance response — beating the other three, on a measure nobody can consciously steer. But the self-reported fear scores did not differ between conditions. Same intervention, opposite verdicts depending on whether you ask the sweat glands or the mouth. The authors suspect the self-reports were contaminated by the very lay theory from 2011 — people who do not believe a technique works may not report it working.
And not every feeling is label-able into compliance. The review literature notes a study that looked at anger and shame side by side: labelling reduced anger but did not touch shame. That fits the crystallising story from the other direction. Anger is about something out there. Shame recruits the whole self, and there is no word big enough to hold it.
The honest state of the field. This is a young area. The Wikipedia summary of it says outright that the mechanisms are hypotheses in need of empirical support, that clinical use is tentative rather than established, and that almost everything has been demonstrated in laboratories. The effect sizes are modest, and one paradigm detail keeps producing different answers: when participants have to generate their own label, the effects tend to appear later; when they pick from a supplied list, the effects tend to appear immediately — possibly because being handed a word feels like support from the experimenter. Which is its own small lesson about how much of “naming” is really interpersonal.
🔗 Why This Matters Outside the Lab
Anyone building something that a person talks to will meet this finding on the first day, whether or not they have a name for it. When someone arrives carrying a feeling they cannot quite describe — a bad week, a grey mood, a pressure they cannot place — the highest-value response is usually not a diagnosis and definitely not a plan. It is a word. “Is it closer to angry or to disappointed?” is doing something measurable in that person’s head: it hands them the conversion step, the one that moves a signal out of the alarm channel and into a category. Jumping to solutions skips the step where the effect lives. The person is not helped by being told what to do about a feeling they have not yet successfully named.
The 2021 result is the part worth carrying into how you sequence things. The folk order — name it first, then reframe it — turns out to have a cost, because the name seems to set the feeling in place. If the goal is to take the heat down, name it. If the goal is to see the situation differently, go straight to the reinterpretation and skip the labelling. Doing both in that order is working against yourself, and it is the exact order that shows up in a hundred self-help scripts and a fair number of therapy sessions.
It is also happening at scale, mostly without anyone noticing. A study of 74,487 Twitter users looked at tweets of the form “I feel…” and tracked the emotional valence of what those accounts posted before and after. Intensity built towards the labelling tweet, peaked right at it, and then decayed quickly back to baseline. Millions of people are performing the intervention on themselves in public, in the form of a sentence they think is just complain-posting.
And it is useful in fiction for reasons that have nothing to do with therapy. The ability to name a state precisely, versus describing everything as “I’m fine” or “everything is awful,” is a clean way to show a character’s inner command without a single line of narration about their emotional intelligence. The character who says jealous instead of annoyed has just told you how much they know about themselves — and a character who can only ever locate the coarsest version of what is happening is already in trouble, and you have not had to say so.
🎲 Bonus
The vocabulary matters as much as the act. Lisa Feldman Barrett’s work on emotional granularity measures how finely a person carves up their own affect — whether “bad” splits into bored, tired, lonely, miserable, or stays as one blunt slab. People who differentiate more finely tend to show better outcomes under stress, and the standard finding in that literature is that they rely less on things like drinking or binge eating to manage a mood. Fewer categories, coarser tool. The word is the smallest unit of self-control that does not require willpower.
One more boundary for the pocket. Because labelling seems to work by turning a signal into a symbol, the trick should get weaker as the feeling gets more total — and that is roughly what the evidence shows. Anger cools. Shame does not. So if you are going to recommend the technique to yourself, aim it at the feelings that have an object and a direction: frustration about a meeting, anxiety about a deadline, envy about a friend’s news. Save the ones that implicate the entire self for something other than a word.