How to Build Emotional Intelligence (The Goleman Model Gets the Mechanism Wrong — Here's What the Research Actually Shows)

By Gwyndalyn Henderson

In 1995, Daniel Goleman published Emotional Intelligence, a book that popularized the argument that a set of capacities — self-awareness, self-regulation, motivation, empathy, and social skill — constituted a form of intelligence that was distinct from cognitive intelligence and in many contexts more important. Goleman's model was compelling and commercially successful. It was also, in important ways, not grounded in the research that would subsequently prove most useful for actually building the capacities he described. The science that makes emotional intelligence genuinely learnable — and that identifies the specific mechanism through which it produces better health, behavior, and relational outcomes — comes primarily from a different line of research. Lisa Feldman Barrett, a neuroscientist at Northeastern University, has spent decades studying how emotions are constructed in the brain and published findings that reframe what emotional intelligence actually is, why it varies between people, and what specifically can be done to develop it. Barrett's central concept is emotional granularity: the precision and differentiation of your emotional vocabulary and the corresponding precision of your ability to recognize and distinguish between emotional states. People with high emotional granularity are not people who feel more intensely or more frequently. They are people who, when they feel something, can identify what they feel with more precision — distinguishing "anxiety" from "dread" from "apprehension," "sadness" from "grief" from "disappointment," "anger" from "irritation" from "contempt" — and who can map the specific features of a situation onto a specific emotional concept rather than collapsing everything into a broad, undifferentiated signal. Barrett's research has found that this granularity predicts meaningful differences in health, behavior, and psychological outcomes. The mechanism explains why — and what you can do about it today.

The Goleman Model: What It Gets Right and Wrong

Goleman's model identified something real: there are individual differences in emotional capacity that predict life outcomes in ways that cognitive intelligence measures do not fully capture. People who understand their own emotional states, who can regulate their emotional responses under stress, who can read other people's emotional signals accurately, and who can navigate social dynamics skillfully do better in many domains — career performance, relationship quality, physical health, and psychological wellbeing — than people who lack these capacities. That observation is well supported. The question is what produces those capacities and how they can be developed.

The Goleman model, as popularized, describes the destination — what high emotional intelligence looks like — without fully explaining the mechanism. "Self-awareness," as a construct, does not specify what produces the variation in self-awareness between people or what interventions specifically increase it. "Self-regulation" does not explain the cognitive or neurological process through which some people regulate their emotional responses more effectively than others, or what specifically can be practiced to improve that process. The research that addresses the mechanism — that identifies what, specifically, you can change about your emotional processing that will produce the outcomes associated with high EQ — comes primarily from Barrett's work on emotional construction and Matthew Lieberman's work at UCLA on affect labeling.

This is not a criticism of Goleman's contribution. Naming the construct and demonstrating its importance was a necessary first step. But the prescriptions that follow from the mechanism-level research are different from the ones that follow from the description-level model. Understanding that emotional intelligence is fundamentally about the precision of emotional concepts — about having more fine-grained emotional vocabulary and applying it accurately — changes what you practice. It is a more specific target. And specific targets produce more specific results.

Barrett: Emotions Are Constructed, Not Detected

Barrett's theory of constructed emotion, detailed in her 2017 book How Emotions Are Made, challenges the model of emotion that underlies most popular emotional intelligence frameworks. The common assumption — inherited from older psychological and neuroscientific models — is that emotions are innate, universal responses that are triggered by situations: anger circuits fire when you are threatened, sadness circuits fire when you experience loss, fear circuits fire when you detect danger. On this model, emotional intelligence is the capacity to detect these firing circuits accurately — to recognize which emotion you are feeling when it is triggered — and to manage the behavioral consequences of the trigger.

Barrett's research, drawing on neuroimaging data and cross-cultural emotion research, finds that the brain does not work this way. Emotions are not detected; they are constructed. The brain receives interoceptive signals — the ongoing stream of information about the body's internal state — and makes sense of those signals by applying concepts from memory. What you call "anxiety" is an interoceptive signal pattern that your brain, drawing on the concept "anxiety" from past experience and cultural learning, has categorized as anxiety. The same interoceptive signal pattern might be categorized as "excitement" by someone whose concept repertoire handles the distinction differently. Emotions are the brain's best guesses about what its internal signals mean, made by matching those signals to the concepts available for the matching.

This constructionist account has a specific implication for emotional intelligence development: the precision of your emotional concepts determines the precision of your emotional perception. If your concept repertoire for negative internal states consists primarily of "bad," "stressed," and "upset," your brain will construct experience from those three options — and the behavioral and physiological responses that follow will be calibrated to those coarse categories. If your concept repertoire distinguishes "anxious" from "apprehensive" from "dread" from "helpless" from "overwhelmed" from "irritable from under-stimulation," your brain can construct more precise experiences from the same interoceptive signal — and the regulatory options available will be correspondingly more precise. More precise emotion concepts produce more precise emotional experiences, which produce more targeted regulatory responses. That is Barrett's contribution to what emotional intelligence actually is: it is not primarily a social skill or a personality trait; it is a conceptual capacity.

Emotional Granularity: The Mechanism That Predicts Outcomes

Barrett and her colleagues operationalized the concept of emotional granularity through research designs that measure how precisely people differentiate between emotional states. In a typical paradigm, participants rate their moment-to-moment experience on multiple emotion dimensions across many occasions, and researchers analyze the degree to which those ratings cluster together (low granularity — experiencing negative emotions as an undifferentiated mass) or differentiate from each other (high granularity — distinguishing between specific negative emotional states). People with high emotional granularity use different words to describe different emotional states in ways that accurately reflect the distinct features of their situations. People with low granularity tend to report that different negative states feel generally similar — they collapse what the research treats as conceptually distinct experiences into a single global negative state.

The outcome differences Barrett's research has documented are substantial. People with higher negative emotional granularity — more precise differentiation among negative emotional states — are less likely to drink excessively in response to stress. They are less likely to aggress against someone who has provoked them. They recover more quickly from setbacks. They show greater flexibility in their regulatory responses to difficult situations. They report better psychological adjustment over time. The mechanism Barrett proposes is that precise emotional concepts enable targeted regulation: if you know you are "anxious about a specific outcome you can't control" rather than simply "bad," you have access to regulation strategies specifically relevant to uncontrollability (acceptance, cognitive reappraisal of the stakes) rather than the default responses to global negative affect (distraction, avoidance). Precision enables targeting. Targeting produces better outcomes than broad-spectrum responses to undifferentiated distress.

Barrett's research also found that positive emotional granularity predicts outcomes independently of negative granularity — more precise differentiation among positive emotional states is associated with better physical health, higher life satisfaction, and greater resilience. The mechanism is analogous: precise positive concepts enable more effective savoring, more targeted reinforcement of valued experiences, and clearer recognition of what conditions produce genuinely meaningful states versus superficially pleasant ones. High granularity across the emotional spectrum is the functional description of what Goleman was pointing to as emotional intelligence. The development path is through vocabulary and attention — specifically, through deliberate practice expanding emotional concept precision.

Lieberman: Why Naming an Emotion Changes It

Matthew Lieberman's neuroimaging research at UCLA provides the neural mechanism that connects Barrett's emotional granularity concept to practical regulation. In a 2007 study published in Psychological Science, Lieberman and colleagues showed participants images of faces expressing strong negative emotions while participants' brains were scanned using fMRI. In one condition, participants labeled the emotional expression in the image with a word. In another condition, they performed a different cognitive task that did not involve labeling. The result: affect labeling — putting an emotional experience into words — significantly reduced amygdala activation in response to the emotional stimulus. The act of labeling the emotion engaged the prefrontal cortex and reduced activity in the brain's primary threat-detection region. Naming the emotion changed its neurological processing.

Lieberman's subsequent research found that the labeling effect was more pronounced for more precise labels. Labeling an image "negative" produced less amygdala reduction than labeling it "angry" or "afraid." More precise labels engaged prefrontal circuits more fully and produced greater regulatory benefit. This finding connects directly to Barrett's granularity framework: a larger, more differentiated emotional vocabulary provides more precise labels, which produce stronger regulatory effects when applied. The two research programs converge on the same prescriptive conclusion: expanding your emotional vocabulary and applying it precisely — in real time, to your own experience — is a neurologically active regulatory intervention. It is not just description. It is mechanism.

The implication for building emotional intelligence is specific: the practice is not "become more aware of your emotions" as a general aspiration. It is: learn more precise emotional concepts, apply them to your ongoing experience as you move through your day, notice the interoceptive signals that differentiate one emotional state from another, and build a habit of labeling your emotional experience at the level of precision that the granularity research identifies as predictive of better outcomes. This is a practice that can be done incrementally, starting with whatever emotional vocabulary you currently have and expanding it deliberately over time.

Quick Win — The Granularity Expansion Protocol

This is a two-part protocol. The first part is a one-time audit that establishes your current emotional granularity baseline. The second part is a daily practice that expands it over time. Both are done in writing, because the externalization of the labeling process makes the precision requirement concrete — you cannot be vague in writing in the same way you can be vague in thought.

1. Baseline audit (one session, 20 minutes): Think of the last three times you felt genuinely stressed, upset, anxious, or emotionally activated in a way that felt unpleasant. For each occasion, write down the emotion label you would have used in the moment to describe what you were feeling. Now, using a more expansive emotional vocabulary — the list below is a starting point — see if you can identify a more precise label for each occasion that distinguishes its specific character from the general category. General: "stressed" → Precise options: overwhelmed (too many competing demands), dread (anticipating something bad that feels inevitable), apprehension (uncertain about a specific outcome), performance anxiety (specifically about evaluation by others), social anxiety (specifically about interpersonal judgment), moral distress (conflict between what you are doing and what you value). For each of your three occasions, write the original label and the more precise one. The exercise builds the habit of matching interoceptive signals to precise concepts — which is the granularity practice. Some vocabulary to start with: agitated, apprehensive, bereft, contemptuous, dejected, despondent, dread, exasperated, forlorn, frustrated, helpless, humiliated, irritable, melancholy, overwhelmed, resentful, resigned, restless, uneasy, wistful. Add the ones that match your actual experience. Retire the ones that don't.
2. Daily labeling practice (2 minutes, twice daily): At midday and at the end of the day, take two minutes to label your current emotional state with the most precise word available. Write it. Rate the intensity from one to ten. If the word doesn't feel quite right — if it's close but not accurate — write the reason it doesn't fit. This acts as a search function: "I feel overwhelmed but it's not that the demands are too many — it's that I feel like none of my effort is producing visible progress." That second clause is the more precise concept. Find or create the word for it. Over weeks of this practice, you are building a personal emotional lexicon — a set of concepts derived from your actual emotional experience, refined through repeated precise labeling, that gives your brain more accurate tools for constructing and regulating your emotional states. This is the practice Barrett's research points to. It is simple, requires no equipment, and takes four minutes per day. The research suggests those four minutes, applied consistently, produce measurable changes in how you experience and regulate emotional activation — the mechanism of what Goleman was calling emotional intelligence, at the level the research has actually located it.

Building emotional intelligence starts with understanding what emotional intelligence actually is — not a social skill or a personality trait, but a conceptual precision that determines how specifically your brain can construct, perceive, and regulate your emotional experience. Barrett's granularity research identifies the mechanism. Lieberman's labeling research confirms the neural pathway. The practice is vocabulary expansion and daily precise labeling — not as a description exercise, but as a neurologically active regulatory practice. If you want the full framework that extends this into an integrated attention and self-observation practice, The Focused Mind gives you exactly that daily structure. Barrett identified the mechanism. Lieberman showed the neural pathway. The Focused Mind gives you the practice.

See also: How to Master Your Emotions for the full Lisa Feldman Barrett theory of constructed emotion and the Ethan Kross linguistic distancing research, and How to Be More Self-Aware for the Tasha Eurich research on why most introspection produces confident but inaccurate self-models.