The Classical View and Its Collapse
Barrett begins by dismantling what she calls the classical view of emotion, a lineage traceable from Aristotle and Darwin through to Ekman and LeDoux. The classical view posits that discrete emotions like fear, anger, sadness, happiness, disgust, are biologically “hardwired,” each with its own characteristic facial expression, physiological pattern, and neural locus. These emotions are thought to be universal across humans and even other mammals.
Barrett systematically reviews a century of evidence refuting this model. Meta-analyses from her lab found no consistent neural fingerprints for any emotion category. For instance, the amygdala, traditionally labeled the “fear center”, was active in only about 25 % of fear-experience studies and 40 % of fear-perception studies, and equally active during anger, disgust, and even cognitive tasks like learning or decision-making. No brain region or combination of regions consistently maps to a single emotion.
Physiological data are equally inconsistent: anger may or may not raise blood pressure; fear can appear with or without amygdala activation. The same diversity holds for facial muscle movements and autonomic responses. Barrett concludes that variation, not invariance, is the rule, rendering the search for fixed emotional signatures scientifically untenable.
Ekman’s Papua New Guinea studies and the “Fore tribe”
- In the 1960s-70s, Paul Ekman and colleagues used the basic-emotion method: posed photographs of Western actors displaying six “universal” emotions (anger, fear, disgust, surprise, sadness, happiness).
- They showed these photos to the Fore, an isolated horticultural population in Papua New Guinea, asking participants to match each photo to a short emotion story translated into their language (e.g., “Which man is happy because his friends came?”).
- The Fore’s above-chance matching was taken as proof of universality.
Barrett’s critique:
- The task itself forces agreement. Providing emotion words or stories constrains responses; without those cues, accuracy collapses.
- The actors in the photos were coached Westerners, not spontaneously emoting individuals; the “faces” were inventions, not observations.
- When subjects are not offered the emotion vocabulary - when simply asked to describe what the person is doing or feeling in their own words - cross-cultural consistency vanishes.
Replication failures in other cultures
- Later research in Japan, Korea, and other non-Western societies reproduced the same “agreement” only when the same forced-choice tasks were used.
- When participants can freely label expressions, they use culturally specific descriptors - e.g., relational or situational terms rather than discrete emotion names - showing that recognition depends on concept learning, not innate decoding.
Infant and child evidence contradicting universality
- If expressions were truly universal and inborn, infants should display adult-like anger, fear, and sadness patterns before language or culture.
- Developmental studies by Linda Camras and Harriet Oster found that infants’ faces in “anger” (arm restraint) and “fear” (growling-gorilla toy) conditions were indistinguishable by objective coding; adults nevertheless judged them differently from the situational context, even when faces were blurred out.
Cultural context determines perception
- When researchers graft an “angry” face onto a body performing a “disgust” action (e.g., holding a dirty diaper), observers label the emotion according to the body or situation, not the face.
- These findings show that facial configurations gain meaning only through context and learned concepts.
Barrett’s interpretation
- The “universality” reported in hunter-gatherer and cross-cultural studies is an artifact of experimental design that supplies Western emotion categories to non-Western subjects.
- Once those categories are removed, there is no spontaneous cross-cultural agreement.
- Thus, the Fore results and similar studies do not demonstrate innate emotional expressions; they demonstrate cultural transmission of emotion concepts and the power of labeling.
The Theory of Constructed Emotion
In place of the classical model, Barrett advances the theory of constructed emotion. Emotions are not reactions but constructions: moment-to-moment predictions generated by the brain. They emerge from the integration of three systems:
- Interoceptive network - regions monitoring and regulating internal bodily states.
- Conceptual (default-mode) network - systems that use prior experience to categorize incoming sensory and interoceptive data.
- Control network - executive regions coordinating prediction and action.
These networks interact dynamically across the whole brain rather than within dedicated “emotion modules.” Barrett likens this to a kitchen model: the brain holds multipurpose “ingredients” (core systems) that combine differently depending on context. As flour, sugar, and butter yield cookies or cake, the same neural resources produce distinct emotional experiences depending on situational constraints and learned concepts.
Each emotion instance is a whole-brain state, not a discrete circuit. The apparent discreteness of “anger” or “fear” reflects categorization, not underlying essence - an application of Darwin’s population thinking to affective science. “Anger” names a diverse population of episodes sharing functional similarity, not biological identity.
Construction Across Levels: From Neurons to Culture
Barrett situates construction at three nested levels:
- Neuroconstruction: Gene expression and synaptic wiring are context-dependent. Experience - social touch, language, caregiver interaction - modifies neural connectivity even at the cellular scale (plasticity). Thus, early cultural environment shapes the neural architecture underlying later emotional life.
- Psychological construction: Core systems combine dynamically to yield cognition, perception, and emotion alike.
- Social construction: Cultural practices provide the conceptual tools - emotion words, scripts, and expectations - that guide how interoceptive sensations are interpreted. Infants acquire emotion concepts from linguistic and social labeling long before they can identify “anger” or “sadness” in others.
By merging these levels, Barrett rejects essentialism: there is no single biological essence of any emotion. Emotions are conceptually mediated predictions that make bodily sensations meaningful within context.
Predictive Coding and the Role of Concepts
Central to the theory is the brain’s operation as a predictive machine. Rather than passively responding to sensory input, it continuously predicts and constructs the sensory world. Perception itself is simulation: from prior experience, the brain infers what inputs mean and prepares appropriate motor and autonomic responses.
Emotions are one class of these inferences. When an ambiguous pattern of interoceptive signals arises, like increased heart rate, facial flush, and stomach tension, the brain consults past experiences encoded as emotion concepts. Depending on contextual cues and learning history, the same pattern may be categorized as fear, anger, joy, or excitement. This process explains why the same physiological state can yield different emotional experiences across situations and cultures.
Body Budget: Lisa Feldman Barrett uses the term “body budget” as a metaphor for how the brain manages the body’s physiological resources. It’s her way of describing allostasis, or predictive physiological regulation. Emotions, in this model, are not distinct reactions but constructed experiences that help manage the body budget. When you feel anxious, what’s happening physiologically is an anticipated energy outlay, your brain predicts that something important might require action. When you feel calm or content, the body budget is in surplus; the brain predicts stability and recovery.
Empirical Evidence for Construction
Barrett’s lab’s meta-analyses across more than 400 fMRI studies revealed consistent activation not of discrete loci but of distributed networks - chiefly the interoceptive and control networks - during experiences or perceptions of diverse emotion categories. Using multivariate pattern classification, the team derived probabilistic “summaries” (not fingerprints) for anger, disgust, fear, happiness, and sadness. Each category corresponded to variable, overlapping activations within those networks, consistent with degeneracy, the principle that multiple distinct neural configurations can yield the same function.
Implications Across Domains
Barrett argues that accepting constructionism demands reevaluating major institutions built on the classical view:
- Medicine: Misinterpretation of symptoms through gendered emotion stereotypes leads to diagnostic error - e.g., women with myocardial infarction misclassified as anxious
- Law and security: Programs such as the TSA’s Screening Passengers by Observation Techniques assumed emotions could be “read” from faces; the $900 million initiative failed empirically. Courts likewise presuppose emotions as involuntary “animal” impulses needing rational control.
- Technology and AI: “Emotion analytics” industries rest on discredited assumptions of universal facial expressions.
- Social policy and education: Understanding emotions as constructed suggests emotional intelligence is teachable, not innate; children and adults can learn finer-grained concepts to regulate experience.
Ultimately, Barrett reframes emotion as a functional category of predictive regulation: a way the brain maintains allostasis by interpreting internal and external signals through culturally informed concepts.
Key Takeaways
- No neural, physiological, or facial fingerprints exist for any emotion category.
- Emotions are emergent predictions, not reactive states.
- Concepts and culture play constitutive (not modulatory) roles in shaping emotional experience.
- Variation and degeneracy are central organizing principles in affective neuroscience.
- The same core systems underlie emotion, perception, and cognition; emotion has no special neural “module.”
Thought Questions
- If emotions are not universal biological reactions but learned constructions, how might that change the way we interpret other people’s feelings, or our own?
- How much of what we call “anger” or “fear” is actually shaped by the words and concepts we grew up with? Would we feel emotions differently if our language had more or fewer emotional terms?
- If facial expressions don’t automatically reveal what someone feels, what do you think makes people so confident they can “read” emotions from faces?
- Imagine growing up in a culture without the word “sad.” What do you think would happen to the sensations or experiences we now call sadness?
- If emotions are built from predictions your brain makes about the body and the world, what kinds of experiences or habits might change the emotions you tend to feel most often?