Here’s the latest state of the art in the cognitive science of emotion, summarized in three words:
Emotions are predictions
More precisely, emotions are predictions generated by the brain using an internal model informed by sensory data and past experiences. What does this mean? Let’s analyze that definition word by word, starting from the end:
- Past experiences. All your emotions are shaped by your memories of the past, categorized by distinct emotion concepts. Some of your experiences are strongly associated with joy, others with hate, others with guilt, disgust, gratitude, regret, and so forth. Especially important are your early childhood memories because they tend to contain many of the most paradigmatic emotion episodes.
- Sensory data. Emotions are situated, always embedded in an ever-changing environment. They dynamically depend on what you’re perceiving in your physical environment (e.g., a setting sun or an off-leash pit bull), your social environment (e.g., a nice compliment by or a frowning face on your partner), and your physiological environment (e.g., high blood pressure or an empty stomach)—the latter interoceptive1 signals are of particular importance for emotion creation.
- Internal model. With these sensory and memory inputs, your brain generates new concepts to drive emotion and behavior. In particular, it uses an internal model that functions as a Bayesian filter and performs three operations:
- Categorize. Your brain asks, “What does the sensory data mean?” and categorizes the sensory input to create meaningful sensations. For example, high heart rate alone doesn’t mean much, but if you also feel yourself sweat, hear yourself stutter, experience unusual muscle stiffness, and see a vast number of faces staring at you, your brain may classify that bundle of sensory data as “public speaking,” roughly speaking.
- Combine concepts. Your brain asks, “What emotion categories, rooted in prior experience, best match these sensations?” and searches for functional similarities to combine potentially fitting emotion concepts in various ways. For example, your brain might find your current physiological profile to be similar to a past ‘fear’ experience, your current goals and desires to be similar to a past ‘excitement’ experience, and the nonverbal cues you get from the people around you to be similar to a past ‘shame’ experience. Accordingly, your brain may create conceptual combinations like [50% fear, 30% excitement, 20% shame], [60% fear, 10% excitement, 20% shame], etc.
- Create expectations. Your brain asks, “Given these conceptual combinations, what sensations are most likely to occur next and what behavior will be most beneficial2 in the current situation?” and constructs competing predictions, a set of prior probabilities about what new sensations and motor actions are expected.
- Brain. The brain is your brain, seat of the internal model that involves constant communication between sensory, motor, and limbic areas.
- Generated. The brain doesn’t retrieve discrete, static emotions from some emotional storage room, be it the “soul” or a certain brain area.3 Instead, it continually and flexibly generates new, more or less “mixed emotions” from conceptual combinations of current interoceptive data and past experiences. This implies that every single emotion you experience is unique.
- Predictions. As the product of your internal model, predictions are ad hoc concepts that anticipate sensations and motor actions. When the brain then receives new sensory input, it compares these data to the competing predictions, asking, “How well do they match? Which prediction best fits the new sensory data?” This matching results in posterior probabilities and in a prediction error (= difference between anticipated sensations and actual sensations) that, if not too high, updates your internal model or, if high, selects a different sample from the emotion concepts in your memory to initiate a whole new emotional event. As active inferences, emotional predictions minimize prediction error by triggering physiological changes.
- Emotions. As predictions, emotions are active constructions rather than canned reactions. Since your brain categorizes sensations, combines concepts, generates predictions, calculates prediction errors, and updates your internal model in less than a second, many emotions may be created—many predictions fine-tuned—before you actually have a conscious emotion experience. This conscious experience then motivates high-level behavior (e.g., fight or flight), in contrast to low-level sensorimotor interactions (e.g., rapid eye movements in response to external motion) which are constantly ongoing. In addition, the emotion experience gets integrated into your past experiences, becoming part of your emotionally categorized memory that will influence your future emotions.
You may wonder why your brain does all that. What’s the point of constantly trying to predict the future? The answer is—allostasis, your body’s mechanism to maintain physiological stability through dynamic changes in physiology, cognition, and behavior. Emotions, as predictions, anticipate what your body will need to stay relatively stable in a given situation and then try to meet those needs as fast as possible. Reacting flexibly to changing needs is critical for survival, and if the body can make predictions to adapt quickly, even before needs arise, it can save a lot of energy. So, in a sense, emotions are physiological tools to optimize energy efficiency in the body.
In fact, allostatic predictions seems to be a fundamental principle underlying all psychological faculties. For example, volition is trivially predictive, and we have robust evidence that perception is predictive, too. As a thought exercise, consider further how the placebo effect may be an allostatic prediction; after all, a placebo pill has an effect precisely because you expect an effect: your mind anticipates an improved physiological state, so your body uses allostasis to meet that prediction in order to reduce the potential energy costs of a high prediction error. As a second thought exercise, think about how this might relate to the power of mind and mindsets.
Hoemann K, Gendron M, Feldman Barrett L (2017). Mixed emotions in the predictive brain, Current Opinion in Behavioral Sciences, Vol. 15, pp. 51-57.
- The Bayesian Brain: An Introduction to Predictive Processing
- How to Self-Generate Emotions in 5 Steps
- How Our Beliefs Undermine Our Happiness
- When Reason Needs Emotion: The Problem of Rational Foresight
- Expectations, Mental Toughness, and My 72-Hour Fasting Challenge
- Interoception is the sense of the body’s physiological condition.
- A behavior is beneficial if it supports homeostasis, usually producing desirable sensations (rewards), for example, an experience of relief or happiness.
- Everything we commonly call emotion is a discrete emotion. We say, “This is pride,” “This is boredom,” and “This is anxiety.” But know that “pride,” “boredom,” and “anxiety” are only simple labels for extremely complex cognitive processes. Having a word for an emotion just means that there is some emotional pattern that is uniform enough, forms frequently enough, enters consciousness often enough, and has enough social relevance as to be useful for us to refer to it in our communication. In reality, however, emotions are overlapping and probabilistic. The brain says something like, “This is, in terms of prior experiences, [60% pride, 20% boredom, 20% anxiety] with 40% probability, [80% pride, 5% boredom, 15% anxiety] with 30% probability, and [40% pride, 20% boredom, 40% anxiety] with 30% probability.” Of course, such an array of predictions is still extremely simplified.