Challenging the classic view, two cognitive scientists argue in a new review that categorization is not a special late stage of sensory processing. Rather, it is a core function that anticipates the body’s needs and movement plans and functions at all levels. Categories are therefore not fixed prototypes stored in “higher-order” areas of the cortex, but are dynamically constructed based on previous experience throughout sensory processing.
In a new review article, “Categorization is baked into the brain,” cognitive scientists Lisa Feldman Barrett, Distinguished Professor at Northeastern University, and Earl K. Miller, Picower Professor at Massachusetts Institute of Technology, argue that categorization is part of a predictive process that the brain uses to efficiently meet the body’s needs in a fast-paced, overwhelmingly sensory world. In that sense, their paper Nature Reviews Neuroscience It challenges decades-old assumptions about how and why the brain summarizes what it sees, hears, smells, tastes, and feels.
A category is a group of things that are similar enough to be considered functionally equivalent. As you walk around your neighborhood, you’ll naturally assume that the furry, four-legged, barking animal in front of you is a dog. In the classic cognitive view, your brain arrives at its classification by absorbing a number of basic sensory characteristics of a dog, such as its shape, size, sounds it makes, and its behavior, and then compares it to the prototype “dog” stored in your memory. A few hundred milliseconds after the first sensory input, you can decide what you want to do with your dog.
Barrett and Miller say that’s not true. Instead, they propose that the brain is able to respond to sensory patterns by predicting motor action plans that are most likely to achieve the needs and goals that present at that moment. These predictive signals can be described as momentary categories that the brain constructs to shape the processing of sensory signals. From the very beginning, incoming sensory signals are compressed and abstracted into their categories, and the optimal predictive plan is efficiently selected. When you’re in an unfamiliar area, your brain may build up a category called “dog” to avoid being bitten, which may result in you slowly backing away while saying “wow!” If you’re on your block and encounter a familiar dog, your brain might construct the category of kneeling down and opening your arms to summon the neighborhood’s adorable pup for a satisfying petting.
In both cases, the category “dog” arises in the context of your needs and predictions from a menu of learned behavioral plans for similar situations. Not from the intellectual exercise of being neutral about sensory inputs, comparing them to a fixed prototype, and planning from there. If your brain really worked in the classically believed way, you would be at a disadvantage when an unfamiliar dog lunged at you.
One of the main things your brain has to do is predict the world. Things take hundreds of milliseconds to process, and in that time the world is moving forward. The brain has to predict things. ”
Earl K. Miller, faculty member of the Picower Institute for Learning and Memory and the Massachusetts Institute of Technology’s Department of Brain and Cognitive Sciences
The most practical and efficient way to survive and thrive in such a world, Barrett said, is to prepare your needs and potential plans for sensory situations. If the prediction is correct, preparations should be in time. If you’re wrong, adjust and learn from it.
“The stimulus-cognition-response model of the brain is wrong,” says Barrett, a faculty member in Northeastern University’s Department of Psychology and co-director of the Interdisciplinary Affective Science Institute. “The brain perceives a stimulus after preparing a response. The brain is not reactive. The brain is predictive. Action planning comes first, and perception is second, depending on the action plan.”
Anatomical and functional evidence
Throughout their review, Barrett and Miller grounded their provocative proposals in the wealth of anatomical, electrophysiological, and imaging evidence regarding the brain. They cite numerous experiments showing how the brain is structured to broadcast memories, create motor plans, and flow backwards toward signals arriving from the body’s sensory surfaces, actively chipping away at memories, shaping them, and giving them meaning.
“The ability to draw similarities from differences to abstractions is built into the structure of the nervous system, and we can tell by looking at what is connected to what and observing the flow of signals,” Barrett said.
For example, when a circuit sends a signal “forward” from a sensory surface (such as the retina) to an area of the cerebral cortex focused on sensory processing (such as the visual cortex), to an area important for executive control (prefrontal cortex) and control of the body (limbic cortex), information is passed from many small, poorly connected neurons to a small number of larger, better-connected neurons. Such architectures compress sensory details into increasingly abstract representations, group many different features into smaller groups of similar features, and in doing so help select predicted plans of action from broader categories that already exist.
“Your brain is a big funnel for taking in the outside world and turning it into output,” Miller says.
Furthermore, anatomical evidence shows that neurons in the cortex maintain more connections to provide feedback from memories that control sensory areas than to forward sensory information. Barrett and Miller write that 90 percent of the synapses in the visual cortex are “feedback” rather than “feedforward.” In other words, the brain is built to use memory to filter incoming sensory signals, consistent with imposing needs and goals on what would otherwise be a flood of sights, sounds, and other sensations.
Yet another piece of evidence is a number of studies from Miller’s own laboratory showing that, at the level of the broad network of information flow in the cerebral cortex, the brain uses beta-frequency waves, which convey information about goals and plans, to suppress the expression of gamma-frequency waves, which convey information about specific sensory inputs.
Finally, the predominance of “feedback” over “feedforward” signals in cortical structures allows for the possibility that sensory signals can be meaningful with respect to predicted plans. If these plans go wrong, the resulting surprises can be consolidated for future use.
“Science has a special name: learning,” Barrett said.
Human thought and its influence on disease
Ultimately, Barrett and Miller’s proposal completely changes the concept of classification, moving it from specific intellectual skills to basic functions for predictively meeting the body’s needs (or “allostasis”).
“Categories may be signal processing events rather than representations that animals do that predictively constrain the meaning of high-dimensional sets of signals in a given situation,” the authors write. “When you classify them, these signals become similar to each other and meaningfully analogous to past allostatic events in terms of some purpose or function.”
Barrett says that humans are able to make classifications that seem overtly metaphorical (e.g., the functional similarity between “climbing a career ladder” and climbing a literal physical ladder) because humans have a relatively large neural network architecture for performing these practical abstractions.
But these processes can go awry in the case of illness, Barrett and Miller note. Depression can be seen as a disorder in which the brain imposes overly broad categories such as “threat” or “criticism” on sensory episodes that do not need to be recognized as such. In contrast, autism is characterized by insufficient compression of incoming sensory signals and insufficient generalization to recognize that a situation is similar to previous situations and select an appropriate plan.
Funding to support this paper was provided by the National Institutes of Health, the U.S. Army Institute of Behavioral and Social Sciences, the Office of Naval Research, the Unlikely Collaborators Foundation, The Freedom Together Foundation, and The Picower Institute for Learning and Memory.
sauce:
Reference magazines:
Barrett, L. F., & Miller, E. K. (2026). Categorizations are “burned into” our brains. Nature Reviews Neuroscience. DOI: 10.1038/s41583-026-01036-2. https://www.nature.com/articles/s41583-026-01036-2
