Compositionality is most obvious in language. A limited vocabulary can generate an effectively unlimited number of sentences because words are assembled according to grammatical rules. The meaning of a sentence depends not only on the words it contains, but also on how those words are combined. This ability allows people to understand entirely new statements, follow unfamiliar instructions and apply old knowledge to new situations. For decades, researchers have treated this systematic flexibility as a defining feature of human cognition and a central ingredient of general intelligence.

The new analysis questions whether compositional thought requires an explicit symbolic architecture. Traditional theories often describe intelligent reasoning as the manipulation of discrete symbols according to formal rules, much like operations performed by a computer program. Under that view, a system must represent objects, concepts or actions as separate symbolic units before it can recombine them. But modern artificial intelligence has complicated this picture. Large language models, trained on enormous datasets and built from neural networks, can produce strikingly novel combinations without being given an explicit grammar or a manually programmed symbolic system.

These models do not simply retrieve sentences from memory. Their behavior suggests that statistical learning across vast numbers of examples can create internal representations that support generalization. A language model may respond appropriately to a combination of words, concepts or instructions that it has never encountered in precisely that form. The researchers emphasize that this does not settle the question of how compositionality works, but it raises a crucial possibility: some compositional behavior may emerge from scale, learning and distributed representations rather than from clearly identifiable symbolic components.

Neuroscience is revealing comparable complexity in animal brains. Studies of animals performing tasks that require flexible reasoning have found evidence for compositional neural codes. In such codes, separate features of a situation—such as an object’s identity, its location, an action or a goal—can be represented in ways that allow them to be recombined when circumstances change. An animal that has learned what an object is and where an action is useful may be able to apply that knowledge to a new combination, even without having experienced the exact situation before.

The technical challenge is determining how neural circuits achieve this flexibility. Biological neurons rarely function as isolated symbols. Instead, information is encoded through patterns of activity distributed across populations of cells. A concept may be represented by the coordinated firing of many neurons, with the same neural population participating in multiple tasks. Compositional computation can therefore arise when these activity patterns are organized so that particular features remain sufficiently stable while other features can be recombined. The result is a neural system that can preserve structure without storing every possible combination separately.

The authors propose studying these mechanisms through a “compositionality continuum,” defined by two interacting properties: the expressivity of computation-specific building blocks and the complexity of the rules used to recombine them. At one end, a system might rely on highly specialized components combined through simple operations. At the other, it might use broad, flexible representations whose interactions are learned through complex distributed dynamics. Between these extremes lies a wide range of possible biological and artificial implementations.

This framework could reshape how researchers compare brains with machine-learning systems. Rather than asking whether an animal, neural network or language model is compositional, scientists could ask what kind of compositional mechanism it uses. Does the system contain reusable representations? Can it apply learned relationships to unfamiliar combinations? Are its recombination rules explicit, implicit or distributed across many units? And how much performance depends on the architecture itself compared with the quantity and diversity of training data?

Answering those questions will require experiments that connect behavior to neural computation. High-density recordings from animals engaged in compositional tasks can reveal how populations of neurons represent individual elements and how those representations change when elements are combined. At the same time, reverse engineering artificial neural networks can identify the internal circuits and activity patterns responsible for flexible behavior. Comparing the two may show whether similar computational principles appear in systems built from biological neurons and systems built from mathematical units.

The debate has implications far beyond language or artificial intelligence. If general intelligence depends on explicit symbolic compositionality, researchers may need to design machines with more structured internal operations. If sophisticated compositional behavior can emerge from large-scale learning and distributed neural dynamics, then increasing model capacity and improving experience may be sufficient to produce abilities once thought to require symbolic reasoning. The proposed continuum does not choose between these possibilities. Instead, it offers a way to measure them, turning a philosophical question about intelligence into a testable problem in neuroscience and machine learning.

Subject of Research: The neural and computational mechanisms underlying compositionality in biological brains and artificial intelligence systems.

Article Title: The compositionality continuum as a principle for studying the neural basis of intelligence

Article References: Riveland, R., Pouget, A. & Driscoll, L. The compositionality continuum as a principle for studying the neural basis of intelligence. Nat Neurosci (2026). https://doi.org/10.1038/s41593-026-02382-1

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41593-026-02382-1

Keywords: compositionality, intelligence, neuroscience, neural networks, artificial intelligence, large language models, neural codes, cognition, symbolic reasoning, general intelligence

The superior temporal lobe, located near the temples, serves as a critical hub for processing complex linguistic and social information. Prior research in adults has noted the lateralization of these functions, with language predominantly processed in the left hemisphere and theory of mind—the cognitive skill allowing individuals to infer others’ thoughts and feelings—processed in the right hemisphere. However, until now, it remained unclear whether such functional segregation was present in the brains of young children or developed later through experience and maturation.

Utilizing functional magnetic resonance imaging (fMRI), the research team scanned the brains of both adults and children aged between 3 and 9 years, while exposing them to carefully designed stimuli. Children heard coherent sentences designed to activate language areas and watched a silent cartoon aimed at engaging theory of mind regions. These stimuli were chosen to invoke specific neural patterns: listening to real language versus nonsensical string of sounds, and observing social interactions eliciting mentalizing responses versus reactions to physical pain cues.

The fMRI data revealed a striking dissociation. The brain regions activated during language tasks were completely distinct from those engaged during theory of mind assessments. This spatial dissociation in children mirrors the patterns previously observed in adult brains, confirming that these cognitive functions are supported by segregated neural architectures from a young age. This runs contrary to earlier hypotheses proposing that language and mentalizing may share overlapping neural substrates during early development.

In addition to spatial separation, the study probed the connectivity patterns of these regions, employing resting-state fMRI scans. Even in the absence of task-driven stimuli, the neural networks underlying language and theory of mind showed unique “connectivity fingerprints” — distinctive patterns of communication with other brain areas. These connectivity profiles remained stable within individuals over time, underscoring the robustness of these functional distinctions throughout childhood.

Crucially, the data suggests that these brain networks did not exhibit increasing segregation with age; rather, their distinctive patterns of connectivity and function were established early and remained consistent from ages three to nine. This challenges models positing that early brain systems overlap extensively and differentiate only through experience and cognitive development. Instead, the findings imply that our brain’s architecture is fundamentally organized into discrete circuits supporting language and social cognition from early childhood.

Interestingly, while children display clear-cut functional boundaries between these regions, neuroimaging of adults revealed a more nuanced interaction—regions associated with theory of mind begin to communicate with areas tied to language. This subtle cross-talk may reflect the complexity of adult cognition, where sophisticated social reasoning and linguistic processing intertwine, enhancing our ability to navigate complex social and communicative landscapes.

Senior author Zeynep Saygin, an associate professor of psychology, emphasizes the evolutionary and developmental significance of these results. “This dissociation suggests that distinct evolutionary pressures shaped neural substrates for language and social cognition separately, equipping humans with dedicated machinery for these core capabilities,” Saygin explains. This insight could reshape our understanding of how uniquely human cognitive faculties emerge during brain development.

The use of advanced fMRI techniques with fine-grained spatial resolution — measuring brain activity at the scale of millimeters and employing 3D voxel analysis — enabled the team to pinpoint these specialized brain regions with unprecedented precision. This approach highlights how technological advancements in neuroimaging continue to deepen our grasp of brain organization and function, particularly concerning the emergence of complex cognitive skills.

Equally remarkable was the longitudinal aspect of this research. By scanning the same children at multiple points, the team confirmed that individual connectivity signatures in language and theory of mind networks remain steadfast over time. This constancy provides strong evidence that these systems are maturing along independent trajectories, governed by preconfigured genetic and neurobiological determinants rather than solely by environmental interaction.

Furthermore, these discoveries have potential clinical implications. A clearer understanding of the separate neural substrates supporting language and social cognition could inform interventions for developmental disorders where these functions are impaired, such as autism spectrum disorder or specific language impairments. Tailored therapies targeting these discrete pathways could enhance efficacy and improve outcomes for affected individuals.

This research contributes to a broader debate about the architecture of human cognition. By demonstrating the early emergence of specialized neural systems for language and mentalizing, it challenges integrated models proposing a shared origin or substantial overlap between these faculties. Instead, it supports frameworks that emphasize modularity and parallel development of complex human skills.

Funded by prominent institutions including the National Science Foundation, the Alfred P. Sloan Foundation, and the National Institutes of Health, this study exemplifies interdisciplinary collaboration bridging psychology, neuroscience, and evolutionary biology. It paves the way for future research exploring how these discrete networks interact dynamically throughout life and adapt to cultural and environmental influences.

As the landscape of cognitive neuroscience evolves, findings such as these provide a compelling glimpse into how our brains are wired not only to acquire language but also to navigate intricate social worlds—two cornerstones of what it means to be human. This foundational knowledge strengthens the bridge connecting biological mechanisms to the richness of human mind and behavior.

Subject of Research: People
Article Title: Functional dissociation of language and theory of mind in the developing superior temporal lobe
News Publication Date: 23-Apr-2026
Web References: https://doi.org/10.1038/s42003-026-10040-2
Keywords: language development, theory of mind, superior temporal lobe, cognitive neuroscience, fMRI, brain connectivity, childhood development, social cognition, neuroimaging, brain lateralization