Decoding the Language of the Brain: How AI and Genetics Are Unlocking Communication’s Secrets
Learning French, reading a novel, or simply chatting with friends – language is fundamental to the human experience. But beneath its seemingly effortless flow lies a remarkable complexity. Cognitive neuroscientists are now leveraging cutting-edge tools, including genetic analysis and artificial intelligence, to unravel the mysteries of both typical and disordered communication.
From Genes to Neural Pathways: An Integrated Approach
Traditionally, language research has often focused on isolated components – genes, brain regions, neural activity, behavior, and computational models – without fully integrating them. “We still tend to study language one level at a time…without fully connecting those levels into a coherent mechanistic account,” explains Tamara Swaab, chairing a symposium on language at the annual meeting of the Cognitive Neuroscience Society (CNS) in Vancouver, B.C. However, this is changing. Researchers are now able to study these connections in greater detail, leading to breakthroughs in understanding how language develops and why it varies across individuals.
AI Models Predict Language Development
One exciting development is the use of AI-based models to test and potentially predict language development in children. Simultaneously, genetic research is revealing links between rhythm disorders and dyslexia. This shift marks a move away from simply identifying where language occurs in the brain to understanding how it happens and why it differs so significantly between people, according to Swaab of the University of California, Davis, and University of Birmingham in the UK.
How AI is Mimicking the Human Brain
The evolution of AI, particularly deep learning models, is providing novel insights into how humans acquire language. Jean-Rémi King, a cognitive neuroscientist at Meta, notes that humans learn language with far less exposure to words than current large language models (LLMs). “With the rise of small and then large language models, using artificial neural networks became, de facto, the most efficient way to model and decode language representations in the brain,” King says.
King and colleagues recently found that LLMs can accurately model neural responses to audiobooks in both adults and children as young as two years old. This research, conducted with the Rothschild Foundation Hospital’s pediatric epileptology unit, involved analyzing neural activity recorded from over 7,400 electrodes implanted in the brains of 46 participants. The findings suggest that high-level language features, like grammar, continue to mature between ages two and ten.
Beyond Broca’s and Wernicke’s Areas: The Brain’s Language Network
Stephanie Forkel of Radboud University Nijmegen in The Netherlands is taking a different approach, focusing on the brain’s wiring that connects language regions. Rejecting the traditional view of language being localized to “Broca’s area” or “Wernicke’s area,” Forkel emphasizes that language is a complex system. Using ultra–high-field 7 Tesla diffusion MRI, her team reconstructed seven major white-matter pathways involved in language in 172 individuals.
The research revealed that language isn’t binary in the brain – it forms a continuum, challenging the idea of distinct “left-brained” versus “right-brained” language dominance. Forkel’s team is now embarking on a five-year project to understand the biological foundations of language development and how to protect or restore it after injury or disease.
The Polygenic Nature of Language
Understanding the genetic underpinnings of language is similarly advancing rapidly, fueled by large datasets from sources like 23andMe and the U.S. National Institutes of Health. Reyna Gordon of Vanderbilt University Medical Center highlights that language is influenced by many genes. While pinpointing the genetic contribution to language skills in individuals is challenging, large population studies reveal significant patterns.
Gordon’s team is integrating genetic data with language and music development questionnaires to show how genetic variation contributes to individual differences. A recent study analyzing data from 1 million participants from 23andMe, along with language testing data, identified multiple genes associated with dyslexia, potentially leading to earlier diagnosis and treatment.
research suggests a shared biological basis between language and music, with 16 genomic regions common to both rhythm impairments and dyslexia. Rhythm impairments may even be a risk factor for language problems and reading disorders.
An Adaptable Brain
Collectively, the research presented at the CNS meeting demonstrates the brain’s remarkable adaptability. “The human brain is not built from rigid blueprints, but rather from adaptable architectures,” Forkel says. Swaab concludes, “Language comprehension is a form of swift, adaptive cognition. We finally start to more fully understand it by linking the story from genes, to brain pathways and networks, to neural decoding and computational models that help explain how the brain comprehends and produces language.”