It is exciting and nerve-wracking at the same time, Martin states about the time before receiving this grant: “This grant represents over a year of preparation, from crafting the proposal to navigating the review process”.
DYNALANG project
The grant will fund Martin’s project, DYNALANG, which integrates two historically divided perspectives in language science. One focuses on grammar and structure, the rules we instinctively follow to form sentences, while the other emphasizes the statistical patterns we learn, such as predicting words in a sentence as large language models do. According to Martin there has been a tendency to see grammar and statistics as separate. But she approaches these as two sides of the same coin.
DYNALANG aims to understand how the brain combines these elements, especially in real-world settings like conversations. The project aims to uncover the hidden patterns of brain activity involved in both speaking and listening, across different languages. By using cutting-edge machine learning techniques, researchers will build "manifold models"—advanced tools that reveal how multiple factors, like speech sounds, grammar, and timing, simultaneously shape the brain's processing of language. These models will help Martin’s team explore how the brain combines diverse sources of information to create rich, composite representations of language, offering a fresh perspective on how we understand and produce spoken language in natural contexts.
Potential applications
While DYNALANG focuses on fundamental research, its potential applications are extensive. A deeper understanding of how the brain processes language could advance technologies like brain-computer interfaces, artificial intelligence, and clinical tools for speech and language disorders.
“Language gives us a window into the brain’s dynamics,” Martin noted. “If we can decode how the brain integrates structure and statistics, we can create more precise models of brain activity. These insights could eventually help develop technologies that enable communication for people with disabilities, improved speech recognition systems, and more effective therapies for language disorders”.
Five year project
The five-year project will also provide opportunities the next generation of scientists. Martin plans to hire two postdoctoral researchers, two PhD candidates, and research assistants to join her team. The research will be conducted at the Max Planck Institute and the Donders Centre for Cognitive Neuroimaging.
Collaboration will play a key role, with contributions from experts like Linda Drijvers at the Donders Institute and Asli Özyürek and Judith Holler at the Max Planck Institute. “Their expertise in gesture and multimodal language will be invaluable”, Martin said.
With DYNALANG, Martin hopes to bridge gaps in language science and inspire future innovations. “This is a new chapter for my research group”, she said. “We are excited to push the boundaries of what we know about the brain and language, including how different types of information come together to be encoded as a whole”.
Read more information about the project and Andrea E. Martin and her team on lacns.github.io, and mpi.nl.