Summary: A new study analyzed a very large functional magnetic resonance imaging (fMRI) dataset from more than 700 participants and, by using more permissive statistical thresholds to capture smaller signals that conventional analyses typically filter out, identified 17 additional, distinct brain regions that respond to language.
These newly revealed language-responsive sites extend well beyond the classic left-hemisphere language centers and include locations in subcortical regions, memory-related structures, and motor-coordination networks.
Key Facts
- The Strawberry Metric: Although the 17 newly identified language nodes are dispersed across diverse brain areas, together they occupy only about 5 percent of an adult brain’s volume—roughly the size of a large strawberry.
- The Resistance Frontier: Senior author Evelina Fedorenko notes that early in her career, language-related activations outside canonical left-hemisphere centers often appeared in her data but were discouraged from being reported. This investigation revisited more than a decade of scans to systematically validate those weaker signals.
- The Language Localizer Grid: To locate these nodes, all 772 participants completed a standardized “language localizer” while in an fMRI scanner: they read or listened to meaningful sentences and to matched strings of meaningless nonwords. The researchers isolated regions that showed reliably stronger responses to real sentences in both reading and listening conditions.
- Infiltrating the Coordination Core: Five of the 17 newly revealed sites lie in the cerebellum, a structure traditionally associated with motor control and coordination.
- Multitask Integration Hubs: By comparing responses from a spatial working memory task, the team found that three cerebellar regions responded to both language and nonlinguistic cognitive demands. These dual-purpose sites may function as integration hubs that route information across distinct cortical systems.
- Deep Brain Geography: Other newly identified language-selective regions are located in the medial frontal cortex, along the inferior surface of the left temporal lobe, and within memory- and emotion-related structures such as the hippocampus and amygdala.
Source: MIT
Overview: For decades, research has emphasized specialized language centers in the left hemisphere. This new MIT-led study demonstrates that language processing engages a broader, distributed network that includes multiple additional regions across the brain.
Using fMRI data from over 700 people scanned in the same lab, the investigators identified 17 additional regions that show reliable language-related activation. These areas span the cerebellum, the hippocampus, and various parts of the cerebral cortex, and together represent about 5 percent of the brain’s volume.
“Even though there are distant components, the overall volume involved is quite limited. You don’t need that much brain tissue to support language,” says Evelina Fedorenko, MIT associate professor of brain and cognitive sciences and senior author of the paper.
The exact roles of these regions in language processing remain to be fully characterized, but the study provides an important roadmap and some initial functional distinctions—particularly for the cerebellar sites.
The paper’s lead author is MIT postdoc Agata Wolna. Other contributors include Aaron Wright, Colton Casto, Samuel Hutchinson, Benjamin Lipkin, and Evelina Fedorenko. The study appears in the Journal of Neuroscience.
Tracking language
Classic language areas include Broca’s region and other left frontal and temporal cortex locations. Right-hemisphere counterparts have been implicated in social and emotional aspects of language. Prior studies and anecdotal observations have also hinted at additional involved regions, but those signals were typically overlooked or attributed to nonlanguage processes.
Fedorenko recalls that early in her work, consistent activations outside canonical language areas were often dismissed by reviewers. The current study was designed to reassess those data and determine whether peripheral activations reliably relate to language.
The researchers compiled data from 772 individuals scanned in Fedorenko’s lab since 2013. Each participant completed a language localizer task that contrasts real sentences with sequences of nonwords. Regions that show reliably stronger responses to sentences—and that respond across both auditory and visual presentation—were treated as language-sensitive.
“This straightforward paradigm identifies a core language system within individual brains,” Wolna explains.
Unlike many prior studies that apply strict statistical cutoffs, the team relaxed those thresholds and performed focused searches in subcortical and low-signal regions to maximize detection of small or weakly responsive language areas.
“We consistently see the frontal–temporal network, but there is mounting evidence that other regions are also important for language,” Wolna says. “By lowering the threshold and examining areas with weaker MRI signals, we aimed to recover small, dispersed components of a broader language network.”
A widespread network
For approximately 490 participants, the team also had fMRI data from a spatial working memory task, a demanding nonlinguistic challenge that recruits the multiple demand system. Comparing responses across the language and working memory tasks allowed the researchers to assess whether newly found regions are selective for language or instead reflect general cognitive demands.
Among the 17 newly revealed sites, five are in the cerebellum. Earlier work by some team members showed that three cerebellar regions respond during both language tasks and certain nonlinguistic tasks; the current study replicated those findings.
“Areas that respond to both language and other cognitive demands are particularly intriguing because they may integrate information from different cortical systems,” Fedorenko says.
The team also identified language-selective regions in the medial frontal cortex, the inferior surface of the left temporal lobe, the hippocampus, and the amygdala. Future work will test specific functional contributions of these regions across diverse language manipulations and nonlinguistic tasks.
“With these regions reliably mapped, we can now evaluate hypotheses from prior work and more rigorously characterize what these areas do—and whether their functions differ from those of the core language network,” Fedorenko adds.
Funding:
This research was supported by the Simons Center for the Social Brain at MIT, the McGovern Institute, MIT’s Department of Brain and Cognitive Sciences, and the MIT Siegel Family Quest for Intelligence.
Key Questions Answered:
A: This study expands the view of language processing from a narrowly focused set of left-hemisphere hubs to a broader, distributed network. By examining weaker MRI signals that were previously ignored, the researchers identified 17 additional language-responsive areas across the brain, including regions involved in movement coordination and memory.
A: Five of the 17 newly found language sites are in the cerebellum. Some of these cerebellar regions also respond during demanding nonlinguistic tasks, suggesting they may serve integrative roles—routing or coordinating information between language and other high-level systems rather than simply processing words in isolation.
A: No. Although the additional nodes are distributed across several brain systems, they are small in total physical volume. Altogether, they account for less than 5 percent of grey matter volume—about the size of a large strawberry—indicating that language processing remains spatially efficient.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full.
- Additional contextual information was added by staff.
About this language processing research news
Author: Sarah McDonnell
Source: MIT
Contact: Sarah McDonnell, MIT
Image: The image is credited to Neuroscience News
Original Research: Open access. “The extended language network: Language-responsive brain areas whose contributions to language remain to be discovered” by Agata Wolna, Aaron Wright, Colton Casto, Samuel Hutchinson, Benjamin Lipkin and Evelina Fedorenko. Journal of Neuroscience.
DOI: 10.1523/JNEUROSCI.0638-25.2026
Abstract
The extended language network: Language-responsive brain areas whose contributions to language remain to be discovered
Language neuroscience has traditionally emphasized core left frontal and temporal regions and their right-hemisphere counterparts, but many other cortical and subcortical areas have been implicated in linguistic tasks. Evidence for these additional areas has come from a range of paradigms that sometimes confound language with perceptual, motor, or task-related demands.
Using fMRI data from 772 participants (438 females, 334 males) who completed a validated language localizer that isolates language processing from other operations, the authors (a) delineate a comprehensive set of areas that respond reliably to language across written and auditory modalities, and (b) assess the selectivity of these areas for language relative to a demanding nonlinguistic task.
Consistent with earlier suggestions, many regions outside the core fronto-temporal network are engaged during language processing, and most show selectivity for language compared with general task demands. The extended, language-selective network includes areas around the temporal poles, regions in the medial frontal cortex, parts of the hippocampus, and cerebellar sites, among others.
Although these regions are distributed across the brain, the extended language-selective network still comprises a small fraction (<5%) of grey matter volume, challenging the notion that language processing engages the entire brain.
These newly identified language-selective areas provide targets for systematic characterization, enabling researchers to test how their contributions to language differ from those of core language areas.