- Atlas
- Humanities
- Neurolinguistics
Neurolinguistics
Neurolinguistics is the study of how the brain makes language possible.
15 frameworks, from 1861 to today · Linguistics · Humanities
- Localizationism 1861-1980
- Holism 1864-1950
- Connectionist diagram model 1874-1990
- Cortical stimulation mapping 1930-Present
- Clinical-statistical aphasia research 1960-Present
- Lateralization of brain function 1960-1990
- Generative biolinguistics 1965-Present
- Stage-based processing models 1970-2000
- Event-related potential neurolinguistics 1980-Present
- Sign language neurolinguistics 1980-Present
- Connectionist and emergentist modelling 1986-Present
- Functional neuroimaging of language 1988-Present
- Dual-stream model 2000-Present
- Voxel-based lesion-symptom mapping 2001-Present
- Predictive processing 2005-Present
Overview
Neurolinguistics is the study of how the brain makes language possible1. It asks where language lives in the brain, which circuits compute sounds, words, and grammar, what happens when those circuits are damaged, and how the brain of a bilingual, a signer, or a child differs from the textbook case. The field matters because language is one of the most complex things a brain does, and because its breakdown—after stroke, injury, or disease—is one of the most devastating losses a person can suffer. Its history is a long argument between two intuitions: that language is carried by specific, localizable patches of cortex, and that it emerges from the whole brain acting together.
The Localizationist Revolution: Broca, Wernicke, and the Diagram
Before the nineteenth century, the dominant view was holistic. The phrenologists of the early 1800s had mapped mental faculties onto bumps of the skull, but serious medicine dismissed them. The physician Jean-Baptiste Bouillaud argued in the 1820s that articulate speech had a frontal seat, and in 1861 the surgeon Paul Broca examined a patient called Leborgne—known as "Tan" for the only syllable he could produce—who had lost the ability to speak but understood what was said to him. After the man's death, Broca found a lesion in the left frontal lobe1. A second case confirmed it, and Broca concluded that articulate language is seated in the left frontal convolution. The discovery did two things at once: it localized a function, and it revealed that the left hemisphere is special for language in most people2.
In 1874 the young Carl Wernicke extended the picture. He described patients who spoke fluently but nonsensically and could not understand speech, with damage to the posterior part of the left temporal lobe13. Wernicke's insight was not just a second location but a connectionist model: he proposed that a "motor" speech center and a "sensory" speech center were linked by a fibre pathway, and that damage to the pathway produced a third syndrome—conduction aphasia—in which patients could understand and speak but could not repeat what they heard. This was the birth of the diagram model of language in the brain: boxes for centers, arrows for connections, and syndromes predicted from where the damage fell. Ludwig Lichtheim and later Norman Geschwind refined the diagrams, and the "Wernicke–Geschwind" model became the standard teaching picture for a century.
The Holistic Reaction: Hughlings Jackson and the Whole Brain
Not everyone was convinced. The English neurologist John Hughlings Jackson, working from the 1860s onward, argued that the brain is a hierarchy of levels and that language is not a thing stored in a box but a function of the whole nervous system. He distinguished propositional language—the voluntary, creative use of words—from automatic utterances like swearing or singing, and he observed that patients who could not speak could often still sing or curse. Jackson's holism was a minority position for decades, but it kept alive the idea that the lesion method alone could not settle the question. The debate between localizationists and holists—sometimes called the "diagram makers" versus the "noetic" or whole-brain school—ran through the late nineteenth and early twentieth centuries, and it never fully died.
Cortical Stimulation and the Clinical-Statistical Turn
A different kind of evidence arrived with brain surgery. In the 1930s and 1940s, Wilder Penfield and his colleagues at the Montreal Neurological Institute stimulated the exposed cortex of awake patients during operations for epilepsy. Stimulation of certain frontal and temporal sites could arrest speech or evoke vocalizations, and Penfield mapped these "speech areas" directly in the human brain. The method was crude by later standards, but it was the first direct, causal probe of language in the living human cortex, and it confirmed that the left hemisphere was dominant for speech in most people.
Meanwhile, the clinical-statistical approach matured. Rather than reasoning from single famous cases, neurologists began to collect series of patients and ask which lesion sites were statistically associated with which symptoms. The work of the Boston group around Norman Geschwind in the mid-twentieth century, and later the aphasia research of Harold Goodglass and Edith Kaplan, systematized the classification of aphasias and their lesion correlates. The Boston Diagnostic Aphasia Examination, published in 1983, became a standard tool. The logic was still the classical one—lesion site predicts syndrome—but the evidence was now quantitative.
The Cognitive Revolution and the Search for Processing
The 1960s and 1970s brought a challenge from a different direction. The cognitive revolution in psychology, and the rise of generative grammar under Noam Chomsky, reframed language as a computational system in the mind rather than a set of associations4. Neurolinguistics absorbed this: the question shifted from where language is to how it is processed. The generative biolinguistics program, associated with Chomsky and later with researchers like David Caplan, treated language as an innate, modular faculty and asked how its components—phonology, syntax, semantics—are computed in real time. Aphasia was reinterpreted not as a loss of stored words but as a breakdown in specific processing operations.
At the same time, the study of hemispheric specialization took a dramatic turn. In the 1960s, Roger Sperry and his colleagues studied patients who had undergone surgical severing of the corpus callosum to control epilepsy—the "split-brain" patients. These studies, especially the work of Michael Gazzaniga, showed that the two hemispheres could act almost as separate minds: the left hemisphere typically dominated for language, while the right could handle some spatial and emotional tasks but could not speak. The finding cemented the left-hemisphere story but also complicated it, showing that the right hemisphere contributes to language in ways the classical model had ignored—prosody, metaphor, and the pragmatic framing of utterances.
Electrophysiology: Watching Language in Milliseconds
A major methodological leap came from event-related potentials (ERPs)—electrical brain activity time-locked to a stimulus, measured at the scalp with millisecond resolution. In 1980, the Kutas and Hillyard study showed that a semantically anomalous word in a sentence (the famous "He spread the warm bread with socks") produced a distinct negative deflection about 400 milliseconds after the word appeared—the N400. Later work identified a different component, the P600, associated with syntactic anomalies5. ERPs gave neurolinguistics its first fine-grained, online window into language processing, and they established that semantic and syntactic processing could be dissociated in time. The method's limitation—poor spatial resolution—set the stage for the next revolution.
Functional Neuroimaging: Seeing the Language Network
In the 1990s, positron emission tomography (PET) and then functional magnetic resonance imaging (fMRI) allowed researchers to watch the living brain while people read, listened, and spoke. Early studies by Petersen, Raichle, and colleagues in the late 1980s and 1990s mapped word reading and listening to a network including left inferior frontal and posterior temporal regions. The new methods confirmed the classical left-hemisphere language areas but also revealed that language tasks engage distributed networks, not isolated boxes. A landmark finding was that the left inferior frontal region—Broca's area—was not a single "speech center" but a set of functionally distinct subregions, and that the same area was active in tasks far removed from speech production.
The imaging era also transformed the study of sign languages. Work from the 1980s and 1990s, notably by Ursula Bellugi and her colleagues at the Salk Institute, showed that deaf signers with left-hemisphere damage suffered sign-language aphasias closely parallel to spoken-language aphasias, and that the brain's language areas were engaged by sign just as by speech. This was powerful evidence that the language system is organized around linguistic structure, not around sound.
Distributed Networks and the Decline of the Box
By the 2000s, the classical box-and-arrow model was under pressure from several directions. Functional imaging showed that language tasks recruit widespread, overlapping networks; lesion studies showed that similar lesions could produce different symptoms and different lesions similar symptoms. The emerging view was that language is supported by distributed networks—large-scale systems of interacting regions—rather than by discrete centers16. The "dual-stream" model, articulated by Gregory Hickok and David Poeppel in the 2000s, proposed that a dorsal stream maps sound onto articulation and a ventral stream maps sound onto meaning, each involving multiple regions. This was a modern descendant of Wernicke's connectionism, but with the arrows replaced by dynamic, reciprocal networks.
Quantitative Lesion-Symptom Mapping
The lesion method itself was reborn. Instead of single cases, researchers began to map lesions across large groups of patients onto standardized brain images and correlate lesion location with behavioral scores—voxel-based lesion-symptom mapping (VLSM), developed in the early 2000s. This allowed statistical inference about which brain regions are necessary for which functions, and it could disentangle the effects of overlapping lesions that had confounded earlier clinical work. The approach has its own limits—it depends on the lesions available in the patient population—but it restored the lesion method as a rigorous tool and connected it to the network view.
Emergentist and Usage-Based Alternatives
A parallel tradition questioned the modular, innate picture altogether. Connectionist models of language, which model processing as the spread of activation through networks of simple units, were developed from the 1980s onward (the parallel distributed processing program of Rumelhart, McClelland, and colleagues)7. These models suggested that language could emerge from general learning mechanisms operating over large amounts of input, without dedicated language circuits. The usage-based and emergentist tradition in linguistics, associated with researchers like Joan Bybee and Michael Tomasello, argued that grammar is built up from concrete usage events and that the brain's language system is an emergent property of general cognitive and social capacities. This tradition has been influential in developmental neurolinguistics and in explaining how language learning shapes the brain.
Stage-Based Processing and Predictive Processing
The cognitive tradition also produced stage-based models of language processing—the idea that comprehension proceeds through a fixed sequence of stages, from acoustic analysis to phonological, syntactic, and semantic interpretation. These models, often tested with ERPs and reaction times, gave neurolinguistics a fine-grained functional architecture to map onto brain activity. More recently, the predictive processing framework has reframed language comprehension as the brain's continuous effort to predict upcoming input and update its predictions when they fail. On this view, the N400 and P600 components are signatures of prediction error and revision, and the language network is a prediction machine. This framework is still being tested, but it has become one of the most active theoretical currents in the field.
The Present: Networks, Methods, and Open Questions
Today neurolinguistics is a methodologically plural field. It combines lesion studies, cortical stimulation, electrophysiology, functional imaging, and increasingly direct recordings from the brain surface in neurosurgical patients. The dominant theoretical picture is network-based: language is supported by distributed, interacting systems, with the left hemisphere playing a leading but not exclusive role. The classical syndromes remain clinically useful, but they are understood as rough summaries of damage to components of a larger system. The big open questions are the ones the field has carried from the beginning: how the language network develops, how it is shaped by experience (bilingualism, literacy, sign language), how it recovers after damage, and whether the brain's language system is a specialized adaptation or an emergent property of more general capacities. The argument between localization and holism has not been settled so much as transformed—into a debate about how networks of regions, rather than boxes, produce the most human of abilities.
References
- Neurolinguistics (introduction, History, Localizations of language processes), Wikipedia
- Split-brain (History), Wikipedia
- Conduction aphasia (History), Wikipedia
- Generative grammar (introduction, History), Wikipedia
- P600 (neuroscience), Wikipedia
- Large-scale brain network, Wikipedia
- Connectionism, Wikipedia
The frameworks of Neurolinguistics
Each framework is a way of working the field has used: when it held, and the question it answers.
- 1861-1980
- 1864-1950
- 1874-1990
- 1930-Present
- 1960-Present
Clinical-statistical aphasia research
Which lesions cause which aphasias?
- 1960-1990
- 1965-Present
- 1970-2000
- 1980-Present
Sign language neurolinguistics
Does sign language use the same brain areas?
- 1986-Present
Connectionist and emergentist modelling
Can networks learn language from input?
- 1988-Present
Functional neuroimaging of language
Which brain networks activate during language?
- 2000-Present
- 2001-Present
- 2005-Present
How the ideas connect
Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.
Holism reacted against Localizationism
Hughlings Jackson argued from the 1860s that language is a function of the whole nervous system organized in hierarchical levels, directly rejecting Broca's and Wernicke's localization of language to discrete cortical centers.
Carried by John Hughlings Jackson.
Connectionist diagram model grew out of Localizationism
Wernicke built his connectionist diagram directly on Broca's localization of a motor speech center, adding a sensory center and a connecting fibre pathway to predict conduction aphasia.
Carried by Carl Wernicke.
Voxel-based lesion-symptom mapping grew out of Clinical-statistical aphasia research
Voxel-based lesion-symptom mapping generalized the clinical-statistical logic of correlating lesion site with syndrome, replacing patient series with voxelwise statistical maps across large samples.
Dual-stream model grew out of Functional neuroimaging of language
The dual-stream model grew out of functional imaging evidence that language recruits distributed networks, recasting the classical centers as dorsal and ventral processing streams.
Carried by Gregory Hickok; David Poeppel.
Cortical stimulation mapping influenced by Localizationism
Penfield's cortical stimulation mapping tested and confirmed the localizationist claim that specific left-hemisphere sites carry speech, using direct electrical probes of the exposed cortex.
Carried by Wilder Penfield.
Clinical-statistical aphasia research grew out of Localizationism
The clinical-statistical aphasia research of the Boston group kept the localizationist logic that lesion site predicts syndrome but replaced single case reports with quantitative series of patients.
Carried by Norman Geschwind; Harold Goodglass.
Lateralization of brain function influenced by Localizationism
Broca's finding that articulate language sits in the left frontal convolution established left-hemisphere dominance for language, the claim split-brain research then probed and complicated.
Carried by Roger Sperry; Michael Gazzaniga.
Sign language neurolinguistics influenced by Localizationism
Sign language neurolinguistics tested the localizationist prediction that left-hemisphere damage produces aphasia, finding sign-language aphasias parallel to spoken-language ones in deaf signers.
Carried by Ursula Bellugi.
Functional neuroimaging of language influenced by Localizationism
Early functional imaging studies tested the classical left-hemisphere language areas, confirming Broca's and Wernicke's regions while revealing them as parts of distributed networks.
Carried by Marcus Raichle.
Clinical-statistical aphasia research influenced by Connectionist diagram model
The Wernicke-Geschwind diagram supplied the syndrome categories and lesion-symptom predictions that clinical-statistical aphasia research then tested across patient series.
Carried by Norman Geschwind.
Stage-based processing models reacted against Connectionist diagram model
Stage-based processing models replaced the box-and-arrow diagram's static centers with sequential real-time operations, reinterpreting aphasia as breakdown in processing stages rather than loss of stored words.
Stage-based processing models influenced by Generative biolinguistics
The generative view of language as modular computation carried into stage-based processing models, which asked how phonological, syntactic and semantic operations unfold in real time.
Carried by David Caplan.
Event-related potential neurolinguistics influenced by Stage-based processing models
ERP neurolinguistics used the stage-based distinction between semantic and syntactic processing as the hypothesis that the N400 and P600 components were designed to test in milliseconds.
Carried by Marta Kutas; Steven Hillyard.
Voxel-based lesion-symptom mapping influenced by Functional neuroimaging of language
Voxel-based lesion-symptom mapping borrowed the voxelwise statistical framework of functional neuroimaging and applied it to lesion data instead of activation data.
Predictive processing influenced by Connectionist and emergentist modelling
Emergentist and connectionist modelling supplied predictive processing with the idea that hierarchical networks learn by minimizing prediction error rather than storing rules.
Beyond the field
Where Neurolinguistics's ideas came from in other fields, and what they led to there: the strongest 12 of 64 links.
Holism reacted against Localizationism in Cognitive Neuroscience
Holism arose as a direct attack on localizationism: Hughlings Jackson and later critics argued that language and other higher functions are distributed and that focal lesions produce effects through the whole system, defining holism against the centre doctrine.
Carried by John Hughlings Jackson.
Cortical stimulation mapping influenced by Experimental Cortical Excitability in Epilepsy
Cortical stimulation mapping grew directly out of earlier experimental work on cortical excitability, in which electrical stimulation of the cortex was shown to evoke movements and sensations; Wilder Penfield and colleagues built their mapping technique on that experimental tradition.
Carried by Wilder Penfield.
Clinical-statistical aphasia research grew out of Anatomical-Clinicopathological Method in Neurology
Clinical-statistical aphasia research grew out of the anatomical-clinicopathological method: it kept the correlation of lesion site with clinical deficit but replaced single case reports with group statistics and quantified lesion analysis.
Generative biolinguistics led to Neurophonetics in Phonetics
Neurophonetics grew out of generative biolinguistics, taking its questions about the mental representation of phonological structure and seeking their neural basis. The programme of locating grammar in the brain carried directly into neurophonetic research.
Carried by Noam Chomsky.
Generative biolinguistics grew out of Universal grammar in First Language Acquisition
Universal grammar is the core postulate of the biolinguistic programme: Chomsky's claim that the child's innate language faculty constrains possible grammars. Biolinguistics grew out of it as the broader biological research programme around that postulate.
Carried by Noam Chomsky.
Stage-based processing models influenced by Information processing in Cognitive Psychology
Information processing as a general framework supplied the serial-stage conception and the assumption of discrete processing levels that stage-based language models adopted. The language models are applications of that general architecture.
Stage-based processing models grew out of Derivational theory of complexity in Psycholinguistics
The derivational theory of complexity, that processing difficulty tracks the number of grammatical transformations, is the direct ancestor of stage-based processing models. Those models generalised the idea into multi-stage architectures for comprehension and production.
Stage-based processing models influenced by Information-Processing Cognitive Psychology in Cognitive Neuroscience
Stage-based processing models of language take the box-and-arrow architecture of information-processing cognitive psychology: input, successive internal stages, output. That architecture, with its serial stages and limited-capacity stores, is what the language models instantiate.
Carried by Ulric Neisser; Cognitive Psychology (1967).
Stage-based processing models influenced Staged production models in Psycholinguistics
Staged production models in psycholinguistics grew out of the stage-based processing models of the 1970s, which decomposed production into discrete levels such as conceptualization, formulation and articulation. The later models kept this serial architecture while adding monitoring and feedback components.
Carried by Willem Levelt; Speaking: From Intention to Articulation (1989).
Event-related potential neurolinguistics influenced by Electroencephalography and Evoked-Potential Analysis in Biomedical Signal Processing
Event-related potential neurolinguistics depends on EEG and evoked-potential averaging, the technique developed for recording and extracting time-locked brain responses. The language work consists of applying that signal-processing method to linguistic stimuli.
Connectionist and emergentist modelling influenced by Connectionism and Neural Networks in Machine Learning
Connectionist and emergentist modelling of language uses the neural-network methods of connectionism: distributed units, weighted connections and learning by error correction. The language models are applications of that machinery.
Dual-stream model influenced by Neuroimaging of language in Psycholinguistics
Hickok and Poeppel's dual-stream model was built directly on functional neuroimaging evidence (PET and fMRI studies of speech perception and production) that had accumulated through the 1990s. The model's dorsal and ventral streams were defined by the activation patterns these imaging studies reported.
Carried by Gregory Hickok; David Poeppel.

Neurolinguistics is a planet in Nousfarer
Its city has a tower for each of these 15 frameworks. Land, walk in, and climb one concept at a time.
Land on NeurolinguisticsRelated fields
Fields its ideas are linked to
- Psycholinguistics 14 links
- Cognitive Neuroscience 8 links
- Sign Language Linguistics 7 links
- First Language Acquisition 4 links
- Neural Engineering 3 links
- Neuroimaging 3 links
- Epilepsy 3 links
- Neuropsychology 3 links
- Language And Literacy Education 2 links
- Phonetics 2 links
- Behavioral Neuroscience 2 links
- Systems Neuroscience 2 links
More in Linguistics
- Corpus Linguistics
- Dialectology
- Discourse Analysis
- First Language Acquisition
- Historical Linguistics
- Language Documentation And Revitalization
- Lexicography
- Linguistic Typology
- Linguistics
- Morphology
- Phonetics
- Phonology
- Pragmatics
- Psycholinguistics
- Second Language Acquisition
- Semantics
- Sign Language Linguistics
- Sociolinguistics
- Syntax
- Writing Systems
See where Neurolinguistics sits on the Map of Human Knowledge