Fact-checked by Grok 2 weeks ago

Language processing in the brain

Language processing in the brain encompasses the neural mechanisms that enable humans to comprehend, produce, and manipulate linguistic information, primarily involving a distributed network in the left cerebral hemisphere.^[1] This process relies on specialized regions such as Broca's area in the inferior frontal gyrus (Brodmann areas 44 and 45), which supports speech production and grammatical processing, and Wernicke's area in the posterior superior temporal gyrus (Brodmann area 22), which is essential for language comprehension and semantic interpretation.^[2]^[3] These areas are interconnected through white matter tracts forming dorsal and ventral streams: the dorsal pathway, via the arcuate fasciculus, handles phonological mapping and sound-to-articulation transformations, while the ventral pathway, involving the inferior fronto-occipital fasciculus, facilitates semantic processing and meaning retrieval.^[1] The foundational understanding of these mechanisms emerged in the 19th century through clinical observations of aphasia patients. In 1861, French neurologist Paul Broca identified a lesion in the left inferior frontal gyrus of a patient with expressive aphasia, establishing Broca's area as critical for articulate speech production.^[2] Over a decade later, in 1874, German neurologist Carl Wernicke described fluent but incomprehensible speech resulting from damage to the posterior superior temporal region, linking Wernicke's area to receptive language functions and proposing an early connectionist model between these regions.^[3] Subsequent neuroimaging studies, including functional MRI, have confirmed and expanded this classical model, revealing bilateral involvement in some aspects of processing and a broader perisylvian network that integrates auditory, motor, and cognitive systems.^[1]^[4] Modern neurobiological models emphasize causal mechanisms grounded in neural dynamics, such as spiking neurons, synaptic plasticity, and attractor states within cortical microcircuits.^[4] The dual-stream framework, refined through lesion and electrophysiological studies, posits that the dorsal stream supports interface between phonological and articulatory representations, while the ventral stream enables comprehension via lexical-semantic access.^[1] Language processing also involves unification operations for combining words into meaningful sentences, occurring rapidly in prefrontal and temporal regions on timescales of hundreds of milliseconds.^[4] Disruptions, as seen in aphasias, highlight the system's vulnerability, yet plasticity allows recovery through compensatory recruitment of homologous right-hemisphere areas or adjacent networks.^[1] Ongoing research integrates computational simulations with invasive recordings to elucidate how these mechanisms adapt during learning and bilingualism.^[4]

Neuroanatomy of Language

Core Language Areas

The core language areas are primarily located in the left cerebral hemisphere and form the foundational neural substrate for language functions in most individuals. These regions, identified through historical lesion studies and modern neuroimaging, include Broca's area, Wernicke's area, and the angular gyrus, each contributing distinct aspects of language processing. Subcortical structures such as the basal ganglia and thalamus provide essential support for coordinating these cortical activities. Language processing exhibits strong lateralization, with approximately 90-95% of right-handed individuals showing left-hemisphere dominance for core language functions, though left-handers display greater variability, with right-hemisphere involvement in up to 30% of cases depending on the degree of handedness.^[5]^[6] Broca's area, situated in the left inferior frontal gyrus (Brodmann areas 44 and 45), plays a central role in speech production, syntactic processing, and the articulation of complex grammatical structures. It was first identified in 1861 by French physician Paul Broca through postmortem examinations of patients with non-fluent aphasia, such as the famous case of "Tan" (Louis Leborgne), whose lesion in this region correlated with severe expressive language deficits despite preserved comprehension. Lesion studies and functional imaging confirm that damage to Broca's area disrupts motor planning for speech and hierarchical syntactic operations, often resulting in telegraphic speech characterized by short, agrammatic phrases.^[2]^[7]^[8] Wernicke's area, located in the posterior superior temporal gyrus (Brodmann area 22) of the left hemisphere, is essential for language comprehension, semantic processing, and the interpretation of auditory verbal input. Discovered in 1874 by German neurologist Carl Wernicke, it was characterized through observations of patients exhibiting fluent but nonsensical speech (Wernicke's aphasia) due to lesions in this region, highlighting its role in mapping sounds to meaning. Neuroimaging evidence supports its involvement in phonological and lexical-semantic analysis, where it integrates auditory signals to form coherent linguistic representations.^[3]^[9]^[10] The angular gyrus, found in the left inferior parietal lobe (Brodmann area 39), facilitates the integration of visual and auditory information crucial for reading, writing, and multimodal language tasks. It serves as a convergence zone for associating orthographic, phonological, and semantic elements, enabling processes like word recognition and cross-modal mapping in literacy. Lesion studies demonstrate that angular gyrus damage impairs reading comprehension and mathematical-spatial language integration, underscoring its role in higher-order associative functions beyond basic speech or audition.^[11]^[12]^[13] Subcortically, the basal ganglia and thalamus modulate language by supporting motor sequencing, procedural learning, and attentional gating of linguistic elements. The basal ganglia, including structures like the caudate nucleus and putamen, contribute to the rhythmic timing and sequential organization of speech production, as evidenced by their involvement in disorders like Parkinson's disease, which disrupt fluent articulation. The thalamus acts as a relay hub, filtering sensory inputs to cortical language areas and enhancing selective attention during comprehension tasks, with its anterior and pulvinar nuclei showing activation in verbal fluency and semantic retrieval paradigms.^[14]^[15]^[16]

Supporting Networks and Connectivity

The arcuate fasciculus is a major white matter tract that connects Broca's area in the frontal lobe to Wernicke's area in the temporal lobe, facilitating the mapping of phonological representations during language processing.^[17] This pathway supports the repetition and production of speech sounds, with disruptions leading to conduction aphasia, characterized by impaired phonological repetition despite preserved comprehension and fluent speech.^[18] Diffusion tensor imaging (DTI) studies since the 2000s have revealed bidirectional connectivity along the arcuate fasciculus, enabling efficient information flow between auditory and motor regions for phonological integration.^[19] The superior longitudinal fasciculus (SLF) forms a key component of the dorsal language pathway, linking parietal and frontal regions to support articulation and sensory-motor mapping in speech production.^[20] DTI evidence indicates that the SLF provides bidirectional connections that coordinate articulatory planning, with integrity correlating to repetition and naming abilities post-stroke.^[21] Complementing this, the inferior fronto-occipital fasciculus (IFOF) constitutes the ventral pathway, connecting frontal and occipital-temporal areas to underpin semantic processing and meaning retrieval.^[22] Lesion and tractography studies highlight the IFOF's role in integrating lexical semantics, with left-hemisphere damage impairing object naming and conceptual associations.^[23] The corpus callosum enables interhemispheric transfer of linguistic information, particularly relevant for bilingual individuals who exhibit enhanced callosal microstructure to manage dual-language systems.^[24] In recovery from brain injury, DTI reveals that preserved callosal integrity facilitates language reorganization by promoting compensatory activation across hemispheres.^[25] Additionally, the default mode network (DMN) overlaps with language networks during narrative processing, supporting internal discourse and story comprehension through connectivity in medial prefrontal and posterior cingulate regions.^[26] This integration allows for the construction of coherent mental narratives, as evidenced by functional connectivity analyses during semantic and episodic language tasks.^[27]

Historical and Current Models

Early Neurolinguistic Models

Early neurolinguistic models emerged in the mid-19th century through clinico-pathological studies of aphasia, driven by the localizationist principle that specific brain regions govern distinct cognitive functions. In 1861, Paul Broca identified a lesion in the left inferior frontal gyrus of a patient with severe speech production deficits but preserved comprehension, establishing this area—now known as Broca's area—as critical for articulate speech. This finding, based on autopsy correlations, marked a pivotal shift toward cerebral localization of language, influencing subsequent aphasiology by emphasizing hemispheric asymmetry, particularly left-hemisphere dominance.^[28]^[9] Building on Broca's work, Carl Wernicke in 1874 described a sensory form of aphasia linked to lesions in the posterior superior temporal gyrus, termed Wernicke's area, characterized by fluent but semantically empty speech and impaired auditory comprehension. Patients with Wernicke's aphasia produce paraphasic errors, such as neologisms or jargon, while maintaining normal prosody and grammar, yet struggle to understand spoken or written language due to disrupted sound-to-meaning mapping. In the 1960s, Norman Geschwind synthesized these observations into the Wernicke-Geschwind model, positing a hierarchical pathway for language processing: auditory input from the primary auditory cortex projects to Wernicke's area for comprehension, then via the arcuate fasciculus—a white matter tract—to Broca's area for speech production. Damage to the arcuate fasciculus was theorized to cause conduction aphasia, featuring intact comprehension and fluent speech but severe repetition deficits and phonemic errors, as the disconnection prevents phonological relay between comprehension and production centers.^[29]^[9]^[30]^[31] This model embodied a modular view of language, treating comprehension and production as discrete, localized modules connected by dedicated conduits, which simplified explanations of aphasia syndromes but overlooked broader cognitive integrations. Critiques highlight its oversimplification of semantics, as it primarily addressed phonological and syntactic processing while neglecting how meaning is constructed across distributed networks. Additionally, the model's strict left-lateralized focus ignored evidence of bilateral contributions to language, particularly in prosody and semantics, and failed to account for variability in lesion outcomes, where arcuate damage does not uniformly produce conduction aphasia. These limitations, revealed through later anatomical and functional studies, underscored the need for more dynamic frameworks beyond lesion-based localizationism.^[30]^[32]^[33]^[34]

Modern Neurolinguistic Models

Modern neurolinguistic models have evolved to incorporate interactive and dynamic processes, addressing limitations of earlier lesion-based approaches by integrating neuroimaging data and computational simulations. These frameworks emphasize parallelism and bidirectional interactions in language processing, revealing how the brain handles comprehension and production through overlapping neural networks rather than isolated modules.^[35] The dual-stream model, proposed by Hickok and Poeppel in the 2000s, posits two partially segregated pathways for speech and language: a ventral stream primarily supporting comprehension by mapping sound to meaning, and a dorsal stream facilitating production and phonological mapping for articulation. This model, informed by functional imaging studies, highlights bilateral processing in superior temporal regions for early auditory analysis, with asymmetries emerging in higher-level functions. Updates from magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) further demonstrate parallel activation across streams during real-time language tasks, enabling flexible integration of sensory input and motor output.^[36]^[37] Predictive coding theories, advanced by Friston in the 2010s, frame language processing as Bayesian inference where the brain generates top-down predictions about incoming linguistic input to minimize prediction errors. In this hierarchical scheme, cortical layers anticipate syntactic and semantic structures, with evidence from fMRI showing reduced neural activity for expected words and heightened responses to surprises, supporting efficient communication. Applied to language, these models explain phenomena like syntactic priming, where prior exposure facilitates prediction of grammatical continuations.^[38]^[39] The discovery of mirror neurons in the 1990s by Rizzolatti and colleagues in premotor areas like the inferior frontal gyrus has informed modern views on action-language interfaces, suggesting these cells activate both during action observation and execution, potentially linking motor simulation to linguistic comprehension. fMRI and MEG studies in the 2000s and beyond have extended this to language, showing mirror neuron involvement in processing action verbs and metaphors, though their precise role remains debated as facilitative rather than essential.^[40]^[41] Connectionist models, utilizing artificial neural networks to simulate language acquisition, challenge strict modularity by demonstrating emergent linguistic abilities through distributed learning rather than innate rules. These networks, trained on probabilistic input, replicate patterns like past-tense formation without explicit programming, aligning with brain-like parallel distributed processing. Critiques of modularity, bolstered by neuroplasticity evidence from recovery studies, argue that language networks reorganize dynamically across development and injury, integrating with broader cognitive systems rather than operating in isolation.^[42]^[43]^[44] In the 2020s, AI-inspired models have advanced syntax prediction in neurolinguistics, with large language models (LLMs) like GPT-3 exhibiting brain-like hierarchical representations that correlate with fMRI activations during sentence processing. These approaches reveal how predictive mechanisms in artificial networks mirror cortical hierarchies for syntactic disambiguation, offering insights into plasticity and offering testable hypotheses for human experiments.^[45]^[46]

Auditory Processing Streams

Ventral Stream Functions

The ventral stream in the brain, often termed the "what" pathway, primarily facilitates the mapping of auditory input to meaning during language comprehension, involving a network of temporal lobe regions that process sounds from basic acoustic features to complex semantic representations.^[20] This pathway originates in early auditory areas and extends anteriorly through the superior temporal gyrus (STG) and sulcus (STS), progressing to the middle temporal gyrus (MTG) and anterior temporal lobe for higher-level integration.^[47] In contrast to the dorsal stream's role in sound-to-articulation mapping for production, the ventral stream focuses on recognition and interpretation without direct involvement in motor output.^[20] Sound-to-lexicon mapping occurs progressively along the ventral stream, beginning in the anterior superior temporal sulcus (aSTS), where phonetic representations are formed from incoming speech signals, and extending to the MTG, which supports lexical access by linking these representations to word meanings.^[48] This process involves hierarchical abstraction, with posterior STS regions encoding spectrotemporal features of sounds and more anterior MTG areas integrating them into recognizable lexical items.^[49] For instance, neuroimaging studies show that aSTS activation correlates with the resolution of phonetic ambiguities in spoken words, facilitating the transition from auditory input to stored lexical knowledge.^[50] Sound recognition in the ventral stream unfolds in distinct stages, starting with the extraction of acoustic features such as frequency and amplitude modulations in primary auditory cortex, followed by phonetic feature assembly in the posterior STG and STS.^[47] These stages enable the transformation of raw sound waves into invariant phonetic units, with increasing tolerance to acoustic variability (e.g., speaker differences) as signals propagate anteriorly along the pathway.^[51] This hierarchical processing ensures robust recognition of speech elements before lexical integration.^[47] Neural adaptations of lexical access models, such as the cohort theory, illustrate how the ventral stream handles word recognition by activating multiple candidate words based on initial phonetic input and narrowing them via contextual cues.^[52] In this framework, posterior temporal regions initiate a "cohort" of phonetically similar activations, while anterior MTG regions resolve competition through semantic and syntactic constraints, mirroring behavioral predictions of the theory at the neural level.^[53] Functional MRI evidence supports this, showing graded activation patterns along the ventral stream during tasks requiring unique word identification from partial acoustic information.^[54] At the sentence level, the ventral stream contributes to comprehension by integrating syntactic structure and semantic content through connections between the temporal lobe and ventrolateral prefrontal cortex.^[20] This pathway supports the unification of lexical items into coherent meanings, with MTG and inferior frontal gyrus facilitating both local phrase building and global thematic role assignment.^[48] For example, during processing of ambiguous sentences, ventral stream activity resolves semantic dependencies, ensuring interpretive coherence.^[55] The ventral stream exhibits bilaterality, with the left hemisphere dominating lexical and syntactic processing, while the right hemisphere contributes to prosody and broader contextual integration.^[56] Right ventral temporal regions, including the STS, enhance comprehension of intonational cues and emotional tone in speech, aiding disambiguation in naturalistic contexts.^[57] Evidence from split-brain studies demonstrates that isolated right hemispheres can process prosodic elements and simple semantic relations, underscoring the pathway's distributed nature.^[58] Disruptions to ventral stream regions, as seen in semantic dementia, lead to profound deficits in conceptual knowledge and word meaning retrieval while sparing phonological and syntactic abilities.^[59] Atrophy in the anterior temporal lobes, particularly the ATL, impairs multimodal semantic associations, resulting in anomia and category-specific comprehension failures.^[60] Lesion studies confirm that ventral pathway damage selectively hinders lexico-semantic access, with preserved dorsal functions allowing fluent but empty speech output.^[59] Recent neuroimaging research highlights the ventral stream's role in predictive semantics, where anterior temporal regions generate expectations about upcoming words based on prior context, facilitating efficient comprehension.^[61] For instance, fMRI studies show enhanced MTG activity during predictive integration in narratives, with top-down signals from prefrontal areas modulating ventral processing to anticipate meanings.^[61] This predictive mechanism, supported by hierarchical models, reduces processing load in real-time language use.^[48]

Dorsal Stream Functions

The dorsal stream in the auditory processing of language serves as the primary pathway for mapping acoustic speech signals to articulatory motor representations, facilitating the transformation of sound into action. This "how" pathway connects the posterior superior temporal gyrus (pSTG) to the premotor cortex through the dorsal branch of the arcuate fasciculus, enabling speech production planning by integrating auditory input with motor output for phonological encoding and articulation sequencing.^[20]^[62] In this process, the stream supports the online adjustment of speech motor commands based on auditory feedback, ensuring precise coordination during verbal output.^[35] A key function of the dorsal stream involves phonological working memory, particularly through syllable rehearsal mechanisms in the left inferior parietal lobule (IPL), which acts as a temporary buffer for maintaining and manipulating speech sounds. This region, part of the phonological loop, relies on subvocal articulation to refresh decaying phonological traces, allowing for the rehearsal of verbal sequences essential for tasks like repetition and sentence construction.^[63] The IPL's role in this rehearsal process highlights the dorsal stream's contribution to short-term storage of sublexical units, distinct from semantic processing.^[64] The dorsal stream also underpins vocal mimicry and self-monitoring during speech production, utilizing efference copies—internal predictions of sensory consequences from motor commands—to distinguish self-generated sounds from external auditory input. These efference signals, generated in frontal regions and propagated through the dorsal pathway, suppress auditory cortex responses to one's own voice, enabling real-time error detection and correction.^[35] Furthermore, the stream integrates auditory phonemes with visual cues from lip movements, providing a neural basis for phenomena like the McGurk effect, where conflicting audiovisual inputs fuse into a unified percept that influences motor planning via posterior superior temporal sulcus projections to premotor areas.^[65]^[66] In addition to transient processing, the dorsal stream contributes to long-term phonological storage by supporting lexicon retrieval for production, with the supramarginal gyrus serving as a dorsal phonological lexicon that maps conceptual representations to articulatory forms. Disruptions to this pathway, such as damage to the arcuate fasciculus, manifest in apraxia of speech, characterized by impaired planning and execution of speech movements despite intact muscle control, leading to effortful, groping articulations.^[67] Recent advances in ultrasound tongue imaging have revealed the dorsal stream's adaptability in accent learning, demonstrating neural plasticity in motor reprogramming as learners adjust articulatory gestures to novel phonological patterns through visual biofeedback.^[68]^[69]

Integration with Linguistics and Cognition

Neurolinguistic Theories

Neurolinguistic theories seek to integrate insights from neuroscience with linguistic principles to explain how the brain processes syntax, semantics, and universal linguistic structures. These theories draw on brain imaging, lesion studies, and genetic evidence to test hypotheses about innate versus learned language mechanisms, often challenging or refining classical linguistic models. For instance, Noam Chomsky's concept of universal grammar (UG), which posits an innate biological endowment for language acquisition including recursive syntax, has been examined through neurogenetic links, particularly the FOXP2 gene discovered in the 1990s as a key regulator of speech and grammatical processing.^[70] Mutations in FOXP2, identified in families with developmental verbal dyspraxia, disrupt orofacial motor control and syntactic abilities, providing genetic evidence for a biological basis of UG-like structures, though the gene's role extends beyond syntax to broader vocal learning circuits. Neural evidence supports specific syntactic operations central to UG, such as the "merge" mechanism, which builds hierarchical phrase structures. Functional magnetic resonance imaging (fMRI) studies show that processing hierarchical dependencies in sentences activates subregions of the left inferior frontal gyrus (IFG), particularly Brodmann areas 44 and 45 (Broca's area), with greater activation for deeper embeddings compared to flat structures, indicating a dedicated neural computation for recursion.^[71] Lesions or disruptions in the left IFG impair hierarchical syntax comprehension while sparing simpler associations. In semantics, the temporal lobes, especially the anterior temporal lobe (ATL) and superior temporal gyrus, encode word meanings in distributed patterns that reflect the distributional hypothesis—that semantic similarity arises from contextual co-occurrences in language use. Voxel-wise modeling of natural language narratives via fMRI reveals semantic maps in the ATL mirroring vector-space representations from distributional models, where words like "carnivore" cluster near related concepts based on usage patterns, supporting a usage-driven rather than purely innate semantic foundation. Embodied cognition theories further bridge linguistics and neuroscience by proposing that language comprehension is grounded in sensorimotor experiences, rather than abstract symbols alone. For example, processing action verbs like "kick" activates motor cortex regions corresponding to leg movements, as shown in transcranial magnetic stimulation (TMS) and fMRI experiments, suggesting meanings are simulated through bodily states to facilitate understanding.^[72] This grounding challenges disembodied views, emphasizing how sensorimotor simulations in premotor and somatosensory areas integrate with linguistic processing for contextual interpretation. Usage-based models critique Chomskyan innatism by arguing that linguistic structures emerge from statistical patterns in input, without requiring a domain-specific UG module; neurolinguistic evidence from electroencephalography (EEG) shows that frequency of exposure modulates neural responses to grammatical constructions, supporting emergentist accounts over hardwired rules. These models highlight neuroplasticity in perisylvian networks, where repeated usage strengthens connections for syntax and semantics via Hebbian learning. Recent 2020s debates on compositionality— the principle that complex meanings arise systematically from simpler parts—have been informed by BERT-like transformer models, which approximate human-like semantic processing but reveal limitations in true hierarchical generalization. Neuroimaging alignments between transformer activations and brain data show that models like BERT capture distributional semantics in temporal regions but struggle with novel recombinations, prompting questions about whether human compositionality relies on innate biases or learned approximations.^[73] Studies decoding fMRI responses during compositional tasks indicate that the left IFG and ATL support flexible meaning integration beyond static embeddings, suggesting neural mechanisms exceed current AI capabilities in handling ambiguity and context.^[74] These findings refine neurolinguistic theories by integrating computational modeling with brain data, highlighting ongoing tensions between innatist and experiential accounts, with 2025 studies further exploring AI-brain alignments for predictive processing.

Language Evolution and Development

The evolution of language processing in the human brain involved key genetic and neuroanatomical milestones that distinguished Homo sapiens from other primates. Mutations in the FOXP2 gene, essential for vocal motor control and sequencing of orofacial movements, emerged after the human-chimpanzee divergence approximately 6–7 million years ago, with two amino acid substitutions shared with Neandertals becoming fixed prior to their divergence around 500,000–800,000 years ago, aligning with early Homo species and potentially enabling complex speech production.^[75] Concurrently, the human brain underwent significant restructuring, including a taller frontal lobe and anterior-medial reorientation of the temporal poles between 100,000 and 35,000 years ago, which enlarged and refined frontotemporal regions critical for integrating auditory and articulatory aspects of language.^[76] These changes supported the phylogenetic origins of language by enhancing neural circuits for symbolic communication, as evidenced by fossil endocasts showing globularization of the brain shape during this period.^[76] Cultural and genetic interactions further shaped language evolution through mechanisms like the Baldwin effect, where learned behaviors transmitted culturally become genetically assimilated over generations, accelerating adaptation. In language contexts, this process may have genetically encoded predispositions for acquiring universal grammatical features after initial cultural transmission refined them across populations.^[77] Recent 2020s comparative neuroimaging underscores proto-language precursors in great apes, revealing experience-dependent plasticity in language-trained individuals, such as human-like asymmetries in the superior longitudinal fasciculus and expansions in Brodmann area 44 (a Broca's homolog), suggesting foundational networks for symbolic communication that prefigure human capabilities.^[78] For instance, studies using diffusion tensor imaging on chimpanzees and bonobos trained in lexigrams show enhanced connectivity in auditory-motor pathways, indicating neural adaptations akin to early human language hubs.^[78] Ontogenetically, language processing develops through sequential stages in children, leveraging high neural plasticity during early life. Preverbal communication begins with gestures and cooing from birth to 6 months, transitioning to canonical babbling around 6–12 months, where neuroimaging reveals activation of dorsal stream components, including motor regions in Broca's area and the cerebellum, as infants link auditory perception to vocal production.^[79] This progresses to holophrastic single-word utterances (12–18 months), two-word combinations (18–24 months), and full syntax acquisition by 3–4 years, with expanding vocabulary and grammatical complexity reflecting maturation of frontotemporal connectivity.^[80] The critical period hypothesis asserts that this plasticity peaks in childhood, declining sharply after age 7–8, as demonstrated by studies on late second-language learners, such as Johnson and Newport (1989), who found that ultimate attainment in L2 grammar declines after age ~7, with immigrants exposed after this age showing non-native-like proficiency despite immersion, highlighting enduring neural constraints on language acquisition.^[81]

Multimodal Language Processing

Sign Language Processing

Sign language processing in the brain relies on visual-manual modalities rather than auditory-oral ones, yet it recruits homologous neural regions to those involved in spoken language, demonstrating significant overlap in core linguistic functions. Deaf signers activate left perisylvian areas, including homologs of Broca's and Wernicke's areas, for processing sign syntax and semantics, such as grammatical structure and meaning integration. Functional neuroimaging studies confirm that these regions support combinatorial processing in sign languages like American Sign Language (ASL), mirroring their role in spoken syntax. The visual ventral stream, spanning occipital-temporal regions, plays a crucial role in sign recognition by discriminating linguistic handshapes and movements from non-linguistic gestures, with early neural tuning observed around 80-120 ms post-stimulus. Additionally, the bilateral superior temporal sulcus integrates manual signs with non-manual features, such as facial expressions and head movements, essential for conveying prosody and syntax in sign languages. fMRI evidence from deaf signers viewing signs shows activation patterns in bilateral fronto-temporal networks (left-dominant) that closely resemble those during spoken language comprehension in hearing individuals, underscoring shared neural substrates for linguistic decoding across modalities. A modality-independent core language network, involving left anterior temporal lobe and ventromedial prefrontal cortex, supports universal grammatical computations in both sign and speech, suggesting that abstract linguistic principles operate beyond sensory input. Disruptions from left hemisphere lesions produce sign aphasias that parallel spoken types, including phonological errors (e.g., handshape substitutions) and agrammatism (telegraphic signing), as documented in case studies of deaf signers with focal brain damage.^[82] Recent studies highlight neural plasticity in late learners of sign language, even into adulthood. For instance, hearing adults undergoing intensive ASL training over eight months exhibit dynamic reorganization, with increased activation in left perisylvian and visual processing areas, alongside shifts in functional connectivity that adapt to linguistic demands.^[83] Short-term training in late learners also recruits frontal and parietal regions similarly to native signers, indicating compensatory plasticity that enhances sign comprehension despite delayed exposure.^[84] These findings emphasize the brain's adaptability, though late acquisition may involve greater reliance on visual and right-hemisphere resources compared to early bilinguals.^[85]

Written Language Processing

Written language processing involves the neural mechanisms that enable the recognition, comprehension, and production of orthographic symbols, distinct from spoken language pathways. Reading primarily engages the ventral occipito-temporal cortex, where visual input is transformed into meaningful linguistic representations, while writing recruits motor planning areas to generate graphemic output. These processes rely on specialized brain regions that develop through literacy acquisition and exhibit variations across writing systems.^[86] A key structure in reading is the visual word form area (VWFA), located in the left fusiform gyrus, which supports invariant recognition of words regardless of font, case, or size. This region becomes tuned to orthographic forms during reading development, facilitating rapid word identification by enhancing sensitivity to recurring visual patterns in print. Functional imaging studies confirm the VWFA's role in abstract orthographic processing, with activation peaking around 150-200 milliseconds post-stimulus onset.^[87]^[88]^[89] The dual-route model describes two primary pathways for reading: the lexical route, which accesses stored word representations directly via the VWFA for familiar words, and the sublexical route, which assembles pronunciations through grapheme-to-phoneme conversion for unfamiliar or pseudowords. The lexical path involves occipito-temporal regions linking orthography to semantics, while the sublexical path engages superior temporal and parietal areas for phonological mapping. Neuroimaging evidence supports this framework, showing distinct activation patterns for exception words (lexical) versus regular nonwords (sublexical) during reading aloud.^[90]^[91] Grapheme-to-phoneme conversion, central to the sublexical route, implicates the superior temporal gyrus in integrating visual letter forms with phonological codes, often within 100-200 milliseconds of visual presentation. This process is modulated by orthographic depth, with shallower systems (e.g., consistent letter-sound mappings) relying more on posterior superior temporal activation. Disruptions in these pathways, such as reduced integrity of the left arcuate fasciculus—a white matter tract connecting temporal and frontal regions—are linked to dyslexia, impairing phonological decoding and reading fluency across languages.^[92]^[93]^[94]^[95]^[96] Writing engages Exner's area in the left premotor cortex (posterior middle frontal gyrus, Brodmann area 6), which stores motor programs for grapheme production and coordinates hand movements during script generation. Activation in this region increases with visual letter presentation, suggesting it integrates orthographic planning with motor execution. In dyslexia, arcuate fasciculus anomalies further compromise writing by disrupting connections between phonological and motor areas, leading to graphemic errors.^[97]^[98]^[99] Cross-linguistic variations highlight script-specific neural adaptations: alphabetic systems emphasize grapheme-to-phoneme mapping in superior temporal regions, whereas logographic scripts like Chinese recruit additional left middle frontal and inferior parietal areas for visual-semantic processing of characters. Bilingual studies show overlapping ventral occipito-temporal involvement but greater fusiform selectivity for logographs due to their morphological complexity. Recent fMRI and eye-tracking research from the 2020s reveals predictive coding in ventral occipito-temporal areas during reading, where prior linguistic context anticipates upcoming words, modulating fixation durations and VWFA responses to enhance efficiency.^[100]^[101]^[102]^[103]

References

[1]
Neural Basis of Language: An Overview of An Evolving Model - PMC
The neural basis of language involves Broca's and Wernicke's areas, the arcuate fasciculus, and two pathways: dorsal (phonological) and ventral (semantic).The Neural Basis Of Language · 1. From Sound To Phoneme · 4. Phonological Processing
[2]
Neuroanatomy, Broca Area - StatPearls - NCBI Bookshelf - NIH
Structure and Function. The primary functions of the Broca area are both language production and comprehension. While the exact role in the production is ...Introduction · Structure and Function · Embryology · Physiologic Variants
[3]
Neuroanatomy, Wernicke Area - StatPearls - NCBI Bookshelf
Jul 24, 2023 · Wernicke area was first discovered in 1874 by a German neurologist, Carl Wernicke. It has been identified as 1 of 2 areas found in the cerebral ...
[4]
Neurobiological Causal Models of Language Processing - PMC - NIH
A neurobiological causal model is a mechanistic description of language processing that is grounded in, and constrained by, the characteristics of the ...Multiple Explanatory... · First Principles Of... · Dynamical Systems View On...
[5]
Left Hemisphere Lateralization for Language in Right-Handers ... - NIH
In >90% of humans, both language and the preferred hand are hosted in the left hemisphere (LH), which has led to genetic theories linking hemispheric dominance ...
[6]
Handedness and hemispheric language dominance in healthy ...
The incidence of right-hemisphere language dominance was found to increase linearly with the degree of left-handedness.
[7]
Re-establishing Broca's Initial Findings - PMC - NIH
Broca's presentations were milestones in the history of the neuroscience of speech, language and the brain, but they were only more defined echoes of assertions ...
[8]
Redefining the role of Broca's area in speech - PMC - NIH
Broca's area is widely recognized to be important for speech production, but its specific role in the dynamics of cortical language networks is largely unknown.
[9]
Wernicke's Area Revisited: Parallel Streams and Word Processing
“Wernicke's area” thus may be better construed as two cortical modules, an auditory word-form area (AWFA) in the auditory ventral stream and an “inner speech ...
[10]
Carl Wernicke of the Wernicke Area: A Historical Review
Carl Wernicke (1848–1905), a neurologist, neurosurgeon, and psychiatrist, is renowned for his field-defining work on aphasia with his first publication in 1874.
[11]
The Angular Gyrus: Multiple Functions and Multiple Subdivisions
This review discusses the involvement of the AG in semantic processing, word reading and comprehension, number processing, default mode network, memory ...
[12]
The role of the angular gyrus in semantic cognition - PubMed Central
The angular gyrus (AG) is widely considered a key brain region for semantic cognition. However, the role of the AG in semantic processing is controversial.
[13]
Angular Gyrus - an overview | ScienceDirect Topics
The angular gyrus is involved in semantic processing of written words, auditory words, and pictures, and plays a role in complex information integration and ...Introduction · Functional Roles in Cognitive... · The Angular Gyrus in...
[14]
The role of the basal ganglia and cerebellum in language processing
The roles of the cerebellum and basal ganglia have typically been confined in the literature to motor planning and control.
[15]
Thalamic and basal ganglia involvement in language-related functions
The basal ganglia (BG) seems to be a pacemaking, propensity-regulating, and sequencing system for the linguistic elements in language production and ...
[16]
Functional roles of the thalamus for language capacities - PMC
Jul 16, 2013 · The thalamus does not engage in proper linguistic operations, but rather acts as a central monitor for language-specific cortical activities.
[17]
Functional Contributions of the Arcuate Fasciculus to Language ...
Jun 25, 2021 · Current evidence strongly suggests that the arcuate fasciculus (AF) is critical for language, from spontaneous speech and word retrieval to repetition and ...
[18]
Conduction Aphasia - StatPearls - NCBI Bookshelf
Feb 25, 2024 · ... arcuate fasciculus cause conduction aphasia. Language production and auditory comprehension remain relatively preserved in conduction aphasia ...
[19]
Diffusion Tensor Imaging Studies on Arcuate Fasciculus ... - Frontiers
Bidirectional transfer of information may indicate that information regarding language production is important for language understanding (Bernal and Ardila, ...<|control11|><|separator|>
[20]
Ventral and dorsal pathways for language - PNAS
Nov 18, 2008 · A dorsal stream is involved in mapping sound to articulation, and a ventral stream in mapping sound to meaning.
[21]
https://www.ahajournals.org/doi/10.1161/STROKEAHA.119.026222
[22]
The left inferior fronto-occipital fasciculus subserves language ...
Apr 18, 2014 · The statistical map was found to substantially overlap with the spatial position of the inferior fronto-occipital fasciculus (IFOF) (37.7%).
[23]
The white matter architecture underlying semantic processing
The 43 included studies suggest that the left inferior fronto-occipital fasciculus contributes to the essential connectivity that allows semantic processing.
[24]
Bilingualism Influences Structural Indices of Interhemispheric ...
If interhemispheric organization varies with bilingual language experience one might also expect to observe differences in corpus callosum anatomy between ...
[25]
Understanding Language Reorganization With Neuroimaging
When the language-dominant hemisphere is damaged by a focal lesion, the brain may reorganize the language network through functional and structural changes ...
[26]
Overlapping Networks Engaged during Spoken Language ...
In addition to these “task-positive” systems, a network normally considered to be “task negative,” known as the default mode network (DMN), is engaged in the ...
[27]
Functional Brain Connectivity During Narrative Processing Relates ...
Many of these studies have highlighted an important role for the default mode network (DMN), and in particular, the posterior medial cortex (PMC) (Fransson and ...<|separator|>
[28]
The Long View of Language Localization - PMC - NIH
May 24, 2019 · Paul Broca's (1824–1880) interpretation of the autopsy findings of Louis Victor Lebornge (Broca, 1861) established the clinico-pathological ...
[29]
Wernicke Aphasia - StatPearls - NCBI Bookshelf - NIH
Wernicke aphasia is characterized by impaired language comprehension. Despite this impaired comprehension, speech may have a normal rate, rhythm, and grammar.
[30]
Language and language disorders: neuroscience to clinical practice
Jul 26, 2019 · The Wernicke–Geschwind model of the 1960s additionally included a role for the angular gyrus in silent reading (with input to Wernicke's area) ...
[31]
Conduction aphasia and the arcuate fasciculus - PubMed
Wernicke, and later Geschwind, posited that the critical lesion in conduction aphasia is in the dominant hemisphere's arcuate fasciculus.
[32]
Broca and Wernicke are dead, or moving past the classic model of ...
We argue that the Classic model (1) is based on an outdated brain anatomy; (2) does not adequately represent the distributed connectivity relevant for language.Missing: critique | Show results with:critique
[33]
The arcuate fasciculus and the disconnection theme in language ...
Furthermore, the Geschwind–Wernicke model predicted that lesions at any point along the course of the arcuate fasciculus result in an identical aphasia. Yet, ...
[34]
Degeneracy in the neurological model of auditory speech repetition
Nov 13, 2023 · These and other observations led Tremblay and Dick to claim that the Wernicke-Lichtheim-Geschwin model is obsolete. More contemporary models ...
[35]
The cortical organization of speech processing: Feedback control ...
The dual stream model (Hickok & Poeppel, 2000, 2004, 2007) holds that early stages of speech processing occurs bilaterally in auditory regions on the dorsal STG ...
[36]
The cortical organization of speech processing - Nature
Apr 13, 2007 · Hickok and Poeppel describe a dual-stream model of speech processing and discuss how this model can account for some of the field's ...
[37]
Revealing the dual streams of speech processing - PNAS
A neuroanatomical model of speech processing proposed by Hickok and Poeppel (8) also emphasizes similar dual route processing. Their dual stream model includes ...
[38]
Predictive coding under the free-energy principle - PubMed Central
This paper considers prediction and perceptual categorization as an inference problem that is solved by the brain.
[39]
Evidence of a predictive coding hierarchy in the human brain ...
Mar 2, 2023 · Critically, and in line with predictive coding theory, our results reveal a hierarchical organization of language predictions in the cortex, in ...
[40]
The Role of Mirror Neurons in Speech and Language Processing
Both authors argue that mirror neurons are an evolutionary pre-curser for the development of speech. In particular, they suggest that mirror neurons evolved to ...
[41]
Language and mirror neurons - Oxford Academic
These neurons, called “mirror neurons,” represent a system that directly matches observed and executed actions. This article first provides an overview of the ...
[42]
Connectionism - Stanford Encyclopedia of Philosophy
May 18, 1997 · Connectionism is a movement in cognitive science that hopes to explain intellectual abilities using artificial neural networks.
[43]
[PDF] Connectionist Modeling of Language: Examples and Implications
Connectionist modeling is attractive as a framework for understanding cognition in general, and language in particu- lar, because it provides an account of the ...
[44]
Challenges to the Modularity Thesis Under the Bayesian Brain Models
Oct 10, 2019 · This review examines how BB models address the assumption of modularity. Empirical evidences of top-down influence on early sensory processes ...
[45]
Artificial Neural Network Language Models Predict Human Brain ...
Artificial neural networks have emerged as computationally plausible models of human language processing. A major criticism of these models is that the amount ...
[46]
Deciphering language processing in the human brain through LLM ...
Mar 21, 2025 · This study shows that neural activity in the human brain aligns linearly with the internal contextual embeddings of speech and language within LLMs as they ...
[47]
Phoneme and word recognition in the auditory ventral stream - PMC
The first two stages involve elemental and concatenative phonetic recognition ... Cortical representation of natural complex sounds: effects of acoustic features ...
[48]
Reconciling time, space and function: A new dorsal–ventral stream ...
We present a new dorsal–ventral stream framework for language comprehension which unifies basic neurobiological assumptions.
[49]
Dynamic speech representations in the human temporal lobe - PMC
We review recent studies that focus on contextual modulations of neural activity in the superior temporal gyrus (STG), a major hub for spectrotemporal encoding.
[50]
Auditory, Visual and Audiovisual Speech Processing Streams in ...
The human superior temporal sulcus (STS) is responsive to visual and auditory information, including sounds and facial cues during speech recognition.
[51]
Ventral and dorsal streams in the evolution of speech and language
This hierarchical organization of the auditory ventral stream with regard to speech-sound processing was recently corroborated with more refined techniques ( ...
[52]
Dorsal and ventral streams: a framework for understanding aspects ...
This cortical processing system then diverges into two broad processing streams, a ventral stream, which is involved in mapping sound onto meaning, and a dorsal ...
[53]
Shared and distinct representational dynamics of phonemes and ...
Jan 24, 2025 · We show that both phonemic and prosodic information is represented along the ventral stream, transferring from posterior to anterior temporal ...
[54]
The Large-Scale Organization of Gestures and Words in the Middle ...
Jul 24, 2019 · The middle temporal gyrus (MTG) has been shown to be recruited during the processing of words, but also during the observation of actions.
[55]
Connectivity of Fronto-Temporal Regions in Syntactic Structure ...
Oct 8, 2024 · The ventral stream consists of multiple fasciculi (uncinate fasciculus, extreme capsule, middle longitudinal fasciculus, inferior longitudinal ...Introduction · Fmri Connectivity Analysis · Discussion
[56]
Right hemisphere ventral stream for emotional prosody identification
Participants were presented with 24 sentences with neutral semantic content spoken with happy, sad, angry, afraid, surprised, or bored prosody.
[57]
Linguistic and emotional prosody: A systematic review and ALE ...
The ventral stream initiates bilaterally, which explains activations of linguistic prosody in both hemispheres ... Aprosodia Subsequent to Right Hemisphere Brain ...
[58]
Split-Brain: What We Know Now and Why This is Important for ...
May 12, 2020 · Variations in handedness may lead to differences in language capabilities in the right hemisphere, and could even underly differences in inter- ...
[59]
Specific disruption of the ventral anterior temporo-frontal network ...
Oct 10, 2022 · The present results underline the contribution of the ventral language stream in lexico-semantic processing and higher cognitive functions, ...
[60]
Direct Exploration of the Role of the Ventral Anterior Temporal Lobe ...
Investigations of semantic dementia have implicated the anterior temporal lobe (ATL) region, in general, as crucial for multimodal semantic memory. The ...
[61]
Diverse Frontoparietal Connectivity Supports Semantic Prediction ...
Jan 29, 2025 · Our findings suggest that both top-down semantic prediction and bottom-up semantic integration in language comprehension rely upon diverse ...
[62]
Roles of ventral versus dorsal pathways in language production
By contrast, the ventral stream has been associated with the ability to map auditory input onto conceptual and semantic representations, as well as syntactic ...
[63]
Phonological Working Memory Representations in the Left Inferior ...
Jun 23, 2022 · This inferior parietal lobe region is different from the speech perception region in the STG and has been proposed as the neural substrate of a ...
[64]
Language Learning Variability within the Dorsal and Ventral ... - NIH
Sep 27, 2017 · In terms of the underlying function facilitated by this dorsal stream, there is extensive evidence linking the phonological working memory ...
[65]
Distinct cortical locations for integration of audiovisual speech and ...
Audiovisual (AV) speech integration is often studied using the McGurk effect, where the combination of specific incongruent auditory and visual speech cues ...
[66]
Distinct cortical locations for integration of audiovisual speech ... - NIH
We aimed to isolate brain areas specifically involved in processing congruent AV speech and the McGurk effect.
[67]
The dorsal stream contribution to phonological retrieval in object ...
A non-auditory, semantically driven speech task, such as object naming, relies on the dorsal route for error-free phonological production.
[68]
Dorsal language stream anomalies in an inherited speech disorder
Feb 23, 2019 · Our findings identify disruption of the dorsal language stream as a novel neural phenotype of developmental speech disorders, distinct from that ...
[69]
The Dorsal Differentiation of Velar From Alveolar Stops in Typically ...
Mar 14, 2021 · This study provides articulatory norms derived from ultrasound tongue imaging for the dorsal differentiation in alveolar and velar stops in TD children.Abstract · Ultrasound Recording · DiscussionMissing: stream adaptability
[70]
Merge in the Human Brain: A Sub-Region Based Functional ...
These and other studies across different languages indicate that the larger region in and around Broca's area in the inferior frontal cortex (IFG) supports ...
[71]
Neural Mechanisms Underlying the Computation of Hierarchical ...
By examining various sentence structures, we recently demonstrated that activations in the left inferior frontal gyrus (L. IFG) and left supramarginal gyrus (L.
[72]
The role of input revisited: Nativist versus usage-based models
This article examines the role of input in nativist and usage-based language acquisition theories, using both first and second language examples.
[73]
Compositionality in language, cognition, and deep neural networks
May 24, 2024 · In this chapter, we survey recent empirical work from machine learning for a broad audience in philosophy, cognitive science, and neuroscience.
[74]
Compositionality in the semantic network: a model-driven ... - PubMed
Aug 1, 2025 · Overall, our work clarifies which brain regions represent semantic information compositionally across contexts and tasks, and qualifies which ...
[75]
Molecular evolution of FOXP2, a gene involved in speech and language - Nature
### Summary of FOXP2 Gene Mutations and Evolution
[76]
The evolution of modern human brain shape | Science Advances
Jan 24, 2018 · ... H. sapiens, the frontal area becomes taller, the parietal areas bulge, the side walls become parallel, and the occipital area becomes ...
[77]
[PDF] Natural language and natural selection - Steven Pinker
Abstract: Many people have argued that the evolution of the human language faculty cannot be explained by Darwinian natural selection.
[78]
Insights From Language‐Trained Apes: Brain Network Plasticity and ...
Sep 9, 2025 · Studying the brains of language‐trained apes (LTAs), such as chimpanzees, bonobos, gorillas, and orangutans, offers a rare opportunity to ...
[79]
Infants' brain responses to speech suggest Analysis by Synthesis
Jul 14, 2014 · We show that 7-mo-old infants activate auditory and motor brain areas similarly for native and nonnative sounds; by 11–12 mo, greater activation ...Infants' Brain Responses To... · Cortical Speech Motor Areas... · Experiment 1
[80]
The 7 Stages of Language Acquisition in Children
Dec 13, 2023 · The 7 stages are: prelinguistic (0-12 months), holophrastic (12-18 months), two-word (18-24 months), telegraphic (24-30 months), early multi- ...Missing: ontogenetic | Show results with:ontogenetic
[81]
Neural adaptations in short-term learning of sign language revealed ...
Feb 13, 2025 · Neuroimaging studies show that spoken and sign language rely on similar areas of the brain in the frontal and parietal regions.
[82]
https://www.frontiersin.org/journals/communication/articles/10.3389/fcomm.2021.748430/full
[83]
The visual word form area: expertise for reading in the fusiform gyrus
This brain specialization is essential to rapid reading ability because it enhances perception of words by becoming specifically tuned to recurring properties ...<|separator|>
[84]
[PDF] The visual word form area: expertise for reading in the fusiform gyrus
The Visual Word Form Area (VWFA) is a brain region, located in the left fusiform gyrus, that is essential for rapid reading by enhancing word perception.
[85]
Position coding in the visual word form area - PNAS
Numerous studies (1–5) have identified a region in a small strip of the fusiform gyrus in the left hemisphere that seems to be specialized for orthographic ...
[86]
The unique role of the visual word form area in reading - ScienceDirect
Reading systematically activates the left lateral occipitotemporal sulcus, at a site known as the visual word form area (VWFA).
[87]
A Dual-Route Perspective on Brain Activation in Response to Visual ...
DRC: A dual-route cascaded model of visual word recognition and reading aloud. ... Neural correlates of lexical and sublexical processes in reading. Brain and ...
[88]
Taking the sublexical route: brain dynamics of reading in the ...
Aug 13, 2020 · Together these results provide further evidence supporting a dual-route model for reading aloud mediated by the interplay between lexico- ...<|separator|>
[89]
The Precentral Gyrus Contributions to the Early Time-Course of ...
Feb 10, 2022 · But basic questions of how a visual stimulus is transduced into an auditory code, known as grapheme-to-phoneme conversion (GPC), remain ...
[90]
Functional MRI evidence for modulation of cerebral activity by ...
The specific process converting graphemes into phonemes is called grapheme-to-phoneme conversion (GPC) or letter-sound (print-to-sound) translation. The GPC ...
[91]
Disrupted white matter connectivity underlying developmental dyslexia
Jan 20, 2016 · For example, disruption of the superior longitudinal fasciculus/arcuate fasciculus was observed repeatedly in dyslexia, which likely ...
[92]
Alterations in white matter pathways underlying phonological and ...
The finding of a disrupted left arcuate fasciculus in Chinese children with dyslexia suggests that this dorsal pathway plays a similar role in reading ...
[93]
Neurobiology of Dyslexia - PMC - NIH
DTI studies often find reduced organization or volume in the left superior longitudinal fasciculus, including the arcuate fasciculus, and corona radiata fibers ...
[94]
Graphomotor memory in Exner's area enhances word learning in the ...
Apr 6, 2021 · We show that late-blind participants, but not sighted participants, learned novel words by recruiting the left dorsal premotor cortex known as Exner's writing ...
[95]
(PDF) The graphemic/motor frontal area Exner's area revisited
Aug 6, 2025 · Exner's area is located in the posterior aspect of the middle frontal gyrus in Brodmann area 6, and this area is shown to have a role in written ...
[96]
The graphemic/motor frontal area Exner's area revisited
Oct 8, 2009 · Visual presentation of single letters activates a premotor area involved in writing. Neuroimage 2003; 19: 1492–1500. 10.1016/S1053-8119(03) ...Abstract · Material And Methods · Discussion
[97]
How Characters Are Learned Leaves Its Mark on the Neural ... - NIH
Dec 21, 2022 · Two typical left hemispheric areas for logographic reading showed increased responses to characters in the brains of proficient alphabetic ...
[98]
The Neural System Underlying Chinese Logograph Reading
Written Chinese as logographic script differs notably from alphabets such as English in visual form, orthography, phonology, and semantics.
[99]
[PDF] Universal and specific reading mechanisms across different writing ...
Each logographic character corresponds to a syllable, but the pronunciation of the character cannot be assembled from parts of the character as in alphabetic.
[100]
The Role of Occipitotemporal Network for Speed-Reading: An fMRI ...
Jun 27, 2024 · These findings highlight the complex connectivity and dynamic interactions within the occipitotemporal network during speed-reading processes.Missing: predictive | Show results with:predictive