. Author manuscript; available in PMC: 2026 May 31.

Published in final edited form as: Nat Rev Neurosci. 2024 Apr 12;25(5):289–312. doi: 10.1038/s41583-024-00802-4

The language network as a natural kind within the broader landscape of the human brain

Evelina Fedorenko ^1,^2,^3,^†, Anna A Ivanova ⁴, Tamar I Regev ^1,²

PMCID: PMC13222024 NIHMSID: NIHMS2142018 PMID: 38609551

Abstract

Language behavior is complex, but neuroscientific evidence disentangles it into distinct components supported by dedicated brain areas or networks. In this Review, we describe the ‘core’ language network, which includes left-hemisphere frontal and temporal areas, and show that it is strongly inter-connected, input and output modality independent, causally important for language, and language-selective. We discuss evidence that this language network plausibly stores language knowledge and supports core linguistic computations related to accessing words and constructions from memory and combining them to interpret (decode) or generate (encode) linguistic messages. We emphasize that the language network works closely with, but is distinct from, both lower-level—perceptual and motor—mechanisms and higher-level systems of knowledge and reasoning. The perceptual and motor mechanisms process linguistic signals, but, in contrast to the language network, are only sensitive to these signals’ surface properties, not their meaning; the systems of knowledge and reasoning (such as the system that supports social reasoning) are sometimes engaged during language use but are not language-selective. This Review lays a foundation for both in-depth investigations of these different components of the language processing pipeline, and for probing inter-component interactions.

1. Introduction

Today, you may have reminded your daughter about her tennis practice, listened to your father complain about his neighbor, skimmed an article about local politics, or wrote a report for work. None of these behaviors would be possible without language. How does our brain support this quintessentially human ability—language processing?

Language is a system of conventionalized symbols that a person can use to communicate specific, detailed meanings to others. To support this sophisticated communication system, our brain carries out complex, multi-component operations that map between meanings and linguistic forms (words or word sequences) and between linguistic forms and the perceptual signals that instantiate those forms. During language comprehension, one needs to perceive linguistic signals (such as speech, sign, or written text) and decode the intended meanings while integrating them with preceding linguistic context and non-linguistic knowledge sources. And during language production, one needs to transform intended meanings into linguistic form and then generate corresponding physical output (speech, sign, or written text). Many brain areas must act in concert to support these complex language processing behaviors. But do all these areas operate as an indivisible whole, or can this language processing pipeline [G] be partitioned into distinct components (“nearly decomposable systems”¹)?

Different components of language (or linguistic) processing have been disentangled through an iterative process of theorizing and empirical testing (Box 1). This iterative process has yielded a detailed picture of the neural infrastructure of language, revealing that the brain’s linguistic capacity is supported by a set of language-specific representations, which capture regularities at the levels of sounds, words, and syntactic structure, along with a large set of form-meaning mappings for words and constructions [G]^2,3. These language-specific representations are used to decode linguistic messages during language comprehension and encode them during language production^2,3. This core language system is distinct from modality-specific perceptual systems that deliver information to it during comprehension^4,5; motor systems that receive information from it during production⁶; and cognitive systems of knowledge and reasoning that interact with it to make transformations between thoughts and linguistic forms and to achieve diverse goals during real-life language use^7,8,9.

Box 1∣. Beyond coarse anatomy: The importance of functional localization.

The question of what constitutes meaningful units of analysis in the brain has long been controversial^14,352. Two main alternatives include: anatomical definitions (brain area x is an area that falls in a particular location in the brain, typically described in terms of macroanatomy such as sulci and gyri), and ii) functional definitions (brain area x is an area that performs some perceptual, motor, or cognitive function: for example, the fusiform face area is an area that supports face perception³⁵³, or visual middle temporal area is an area that processes visual motion³⁵⁴). In many parts of the brain, the relationship between anatomy and function is complex, which makes it difficult to predict function from an anatomical location alone³⁵⁵. Consider the images of the language network in individual participants in Fig. 2B: even with a naked eye, it is easy to appreciate the inter-individual variability in the precise anatomical locations and sizes and/or shapes of these functional areas (see refs.^19,23,49,51 for quantitative evidence). Yet, for many years, both patient and neuroimaging studies on the neural basis of speech and language have used anatomical definitions. Terms like Broca’s area and Wernicke’s area have been extensively used, but they were defined anatomically, not functionally, and these definitions have varied substantially across researchers²²⁰, leading to massive confusion and occasional nihilism about the very idea of localization of function.

A key problem for anatomical definitions is the functional heterogeneity of the association cortex³⁵⁶. For example, language areas lie adjacent to a few distinct functional areas, including lower-level speech perception and articulation areas and areas that belong to non-language cognitive networks. As a result, any macroanatomical area (for instance, some portion of the inferior frontal gyrus or of the superior temporal gyrus) will inevitably encompass multiple functionally distinct areas. The combination of inter-individual topographic variability and functional heterogeneity creates a massive problem for analyses that average brains voxel-wise in a common space, as much past speech and language research has done. Because any given voxel in a common space often corresponds to different functional areas across individuals¹³¹, such analyses lead to blurring of nearby areas and information loss^14,15.

Functional localization provides a solution^14,19,353 (see ref.³⁵⁷ for an alternative approach). This approach relies on extensively validated paradigms (‘localizers’) that target a particular perceptual, motor, or cognitive process. The localizer contrasts are typically motivated by work in experimental psychology and patient investigations. Localizer paradigms afford greater confidence that the ‘same’ brain area/network is referred to across individuals, studies, labs, and species, and—through their consistent use—enable meaningful accumulation of knowledge. Brain areas and networks identified by well-validated localizers correspond closely to areas and networks that can be identified bottom-up from large amounts of resting state data by clustering voxel time courses^23,358. This correspondence suggests that localizers are simply an efficient way to identify the relevant area(s), and in doing so, they respect the brain’s intrinsic organization rather than imposing structure where there is none. Finally, although functional localization originated in fMRI, it has been successfully ported to intracranial human recordings, MEG, and animal physiology^{155,252,359,360,361}.

Perhaps the most common functional localizer for the language network relies on a contrast between reading or listening to sentences¹⁹ (e.g., nobody could have predicted the earthquake) versus pronounceable nonword sequences (e.g., u bizby acworrily mape las pome; Fig. 4c). However, an important feature of an effective localizer is that it is generalizable. Indeed, the sentences>nonwords localizer successfully generalizes to diverse other contrasts between a language stimulus and a perceptually matched condition, including auditorily presented sentences or passages versus muffled and/or acoustically degraded sentences or passages^48,50; and forward speech versus reversed speech⁴⁷, (including using audiovisual stimuli⁷⁰). The fact that the specific localizer paradigm for the language network does not matter is essential to showing the validity of the functional localization approach in language neuroscience, and such robust paradigms show a strong correspondence with subdivisions that can be discovered from task-free functional correlation analyses^23,358.

The separation between either linguistic versus perceptual processing or linguistic versus cognitive processing is not a given. Indeed, one can imagine an architecture where comprehension of auditory language (speech) fully takes place within the auditory cortex and processing of written language (text) fully takes place within the visual cortex. Or one can imagine an architecture where language leverages the same neural resources as those required for thought, as in proposals arguing for a universal syntactic composition engine^10,11,12 (see ref. ¹³ for an alternative proposal). Part of the difficulty in discovering the dissociations we discuss in this piece may have to do with methodological limitations of past work. Most early functional MRI (fMRI) work on language had relied on a group-averaging approach, which assumes voxel-wise alignment across brains, leading to information loss and blurring of the activation patterns (Box 1). Combined with reliance on paradigms that conflate language with speech or with general cognitive effort, the group-averaging approach contributed to the lack of clarity on the relationship between language processing and perceptual, motor, and cognitive processes in the brain. However, a gradual shift in cognitive neuroscience to individual-subject analyses^{14,15,16,17,18}, including in language research^{19,20,21,22,23,24,25} has produced a set of robust and replicable findings about the language network and its relationship with other systems, suggesting that language processing carves out a specific set of regions in the human brain.

In this Review, we discuss brain areas that are specific to language — what we will refer to as the language network — and position them in relation to perception, motor planning, and cognition (Fig. 1). Drawing on evidence from brain imaging studies, intracranial recording and stimulation work, and investigations of patients with developmental and acquired speech, language, and cognitive disorders, we show that the language network constitutes a natural kind—an ontologically meaningful grouping of brain areas based on their response properties and functional cohesion. For brain imaging, we primarily draw on fMRI data from studies that have relied on the individual-subject functional localization [G] approach^14,19; Box 1), which was essential in clarifying the distinctions discussed. We first introduce the language network and summarize its key properties. Then, we discuss the internal structure and computations of the language network. Next, we overview the perceptual and motor brain areas that subserve comprehension and production but are functionally distinct from the language network: we show how, unlike the language areas, these areas are not sensitive to the meaning of linguistic messages, only to their surface properties. We then examine several networks that support human thought (knowledge and reasoning abilities) and show that although they interact with the language network during real-life language use, they are distinct from it in that they are not language-selective. Last, we summarize the contributions of this body of work, highlight some open questions, and make suggestions for addressing these questions in future research.

Fig. 1 ∣ — Both language comprehension (top arrows) and production (bottom arrows) require the core language system (purple) to decode and encode linguistic messages, respectively, but also require lower-level perceptual and motor mechanisms (blue and red, respectively) and higher-level systems of knowledge and reasoning (green shades; note that we do not imply that no other brain areas or networks contribute to thought processes — we simply focus on three well-characterized networks that must interact with the language network during real-life use). For comprehension, the bottom-up processing of linguistic input (solid arrows) is complemented by top-down influences (dashed arrows) of both general knowledge and reasoning on linguistic interpretation^{321,322,323,324} (Box 2), and linguistic knowledge on perceptual processes^249,250,251. All brain areas schematically represent average anatomical location (see Supplementary methods for details); functional areas in individual brains are smaller than these schematic representations and vary in their precise locations.

2. The language network as a natural kind

Every typical adult human brain contains a set of areas that are specialized for language (Fig. 2). Below, we describe the core functional properties of the language areas and show that they form an integrated network.

2a. Anatomy and inter-connectivity

Despite substantial inter-individual variability in the anatomy-to-function mapping (Box 1; Fig. 2B), the language network occupies a well-defined position in the brain (Fig. 2a). This network consists of lateral frontal areas (located within the inferior and middle frontal gyri) and lateral temporal areas (located within the superior and middle temporal gyri, stretching from the temporal pole to the posterior extent of the temporal lobe). These areas are lateralized to the left hemisphere in most individuals, as evidenced by stronger and more spatially extensive responses to language in the left hemisphere and by a higher likelihood of aphasia following left-hemisphere damage^26,27,28. The reason for the left-hemisphere language dominance remains debated^29,30,31, but left-hemispheric lateralization does not appear to be critical for language function. First, in some individuals with no known neurological or genetic conditions, frontal and temporal language areas may not show a hemispheric bias or may show right-hemispheric lateralization, without any effect on linguistic abilities^32,33. And second, in the presence of early left-hemisphere damage (for example, due to early-childhood stroke), a language network can develop in the right hemisphere, again with no discernable behavioral consequences^34,35,36. That said, in many developmental disorders (such as autism, developmental language disorder and epilepsy) language responses have been reported to be more bilateral^37,38,39, and at least in some studies, more bilateral responses have been linked to worse behavioral outcomes⁴⁰.

In addition to the lateral frontal and temporal areas, the language network includes other cortical areas such as the homotopic areas in the non-language-dominant hemisphere^41,42 (Fig. 2a-b), cortical-midline areas²³ and an area on the ventral temporal surface^5,43; subcortical areas⁴⁴; and cerebellar areas^45,46. The contributions to language processing of these different components of what is sometimes referred to as the ‘extended language network’ remain an active area of research. In this Review, we focus on the ‘core’ frontal and temporal areas in the language-dominant hemisphere.

The language network can be identified in individual participants in a few minutes of fMRI scanning by contrasting brain responses during language processing versus the processing of a stimulus that is similar to language in surface form but lacks linguistic meaning and structure. Common localizer contrasts [G] (Box 1) include reading or listening to sentences versus nonword-lists, or listening to speech in one’s native language versus speech played backwards or speech in an unfamiliar language^{19,47,48,49,50}. Although variable across individuals (Box 1; Fig. 2b), the topography of the language-responsive areas and their properties (for example, the magnitude of response to language or the degree of left-hemispheric lateralization) are stable within individuals over time⁵¹ (Fig. 2c) (but see ref.⁵² for evidence that this does not hold for studies that don’t use localizers).

Furthermore, the language network is strongly inter-connected. Several white-matter tracts connect different parts of the language network^53,54, although the precise contributions of the different tracts remain debated⁵⁵. The network is also strongly functionally connected [G], as evidenced by a high degree of correlated activity among the language areas during so-called naturalistic cognition paradigms [G], like resting state or story comprehension ^23,56,57. This strong functional connectivity suggests that the different areas work together in the service of a common goal. In fact, the language network can be identified based on functional connectivity alone²³, further establishing its existence as a natural kind and not a product of a single methodological approach, such as functional localization (Box 1).

Note that although throughout the Review, we talk about functional brain areas as discrete entities with sharp borders, our arguments do not critically require this property and are compatible with gradual changes between nearby functional areas (see ‘Open questions and a way forward’ for discussion).

2b. Responsiveness to different kinds of language

As elaborated below, the language areas engage during both comprehension and production; are input-modality and output-modality independent; respond in different task contexts; and are similar across languages. Furthermore, these areas are sensitive to linguistic regularities [G] at different information scales, from sequences of phonemes to words to sentences.

First, the language network is engaged during both comprehension and production (Supplementary Fig. 1). A number of studies have revealed strong overlap in both frontal and temporal areas during comprehension versus generation of linguistic content^58,59,60. These results overturn the classic model of the neurobiology of language that separates production and comprehension^61,62,63 (but see ‘The language network and language-relevant perceptual and premotor areas’ for evidence of segregation at the level of perceptual and motor processes). Second, the language areas are modality-independent [G]: they respond to diverse kinds of input, including spoken, written and signed language^{19,64,65,66,67}; Fig. 2d), and they respond during generation of diverse kinds of output, including speaking and typing⁶⁰ (Supplementary Fig. 1). Third, the language network responds to language in different task contexts [G], whether individuals are passively comprehending (including when processing rich naturalistic stimuli, such as stories, dialogs and movies), trying to remember words when reading or listening to sentences, making judgments about words or sentences, or answering comprehension questions^24,68,69,70. Fourth, the language network is similar in its topography and properties across diverse languages both across speakers⁵⁰ and within bilingual or multilingual individuals^71,72,73. And last, the language network is sensitive to linguistic regularities at different scales, from sub-lexical structure (how phonemes, syllables, and morphemes go together), to word forms and meanings, to phrase-level combinatorial (syntactic and semantic) structure. This sensitivity is evidenced by the language network’s engagement by diverse paradigms — including those that use single, unconnected words (word reading or listening, picture naming, semantic judgments, verbal fluency or verb generation; ^60,74 and sentences (sentence reading or listening, sentence judgments or sentence-picture matching;^65,75 — its sensitivity to linguistic manipulations at these different scales^19,76; see ‘The internal structure and core computations of the language network’), and the ability of researchers to decode and encode⁷⁷ diverse linguistic features (phonemic, semantic, and syntactic) using the language network’s activity^{78,79,80,81,82}.

Together, these properties suggest that the language network stores abstract (modality-independent) linguistic knowledge and carries out core linguistic computations that are necessary for decoding and encoding linguistic messages across tasks and languages.

2c. Causal role in language ability

The left-hemisphere frontal and temporal brain areas, which house the language network, are causally important [G] for language. Damage to these areas in adulthood leads to aphasia — deficits in language comprehension and production^{26,83,84,85,86,87} The literature on aphasia is extensive and complex. This complexity stems, in part, from the fact that brain damage doesn’t typically respect functional subdivisions and often affects multiple nearby distinct areas. Moreover, the precise locations of functional areas vary across individuals and are not predictable from macroanatomy (as discussed in Box 1). Because individuals with aphasia will not have typically undergone systematic functional mapping prior to their brain injury, it is often difficult to determine which functional areas are affected by the lesion⁸⁸. Nevertheless, several brain-behavior associations have emerged robustly (some of which will be discussed in ‘The language network and language-relevant perceptual and premotor areas’).

With respect to the language network proper, circumscribed damage is typically associated with relatively quick recovery^83,89,90. Extensive damage that affects multiple network components and the underlying white-matter tracts is typically required for long-lasting linguistic deficits⁹⁰. Quick recovery from circumscribed damage suggests that the language network is characterized by some degree of redundancy, although its posterior temporal component may be the most critical and irreplaceable one, given that damage to it is associated with longer-lasting and more severe deficits^83,90.

2d. Selectivity for language versus non-linguistic inputs and tasks

The language areas are highly selective for language processing relative to diverse non-linguistic inputs and tasks⁹¹(Fig. 2d). Language has been argued to share machinery with numerous cognitive functions and domains, including the processing of hierarchically-structured inputs^10,12, executive functions^92,93,94, conceptual thought^95,96,97, and action observation^98,99. However, these claims have not found empirical support. When the language areas are identified functionally in individual participants (Box 1), they show remarkable selectivity for language (Fig. 2d; Supplementary Fig. 1). In particular, the language areas are not engaged when participants listen to music — a rich, hierarchically-structured stimulus^{91,100,101,102} — and show little or no response when participants perform demanding cognitive tasks, including solving arithmetic problems^{22,91,103,104,} and logic puzzles^103,105, understanding computer code^106,107, or carrying out working memory and cognitive control tasks^{24,50,91,108,109}. Language areas are also not recruited for conceptual or semantic tasks on pictorial inputs ^60,110; but see refs.^111,112 and show little response when individuals process socially-relevant information, including faces, bodies and hands^101,113,114; when they observe others’ actions¹¹³, including communicative signals such as eye-gaze and speech-accompanying gestures^113,115; or when they reason about others’ mental states—what is often referred to as mind-reading or having a theory of mind [G]^101,116,117. Finally, language areas only show a weak response when participants perceive^{19,50,118,119} or articulate^60,120 meaningless nonwords or syllable sequences (the existence of these weakly positive responses can be explained by the surface similarity of these stimuli to well-formed language, see ‘The internal structure and core computations of the language network’).

This extensive neuroimaging evidence that the language network is selective for language versus non-linguistic inputs and tasks is complemented by behavioral evidence from aphasia. If brain damage is primarily restricted to the language areas, individuals with aphasia lose their linguistic abilities but retain their non-linguistic abilities, including music perception¹⁰², math^121,122, general reasoning^123,124, conceptual semantics¹¹⁰, and social skills including theory of mind^{123,125,126,127,128}. This preservation of non-linguistic abilities even extends to cases where the relevant task elicits some response in the language areas in healthy adults. For example, a meaning-judgment task on pictures engages the language network to some degree (albeit less than linguistic inputs; Fig. 2d), but patients with severe aphasia are not impaired on this task¹¹¹. This pattern suggests that a weakly positive response to some non-linguistic inputs/tasks in the language areas may be epiphenomenal (i.e., not indicative of the critical role of the language areas in the relevant ability). It is also worth noting that although both linguistic and non-linguistic abilities may be affected in some cases of brain damage^129,130, this is neither surprising nor informative given the proximity between the language system and other cognitive systems^23,131,132; cases showing dissociations between linguistic and non-linguistic abilities are much more informative.

The constellation of properties of the language-responsive areas that we have delineated in this section justify the term ‘the language network’. The word ‘language’ indicates that the core function of these brain areas is language processing: these areas respond to language in all its manifestations, are causally important for language function, and are selective for language (see Box 2 for caveats). The word ‘network’ conveys that this distributed set of brain areas form an interconnected system that is distinct from other areas of the brain. Identifying stable systems within the brain with a unique functional profile is a core goal of cognitive neuroscience: to partition the brain into natural subdivisions (Box 2). Another core goal is to understand how each subdivision contributes to perception, action, and cognition. Next, we summarize progress in the latter effort — to understand how the language network supports language processing.

Box 2∣. The language network versus language: ontological kinds at different levels.

In this Review, we describe the language network — a physical component of the brain — and its relationship to language — a function of the mind. Mapping between neuroscience and cognitive science in such a way is not always straightforward³⁶²: one needs to postulate and rigorously evaluate the linking functions between the brain and the mind, which includes identifying the right units of analysis that can be linked. On the neuroscience side, much progress in identifying the right units has been made thanks to functional localization approaches (Box 1) and data-driven parcellations of individual-specific functional connectivity maps^23,301,363. On the cognitive side, the relevant concepts come from folk psychology (‘language’ or ‘reasoning’) and disciplines, like linguistics and cognitive science, that break these broad folk-psychological concepts into distinct sub-components (‘syntax’ and ‘pragmatics’ for language, or ‘induction’ and ‘deduction’ for reasoning), often on theoretical grounds. Such a priori cognitive concepts typically do not neatly map onto the brain, requiring iterative refinement of the cognitive ontology so that it aligns best with the neural subdivisions discovered in neuroscience³⁶⁴.

As we argue throughout the Review, there exists a remarkable degree of alignment between the concept of language derived from folk psychology and/or linguistics and the set of functions performed by the language network. However, this alignment is imperfect: for instance, discourse processing is an essential component of language as a cognitive function³⁶⁵, and yet the language network does not support it (see ‘The internal structure and core computations of the language network’). Thus, language in ‘the language network’ is an approximation, a convenient shorthand, for the network’s function rather than a precise description. Furthermore, other ontological distinctions from linguistics, like syntax versus semantics, do not map onto dissociable brain components at all. Thus, much caution is needed when attempting to assign cognitive or linguistic labels to brain areas: our cognitive ontologies may not align 1:1 with the neural architecture.

Some have argued that the mapping endeavor should be abandoned altogether in favor of a holistic interactionist study of the brain³⁶⁶ and/or an ‘inside out’ approach to neuroscience, with bottom-up discoveries driving the process of clustering brain computations rather than assigning labels from the top-down^367,368. Although bottom-up label-free approaches are doubtlessly valuable for making progress in neuroscience, our Review is testament that top-down approaches also provide value and that cognitive labels like ‘language’ are a rich source of useful initial hypotheses for what a brain area or network might do.

3. The internal structure and core computations of the language network

In this section, we dive into the contributions of the language network to language processing. We focus on two key properties of this network: its sensitivity to linguistic structure at multiple information scales, and functional similarity of its different areas.

3a. Sensitivity to regularities at sub-lexical, word, and phrase and sentence level, but not discourse level

Human languages are characterized by structure (regular patterns or ‘regularities’) at different levels, from how sounds and word-parts go together (phonotactics [G] and morphology [G]), to what meanings words carry (lexical semantics [G]), to how words combine to create phrases and sentences (syntax [G] and compositional semantics [G]), to how sentences go together to create coherent narratives and conversations (discourse [G]). The human brain could, in principle, process these different kinds of regularities in separate specialized areas (see refs.^{133,134,135,136,137,138} inter alia for examples where such distinctions have been advocated). On the other extreme, the brain could use the same area to process them all. The reality is closer to the second possibility.

Much evidence suggests that linguistic patterns at different levels are processed within the language network. In particular, the language network is sensitive to phonotactic ‘well-formedness’ of letter or sound strings⁷⁶ (Fig. 3a,b) and to morphological structure¹³⁹. It also supports the retrieval of individual word meanings, as indicated by a stronger response to real words than nonwords^19,118 (Fig. 3a,b) and by responses during tasks that require semantic processing of words (such as meaning relatedness judgment tasks; refs.^74,104. Furthermore, the language network supports combinatorial syntactic and semantic processes, as indicated by a stronger response to sentences than word-lists^19,140,141 (Fig. 3a,b), modulation of neural responses by the difficulty of sentence construction — including the costs associated with predicting upcoming words^142,143,144 and integrating incoming words into an evolving representation of sentence structure in memory^{75,117,145,146}— and sensitivity to many other syntactic and semantic manipulations^147,148,149.

Fig. 3 ∣ — **a-b,** Responses in the five left-hemisphere language areas (a) and in the left-hemisphere language network overall (b) to linguistic manipulations at three information scales: discourse structure¹⁶², syntactic structure¹⁵⁰, and phonological structure⁷⁶. The language areas all show a similar response profile (despite slight apparent differences, no region-by-condition interactions come out as reliable¹⁹⁰). They show a lack of sensitivity to discourse structure, as evidenced by a similarly strong response to unconnected sentences as to connected paragraphs; in other words, the costs associated with inter-sentence connections are not processed by the language areas (see ‘The networks the language network works with to support real-life language use’ for discussion of sensitivity to discourse structure in theory of mind and default node network areas). By contrast, the language areas show strong sensitivity to syntactic structure and to phonotactic well-formedness, with stronger responses to more linguistically well-formed stimuli. For syntactic manipulations, effects of structure are present in both linguistic manipulations consisting of real words and manipulations where words are replaced with nonwords (so-called ‘Jabberwocky’ materials), although the response to the latter is overall weaker, which shows that lexical information strongly modulates responses in the language network (see also refs.^19,47). On the brain template, the language areas schematically represent average anatomical location (see Supplementary methods for details); functional areas in individual brains are smaller than these schematic representations and vary in their precise locations (see Fig. 2b-c for sample individual areas). C. Correlations among the five language areas with respect to their responses to 1,000 diverse sentences¹⁸⁷ (the figure also shows correlations between each of the five language regions and the language network as a whole, in the bottom row, which is less critical to the current point). All correlations are high (between 0.47 and 0.77; average: 0.667), which suggests that even at a fine-grain level of individual linguistic stimuli (such as sentences), the language areas have similar preferences.

That so-called ‘Jabberwocky’ sentences (made up of nonwords; for example: ‘The flumpy blork was clooding in the meaves’) elicit a stronger response than nonword-lists^19,141 (Fig. 3a,b) further suggests that the language network is sensitive to abstract syntactic patterns. Importantly, however, lexical-semantic processes (related to accessing word meanings from memory) and combinatorial processes (related to combining words into phrases and sentences) are strongly integrated in every area of the language network during both comprehension^19,47,150 and production⁶⁰ (Fig. 3a,b), contrary to proposals that syntactic processing is spatially separable from lexical-semantic processing and other aspects of language^{133,138,151,152,153}. Even at a finer spatial scale, as probed with multivariate pattern analyses¹⁵⁴ or intracranial recordings^155,156, every neural population within the language network that supports combinatorial processing also processes word meanings. This evidence aligns with the idea that natural language syntax is highly lexicalized, such that the rules for how words combine depend strongly on the particular words^2,157,158.

In contrast to regularities at and below the sentence level, computations related to building discourse structure are not supported by the language areas (see Box 2 for discussion): the language network responds as strongly to lists of unconnected sentences as to connected passages^{159,160,161,162} (Fig. 3a,b); instead, other brain networks support these computations (see ‘The networks the language network works with to support real-life language use’).

3b. Functional differences among the language areas?

Past claims about dissociations within the language network.

Many claims have been made about dissociations within the language network based on functional brain imaging studies and reviews or meta-analyses thereof^{136,163,164,165,166,167}; voxel-based morphometry and voxel-lesion symptom mapping studies of patients with stroke aphasia^84,87,168 or primary progressive aphasia^169,170; and transcranial magnetic stimulation^171,172 or intracranial stimulation^{173,174,175,176} studies. These dissociation claims have sometimes been based on paradigms where the conditions of interest are not matched for difficulty and/or where critical control conditions are missing, which complicates interpretation. Some claims have further contradicted one another (for example, different researchers have argued for a different language area being the main syntactic hub; see ref.¹⁴⁷ for discussion). Most importantly for our purposes, much prior work has not functionally identified the language areas, which makes these dissociation claims challenging to evaluate with respect to the proposal outlined in this Review. In particular, it is impossible to determine whether any dissociation claimed is between two language areas or between a language area and a nearby functionally distinct area like a lower-level perceptual or motor area (see ‘The language network and language-relevant perceptual and premotor areas’) or a higher-level cognitive area (see ‘The cognitive networks the language network interacts with to support real-life language use’).

In addition to these past claims about functional dissociations between language areas, some have argued for dissociations among different aspects of language without specifying particular brain areas implicated, typically based on behavioral dissociations in individuals with brain damage (see also ref.¹⁷⁷ for a review of investigations of individual differences in linguistic abilities). For example, some have argued for the separability of noun versus verb processing^178,179 or for the separability of syntax from other aspects of language (‘agrammatism’^180,181. However, many claims about dissociations within language per se — in contrast to dissociations that could be explained by a combination of a linguistic impairment and an impairment in the lower-level perceptual or motor processing or higher-level cognitive processing — have been questioned empirically and/or conceptually (for example, see refs.^182,183,184 for evidence against the idea that agrammatism is associated with a selective loss of syntactic ability; see ref.¹⁸⁵ for a review).

Similar functional profiles for the different language network areas.

The evidence from the functional localization approach in fMRI has so far revealed highly similar functional profiles among the language areas, in contrast to clear dissociations between language areas and perceptual or motor areas (see ‘The language network and language-relevant perceptual and premotor areas’). All language areas are sensitive to sub-lexical regularities, word meanings, and combinatorial linguistic structure (Fig. 3a,b), support prediction and integration during syntactic structure building ^{75,117,142,143,144,186} and are similarly modulated by diverse linguistic manipulations^147,148,149. These similarities hold even at the grain of preferences for particular linguistic stimuli: for example, the relative magnitudes of response to specific sentences are highly similar across the different language areas, with all areas showing maximal responses to sentences that are somewhat difficult to process¹⁸⁷ (Fig. 3c).

However, one brain area that was originally included as part of the language network¹⁹ — the language-responsive area in the angular gyrus — has been found over the last decade to consistently pattern differently from the rest of the network. Although, like other language areas, this area responds more to sentences than control conditions like nonword-lists, it is less strongly functionally connected with other language areas^56,104,117, less selective for language, showing a strong response to non-linguistic meaningful stimuli^111,117,188, and not sensitive to syntactic complexity^{75,141,143,150,189}. Thus, we no longer consider this area to be a part of the language network.

Making meaningful progress in understanding the internal architecture of the language network.

More work is needed to understand the internal organization and information flow within the language network. To make meaningful progress, it would be helpful to agree on the evidential standard for functional dissociations. Minimally, for fMRI investigations, the following two criteria have to be met to argue that two language areas differ functionally: the language areas should be identified functionally in individual participants using a previously validated approach (note that this criterion rules out the utility of group-averaging studies or meta-analyses of activation peaks), and a reliable area-by-condition interaction statistic should be reported: it is not sufficient to show that language area L1 is sensitive to manipulation M and language area L2 is less or not sensitive to manipulation M, it is critical to also show that the two areas reliably differ in their sensitivity to M¹⁹⁰. Furthermore, if reporting a single dissociation [G] (as in the example in the preceding sentence), it is important to account for the overall level of neural response and the level of responsiveness to language stimuli in the relevant language areas. In our example, area L2 may have overall weaker neural signal, which would lead to lower responsiveness to M (but may not reflect a true functional difference from L1). In general, as with patient studies¹⁹¹, double dissociations [G] are more powerful: showing that area L1 is more sensitive to manipulation M1 than M2, and area L2 is more sensitive to manipulation M2 than M1. In addition to these criteria, to minimize the chances of spurious findings, and in line with the field’s increasing emphasis on robustness and replicability¹⁹², the result should also be replicated with a new group of participants and ideally generalized to a different paradigm that targets the same cognitive process(es) (for example, if areas L1 and L2 dissociate in their sensitivity to syntactic complexity in a controlled paradigm, they should show a similar dissociation during naturalistic comprehension). Similar criteria can be formulated for other approaches, including patient investigations and intracranial recording and stimulation studies.

Intracranial human recording approaches, which continue to gain popularity, perhaps combined with the use of neural network language models^{187,193,194,195,196}, may lead to substantial advances in the coming years. Indeed, intracranial recordings have already helped uncover functional heterogeneity within the language network that does not correspond to regional boundaries. Regev, Casto et al.¹⁹⁷ reported evidence for neural populations with distinct ‘temporal receptive widows’¹⁹⁸, displaying sensitivity to different-length contexts ranging from ~1 word to 2-word sequences to multi-word spans (see also refs.^199,200,201). These neural populations were distributed across the language network, which suggests that all language areas have direct access to linguistic representations at multiple scales, including single-word processing and combinatorial syntactic and semantic processing, as discussed earlier in this section. Importantly, this functional heterogeneity does not undermine the very idea that the language network is a ‘thing’: although different cells and cell populations within the network likely perform distinct computations or represent different information, they share something deep about their functionality—they jointly contribute to some aspect of language processing.

To summarize, many distinctions within the language network that have been advocated in the past have not withstood empirical and statistical scrutiny, or can be explained as dissociations between a language area and a nearby functionally distinct area. Moreover, aspects of language that are often studied by different subfields of linguistics (such as phonology, morphology, or syntax; but see refs.^3,202,203 for usage-based approaches, which do not draw sharp boundaries between structure at different levels) are all processed by the same set of brain areas comprising the language network, although some heterogeneity among interleaved neural populations has already been uncovered and more distinctions are likely to be found in future work.

4. The language network and language-relevant perceptual and premotor areas

As noted above, the language network is input-modality and output-modality independent (Fig. 2d; Supplementary Fig. S1). How can it handle such diverse inputs (like speech, text, and sign) and generate such diverse outputs? The answer lies in the plethora of brain areas that support language-relevant perceptual and motor processes. These areas are distinct from the language network as well as from general-purpose sensory and motor areas, like primary auditory or primary sensorimotor cortex. Below, we highlight the key functional difference between the language areas and specialized perceptual and premotor areas, and discuss evidence for this difference with respect to both spoken and written language.

A body of work has demonstrated the existence of brain areas that are selective for speech perception, speech articulation, and reading. A key feature that distinguishes these areas from the language network areas is their insensitivity to meaning [G]. In contrast to the language areas, which are robustly sensitive to the meaning of linguistic signals, specialized perceptual and premotor areas are not (Fig. 4). Consider a nonword like ‘blork’: although it is not associated with meaning in our lexicon, one can discern the speech sounds when one hears it, repeat it out loud, and read or write it, if they are literate. Thus, the perceptual and premotor areas can still perform their computations given that they exclusively deal with the surface properties [G] of linguistic stimuli, like the speech sounds or the visual shapes of the letters. For example, the speech perception area responds as strongly to nonword-lists or speech in an unfamiliar language as it does to sentences in a familiar language^204,205; this pattern of similar responses to meaningful versus meaningless linguistic stimuli also holds for the premotor articulation area¹²⁰ and the reading area^5,206,207. (Note that although some have reported sensitivity to lexical and semantic effects within the superior temporal gyrus (STG)^{208,209,210,211}, it is impossible to determine — without functional localization — whether these effects arise within the speech perception area or the nearby language area (Box 1, Fig. 4c).) The insensitivity of these areas to the meaning of linguistic stimuli rules out their contribution to higher-level language processes, like lexical access and syntactic structure building.

Fig. 4 ∣ — a, Primary sensory (dark blue) and motor (dark red) areas, the specialized perceptual areas (blue), which include the speech perception area and the VWFA, and motor areas (red), which include Broca’s area and the putative Exner’s area, and the language areas (purple) respond to visual and/or auditory linguistic stimuli modalities. b, The language network (identified with a language network localizer as in c-purple box) is strongly selective for language (showing little or no response to non-linguistic inputs and tasks) and sensitive to the meaning of linguistic messages (with stronger responses to sentences, which convey compositional meanings, weaker responses to word-lists, which only convey word-level meanings, and weaker still responses to meaningless nonword-lists). This pattern holds for both language comprehension (listening and reading) and production (speaking and writing). The perception areas — the speech perception area (listening) and the VWFA (reading) (identified with localizers as in c-blue boxes) — and the motor planning areas — Broca’s area (speaking) (identified with a localizer as in c-red box) and the putative Exner’s area (writing) — are also strongly selective for language relative to non-linguistic stimuli. However, in contrast to the language network, these areas are not sensitive to linguistic meaning (showing similar response profiles to sentences, word-lists, and nonword-lists), only to their surface properties. These specialized perceptual and motor areas have, in turn, distinct profiles from primary sensory and motor areas, which are not selective for linguistic inputs. (It is worth noting that the speech perception area may correspond to Wernicke’s area⁶³ as originally defined (an area in the superior temporal cortex that stores ‘sound images’ of words; cf. anatomical definitions that have dominated the field²²⁰), but this relationship deserves a more extensive discussion, which is beyond the scope of this Review.) c, Two sample localizers for the speech perception area (listening to meaningless (to the participant) speech sounds versus acoustically matched control conditions), for VWFA (viewing meaningless letter strings versus visually scrambled letter strings or object line drawings), for Broca’s area (producing meaningless syllable sequences versus non-speech oral-motor movements or hand movements; note that the latter contrast would also activate the sensorimotor cortical areas that control the mouth), and the language network (listening to or reading sentences versus perceptually similar conditions). All brain areas schematically represent average anatomical location; functional areas in individual brains are smaller than these schematic representations and vary in their precise locations (Box 1). All profiles are schematic but based on data in published studies (speech perception area: refs.^4,204,205; VWFA: refs.^5,206,207; Broca’s area: refs.^223,120; Supplementary methods).

4a. Speech

Speech perception.

An area in the STG and sulcus bilaterally responds robustly to speech^4,212,213. Although debated at some point²¹⁴, Norman-Haignere et al.²⁰⁴ established that this area is selective for speech over diverse natural sound categories (see also refs.^205,215,216). This speech selectivity cannot be reduced to lower-level acoustic features: such features explain less than half of the variance in the speech perception area’s responses to sounds²⁰⁴. As such, this area must be selective for higher-level spectro-temporal structure [G] that creates the percept of speech, and it may correspond to the area that Wernicke discovered in patients with selective auditory comprehension deficits⁶³.

Although the computations that the speech perception area performs is a matter of ongoing research, some findings provide important hints. First, this area is sensitive to speech-specific temporal structure, as evidenced by reduced responses to temporally-scrambled speech^4,204, and its temporal receptive window is approximately 500 milliseconds⁴ —a timescale that falls between syllables and words, in contrast to the longer receptive window of the language areas, as discussed above. Second, neural populations in the STG, which encompasses the speech perception area, plausibly support the processing of phonemes and syllables: some cells and neural populations show selectivity for particular phonetic features like the place and manner of articulation^213,217, reflect categorical perception of consonants²¹³ or the mapping between the formants and vowel categories²¹⁸. In tandem, the evidence suggests that the speech perception area is selective for speech sounds relative to other auditory inputs, tuned to spectro-temporal properties of speech, and supports the early stages of speech processing before word forms are linked to meanings.

Speech articulation.

Broca²¹⁹ described a patient with selective difficulties in articulation. The patient could only utter a single syllable (“tan”²¹⁹) but could make non-speech sounds and oral-motor movements and exhibited no other linguistic or cognitive deficits. A post-mortem examination revealed a lesion in the posterior portion of the patient’s left inferior frontal gyrus. The lesion site has since been known as ‘Broca’s area’, although anatomical rather than functional definitions of this area have dominated the literature ²²⁰ (Box 1), leading to much confusion given that the left inferior frontal cortex consists of diverse functional areas¹³¹. Broca’s discovery has stood the test of time, with many subsequent investigations confirming the existence of an area in the inferior frontal cortex that supports articulation^{6,120,221,222,223,224}.

However, the field did go through a period when the role of Broca’s area in articulation was questioned. Dronkers²²⁵ performed a lesion overlap analysis for a set of patients who had suffered a stroke in the left middle cerebral artery (MCA) and exhibited articulatory deficits (‘apraxia of speech’ [G]). The only shared area of damage in these patients was the superior precentral gyrus of the insula (SPGI), leading to a claim that the SPGI, not Broca’s area, is responsible for articulation. However, as Hillis et al.²²¹ pointed out, Dronkers did not examine the probability that a lesion in SPGI causes articulatory deficits, which is important given the generally high vulnerability of the anterior insula to damage caused by MCA stroke. Hillis and colleagues examined patients with and without damage in the left insula and did not find an association between insular (including SPGI) damage and articulatory deficits; instead, their analysis revealed a consistent association between damage to the posterior left inferior frontal gyrus (Broca’s area) and articulatory deficits (see also ref.²²⁶). Moreover, Fedorenko et al.²²⁷ did not find support for the claim that SPGI is selective for articulation: using fMRI, they showed that SPGI responds strongly to non-speech oral-motor movements and shows little to no response to articulation, in contrast to Broca’s area, which responded strongly to articulation, and showed sensitivity to articulatory complexity.

Of course, Broca’s area is not the only brain region that enables speech articulation. Brain imaging studies have found that Broca’s area, functionally defined (Box 1, Fig. 4c, works closely with several other brain areas as part of the ‘articulation network’^6,223. One such area is the ventral sensorimotor cortex (vSMC). (Although in Fig. 4, we talk about primary motor areas, somatosensory cortex has long been shown to work closely with the adjacent motor areas for the relevant effectors during movement production^228,229,230.) As measured with fMRI, Broca’s area and vSMC both respond during the production of speech sounds and show sensitivity to articulatory complexity^223,231, but their contributions to articulation differ. First, they are active during different stages of speech production: Broca’s area is active during the preparation of speech motor plans whereas vSMC is active during their execution²²². In line with the somatotopic organization of vSMC²³², neural populations therein show selectivity for particular articulators (lips, tongue, vocal cords), and contain information about articulator states (e.g., whether the lips are pursed or the position of the tongue)²³³. Second, Long et al.²²⁴ showed that interfering with the activity of Broca’s area leads to slower speech (apraxia-like symptoms^219,234), whereas interfering with vSMC leads to slurring (dysarthria-like symptoms [G]²³⁵). Last, Broca’s area, but not vSMC, shows some selectivity for speech production relative to the production of non-speech oral-motor movements^120,223, which explains the existence of selective articulatory deficits in patients with damage to this brain region^219,221,226. In tandem, the evidence suggests that Broca’s area is a critical, most speech-articulation-selective node of the articulation network: it prepares speech motor plans and sends them to vSMC, which implements articulatory movements (but also supports non-speech oral-motor movements; Fig. 4c).

4b. Reading and writing

Approximately five thousand years ago, human societies began to develop writing systems. Currently, over half of the ~7,000 of the world’s languages have written forms²³⁶. How do our brains process written language —a culturally recent invention?

Reading.

Dejeurine²³⁷ described a patient with a selective deficit in recognizing letters and written words. This patient could recognize other visual objects, perceive and copy letter shapes, and exhibited no other linguistic or cognitive deficits (even their writing was intact). A post-mortem autopsy revealed a lesion on the ventral surface of the left temporal lobe. Over a century later, brain imaging investigations identified an area in the left ventral visual cortex that appears to support reading ability²³⁸. This area is known as the visual word-form area (VWFA). When properly defined (for example, by a contrast between words or nonwords and line drawings of objects; cf. coarser contrasts like words>checkerboards; Fig. 4c), this area responds similarly strongly to strings of letters and real words^206,207. Thus, like the speech perception area, which processes the auditory form of words (or nonwords), the VWFA processes their visual form.

Like the speech perception area, the VWFA is strongly selective: it responds more to letter strings in a familiar script than to other visual stimuli^{5,206,207,239}. This area serves as a prime example of experience-driven specialization: it develops as individuals learn to read²⁴⁰ and it only responds to scripts that individuals have familiarity with (for example, it shows a low response to Hebrew letters in individuals unfamiliar with the Hebrew alphabet²⁰⁶). Furthermore, in line with evidence from patients with reading deficits, interfering with the VWFA’s activity leads to temporary inability to identify letters or read words²⁴¹. In tandem, the evidence suggests that the VWFA is a critical, visual-word-form-selective area: it is tuned to visual properties of familiar scripts and supports the early stages of reading, before word forms are linked to meanings.

Writing.

The neural basis of motor planning and execution during written language production has received relatively limited attention (but see ref.²⁴²). As expected, execution of hand-motor movements recruits the finger and hand areas of the sensorimotor cortex^243,244. However, whether Broca’s area plays a role during written (or signed) language production, in addition to speech articulation, is not known. Broca’s area may be output-modality-independent and have a role in creating motor plans regardless of whether they are eventually implemented by the lips, tongue and vocal cords, or by the fingers and hands. Some evidence seems to support this possibility^245,246. It is also possible that Broca’s area is selective for planning speech motor movements, and a distinct area supports planning motor movements during writing (and maybe signing). Such an area was proposed by Exner^247,248, but the evidence remains scarce²⁴³. Individual-subject-level investigations may help further illuminate the motor planning and execution stages of written (and signed) language production.

In summary, the language network (Fig. 2-3) is distinct from both perceptual and motor mechanisms (Fig. 4) during language processing (Fig. 1). During comprehension (Fig. 1), linguistic information flows from general-purpose sensory systems (most commonly, primary auditory and visual cortex areas) to specialized perceptual systems (such as the speech perception area or the VWFA) before reaching the language areas. Of course, linguistic context (processed in the language network) affects these perceptual processes^249,250,251. During production (Fig. 1), the language network formulates linguistic messages (by converting conceptual representations into word sequences), and then—for spoken production—passes them to Broca’s area, which prepares motor-articulatory plans and sends them to primary motor areas for execution. The passing of information from the language areas to Broca’s area may proceed via the speech perception area (which may be needed to convert words into sound sequences) or a subset thereof that specifically implements the auditory-to-sensorimotor transformation^137,252. The systems that support motor planning and execution for written and signed language production deserve more attention.

5. The cognitive networks the language network interacts with to support real-life language use

The language network is neither the starting point during production nor the endpoint during comprehension (Fig. 1). During production, it receives input from the systems that support our ability to think and encodes them into a word sequence. During comprehension, the language network decodes linguistic input and passes it to other cognitive systems, which can use this new information to update existing knowledge structures and to reason about and act in the world.

A tight integration between language and the rest of the mind is critical for efficient language use in real-world situations. Language is a tool that allows humans to acquire knowledge, share new ideas, build relationships, make and follow requests (see ref.²⁵³ for a recent discussion). These behaviors necessarily require fast and continuous interaction between language processing and other cognitive capacities, such as executive functions, social cognition and general world knowledge. Moreover, empirically, some cognitive processes that may seem like essential parts of language — such as building discourse-level structure by connecting information across sentences — draw on cortical areas that lie outside the language network and support computations that are not specific to language (Fig. 5; Box 2).

Fig. 5 ∣ — The language network responds to diverse linguistic content but is strongly selective for language over diverse non-linguistic input and tasks, including math^91,103 and meaningful pictorial or video inputs^111,117. This selectivity of the language system (purple) is reflected by strong responses to random unconnected sentences (A) and semantically diverse sentences (B1, B2, and B3) but not mathematical expressions (C1) or meaningful pictures or video (C2, C3). By contrast, systems of knowledge and reasoning respond to content irrespective of whether this content is delivered verbally or in another format so are not selective for language. For example, the multiple demand (MD) network (salad green) responds to math statements presented as sentences (B1) or as mathematical expressions (C1)²², but not to sentences that do not have the right kind of content and that are not accompanied by a task (A)⁶⁸. The theory of mind (ToM) network (teal) responds to stimuli that invoke thoughts about mental states (knowledge, beliefs, desires) presented verbally (B2) or as pictures or videos (C2)^69,281,283, but not to sentences that lack mental-state content (A). The default mode network (DMN) (dark green) responds to connected narratives, verbal (B3) and pictorial (C3)^308,309, but not to unconnected sentences (A). All brain areas schematically represent average anatomical location; functional areas in individual brains are smaller than these schematic representations and vary in their precise locations (Box 1). All profiles are schematic but based on data in published studies (Supplementary methods).

Below, we discuss three neural systems that work with the language network to support real-life language use (see ‘Open questions and a way forward’ for a discussion of linguistic versus conceptual meaning). A key feature that distinguishes these systems from the core language areas is their lack of selectivity for language (Fig. 5). In contrast to the language network (Fig. 2d), they respond to cognitive demands or particular semantic content regardless of whether the information is delivered through language or other representational means.

5a. Multiple demand network: task demands beyond comprehension and production

Network properties.

The multiple demand (MD) network comprises bilateral frontal and parietal, medial prefrontal, and posterolateral inferior temporal domain-general areas that are active during diverse cognitively challenging tasks, with stronger responses during more difficult conditions or tasks^{7,254,255,256,257,258,259} (Fig. 5). The hard>easy response signature holds across tasks that differ in the representation format (spatial versus verbal working memory tasks) and the nature of the task (math problems, logic puzzles, novel task learning)^{254,256,258,259}. Activity in the MD network has been linked to constructs like working memory, cognitive control and attention — all critical ingredients of goal-directed behaviors. Individual differences in the activity in this network have been linked to differences in fluid intelligence²⁶⁰ and damage to this network is associated with a reduction in executive abilities and fluid intelligence^{261,262,263,264}.

Dissociation from the language network.

The MD network and the language network are robustly dissociated, as evidenced by neuroimaging studies, both task-based and naturalistic^{23,56,57,108,132,256,265} and studies of individuals with brain damage^266,267. This dissociation holds even when examining responses to linguistic difficulty: in particular, linguistic difficulty manipulations tax the language network and elicit little or no response in the MD network^{24,68,117,143,268} (see ref.¹³ for review).

Contributions to language.

First, the MD network is recruited in the presence of task demands beyond language comprehension or self-generated language production, such as answering a question, deciding whether a word was present in a sentence, or naming a picture on demand^60,68. Second, it is required for processing certain types of content, such as mathematical or logic statements^{22,103,105,106,269}. As noted in its properties above, in stark contrast to the language network, the MD areas process these types of content regardless of the format in which it arrives (for example, the expression ‘601 – 3 + 2 =’ versus the sentence ‘Subtract three from six hundred and one, then add two’; Fig. 5). Last, the MD network is engaged in some cases of effortful language comprehension, such as when processing acoustically degraded, temporally compressed or accented speech^65,270,271; when listening to speech in a foreign language that one has limited proficiency in⁷³; or when reading at fast presentation rates⁶⁵. Some have also argued that parts of the MD network can support language processing following damage to the language network^{272,273,274,275,276}. However, this claim remains controversial: a few studies and meta-analyses have failed to find evidence of the MD network’s engagement during language processing in individuals with aphasia^{277,278,279,280} (see also ref.²⁷⁷ for plausible alternative explanations of the putative evidence for MD network’s engagement).

5b. Theory of mind network: pragmatics and social reasoning

Network properties.

The theory of mind (ToM) network comprises, most prominently, brain areas in the bilateral temporo-parietal junction and along the cortical midline that are engaged when one is thinking about one’s own or others’ mental states^{8,281,282,283}. This engagement generalizes across the mental state’s content, format (linguistic versus pictorial or video), and evidence source^{8,281,284,285,286,283} (see ref.²⁸⁷ for review). By adulthood, these areas, especially the most selective component of the ToM network, which resides in the right-hemispheric temporo-parietal junction, become exquisitely selective for theory of mind relative to processing diverse other kinds of socially-relevant information: they do not respond to social stimuli (such as faces, voices, or biological motion¹⁰¹), to descriptions of another person’s physical attributes, bodily sensations, or broadly social attributes^282,283,288, or to general executive demands²⁸⁶.

Dissociation from the language network.

The ToM network is anatomically and functionally distinct from the language network. These networks show different response patterns in diverse experimental paradigms^101,116,117 and they track different information when processing rich naturalistic stimuli⁶⁹. They also dissociate in their patterns of inter-area correlations during naturalistic cognition^23,57,132. Last, even severe damage to language-processing mechanisms can leave social reasoning unimpaired^{124,126,127,289,290}.

Contributions to language.

First, the ToM network is engaged during some aspects of nonliteral language comprehension, including phenomena such as sarcasm, indirect speech and conversational implicature, where understanding the meaning of an utterance requires inferring the beliefs, desires and intentions of the speaker^{291,292,293,294} (see ref.²⁹⁵ for review; see also ref.²⁹⁶ for meta-analysis). Second, this network — along with the default mode network, as discussed below — supports the processing of discourse-level structure: relationships between clauses in narratives^162,297. Third, it is more strongly engaged during the processing of conversations compared to monologues⁷⁰, presumably because of the greater demands on representing different perspectives. Last, the ToM network supports the processing of semantic content related to mental states in both texts and movies^{69,282,283,297}.

5c. The Default Mode network: narratives and situation modeling

Network properties.

The default mode network (DMN) owes its name to the fact that it is more active at rest (the ‘default’ state) than during externally-oriented demanding tasks²⁹⁸. It comprises, most prominently, bilateral areas in the medial prefrontal and medial parietal cortex, around the temporo-parietal junction, and in the temporal pole. Although the DMN’s topography resembles that of the ToM network and some have argued that the two networks are overlapping or the same^299,300, recent evidence from individual-subject approaches (Box 1) shows that these two networks are functionally distinct although their component regions are located in close proximity^{23,114, 132,301,302,}. The exact role of the DMN in cognition is debated, with different studies linking it to episodic projection^9,132,302, self-directed processing^303,304, internal scene construction³⁰⁵ or spatial information processing¹¹⁴. The key property of the DMN that concerns us here is its sensitivity to long-range temporal contexts. In contrast to the language network’s receptive window of a few words, the DMN’s receptive window is on the scale of minutes, which allows it to connect information across multiple sentences/events^{159,161,198,306}. The DMN’s activity reflects high-level differences in people’s information processing (for instance, whether the person watching a movie knows the ending³⁰⁷), and it encodes specific event schemas, such as going to a restaurant or boarding a plane³⁰⁸. Thus, the DMN appears to track abstract, input-invariant, global situational context.

Dissociation from the language network.

Like the DM and ToM networks, the DMN is not language-selective: it tracks long-range information for both linguistic and non-linguistic inputs^308,309. Furthermore, the DMN encodes information in an input-invariant way, such that patterns of response to a particular event schema (such as going to a restaurant) are as similar between a story and a movie as between two stories³⁰⁸. Thus, the informational content of the DMN is abstract. Last, the DMN clearly dissociates from the language network in its response profile^132,265, functional correlation patterns^23,132 and effects of brain damage or stimulation on behavior^310,311.

Contributions to language.

The DMN’s long temporal receptive window allows it to track not only sentence-level, but also discourse-level linguistic structure. Indeed, this network is recruited for coherent texts more than for disconnected sentences^297,312,313 and can integrate information over multiple sentences^{159,161,193,199,201}. Thus, the task of constructing a coherent overall representation of a narrative or a dialogue requires not only the language network, but also the non-language-specific DMN areas. (If the hypothesis about the bias of the DMN toward spatial information is correct¹¹⁴, then this network may also be important for interpreting linguistic meanings that deal with spatial layouts.)

Thus, real-life language use involves joint recruitment of the language network and brain networks that are required for other cognitive functions. We discussed three higher-order networks whose role in language processing is best understood (Fig. 5). Other systems, including the emotion-processing centers, visual object recognition areas, and areas that support social perception almost certainly come into play as well in different language-use scenarios. Understanding how the different networks interact and share information remains a challenge for future research.

6. Open questions and a way forward

How does neuroscience contribute to our understanding of language and its relationship with the rest of the human mind? Perhaps most importantly, it enables us to build an empirically validated cognitive ontology^314,315 of basic processes that contribute to language comprehension and production (Box 2). Some of the distinctions that we discussed here have been previously postulated on theoretical grounds; some have further been supported by evidence from behavioral experiments and/or studies of patients with brain damage. However, fMRI as a tool is unparalleled for uncovering the structure of the mind: its non-invasive nature, relatively high spatial precision and whole-brain coverage, when combined with careful experimentation, can reveal which cognitive processes share machinery and which are supported by distinct resources³¹⁶. The functional localization approach (Box 1) has proven a powerful way to accumulate knowledge and has helped paint a rich and detailed picture of the ventral visual stream³¹⁶, and higher-level cognitive systems, like the theory of mind network²⁸⁷ and now language processing (Fig. 1).

What is next in cognitive neuroscience of language? We outline some directions below.

6a. The problem of meaning

A core distinction between the language network and adjacent perceptual and premotor areas is that the language network is sensitive to meaningful linguistic content. Does this sensitivity imply that the language network is sensitive to any meaningful (semantic) content, whether it’s verbal or non-verbal? We argue this is not the case: meaningful nonverbal inputs (such as pictures or videos) activate the language areas weakly or not at all^{60,106,110,111,112,188} (Fig. 2d, Supplementary Fig. 1), and individuals with damaged language areas retain the ability to understand the world around them^{110,111,124,125}. Thus, the language network responds to meaningful language, not to meaning in general.

Furthermore, the kinds of meaning that the language network responds to can be superficial and not make sense with respect to one’s world knowledge. For example, the language areas respond strongly to well-formed but incoherent sentences, such as “Colorless green ideas sleep furiously”³¹⁷. At the same time, naturalistic language understanding engages a broad set of brain areas beyond the language network^20,159. Thus, we speculate that language understanding includes two distinct meaning-making components: the language network extracts the lexical forms (words) from linguistic inputs, accesses their meanings, and combines them using syntactic and compositional semantic rules; and this resulting representation is integrated with information from other brain regions to enable deep, context-sensitive, and actionable language understanding.

How does the language network pass information to downstream brain systems? What is the format of this information and what computations are used to integrate linguistic information with other inputs and the brain’s internal states? These questions remain wide open. Some have argued that semantic information is broadly distributed across the brain²⁰; others that certain regions act like amodal semantic ‘hubs’³¹⁸; and yet others that linguistic meaning is fundamentally grounded in sensorimotor representations³¹⁹. In this Review, we do not commit to a particular theory of meaning, although we recognize that the account of language as a system of computations that transform form to meaning should ultimately specify what the endpoint of those transformations looks like.

Finally, our account of language comprehension as information transfer from perception areas to the language network to higher-order cognitive networks (Fig. 1) will eventually need to account for the vast body of psycholinguistic work that shows that non-linguistic sources of knowledge can affect language processing early on^{320,321,322,323,324,325}. The blurred lines between linguistic and non-linguistic modulators of language processing have sometimes led to skepticism regarding the separability of language from the rest of cognition^326,327 (Box 2). However, we believe that this tension can eventually be resolved through a better understanding of the top-down influences of higher-order cognitive networks on the language network and of the ways in which some of these effects may be explained by language-network-internal processes without requiring sophisticated cognitive processing (after all, distributional linguistic information, which is presumably stored within the language network, contains a wealth of knowledge about the world^328,329,330).

6b. Sharp edges versus functional gradients

We have discussed a few functional distinctions, including those between the language areas and perceptual and motor areas (Fig. 4), and between the language areas and areas that belong to other cognitive networks (Fig. 5). Although we have schematically drawn these various areas as ‘blobs’ with sharp edges, none of our arguments hinge on this property of the functional regions. Future studies should use high-spatial resolution tools to test the nature of the boundaries between nearby areas. For example, one could test whether there is a gradual functional change between the speech perception area and the nearby temporal language area, with some cells that fall between the speech and the language area showing mixed selectivity. Importantly, the existence of such gradients would not detract from the fact that the peaks in this functional landscape are robustly functionally different, which makes them interesting as objects of investigation (see ref.³¹⁶ for discussion).

6c. A mechanistic understanding of language processing

One set of questions concerns the separable individual components of language comprehension and production (Fig. 1). A rich characterization of the functional profiles of these different components, including their selectivity for particular kinds of inputs or tasks, critically constrains the computations they are likely to support, but more work is needed to actually decipher the nature of those computations. Doing so will likely require synergistic use of diverse research tools, including the emergent ones, like intracranial human recordings, which continue to increase in their sophistication^217,331, and artificial neural network models, which accurately capture neural responses in diverse perceptual, motor and cognitive domains^332,333, including speech and language^{193,194,195,196,334,187} (see ref.³³⁵ for a review).

Another set of questions concerns the interaction of different components with each other during complex behaviors. For example, how do representations get transformed and/or compressed as they are passed from the speech perception area to the language network and then to higher-order cognitive systems? Or how do non-linguistic information sources affect language processing, as has long been shown to be the case^{320,321,322,323,324,325}? Such questions require high-density temporally-resolved recordings in multiple brain areas simultaneously — methods that are not currently available, even in animal neuroscience. Building multi-componential artificial neural networks — for example, networks that combine a language component and a component for world knowledge or some aspect of reasoning³³⁶ — may help get some traction on these questions.

6d. The role of language in development, evolution and cognition

Differentiating the language network from other components of language processing is likely to accelerate progress in our understanding of language development. Developmental research has already provided ample evidence for dissociation between language, speech perception and articulation, and higher-level cognition. For instance, the ability to discriminate phonemes and languages develops long before language comprehension^337,338, babbling precedes word production^339,340, and higher-order cognitive abilities (like executive control and theory of mind) develop later than linguistic abilities and take time to mature^341,342. Tracing the emergence of the earliest speech perception and articulation brain mechanisms, understanding their role in language acquisition and in the development of the language network as well as the integration of the language network with other brain areas over developmental time, remain critical areas of future work. Disentangling different components supporting language behavior is also critical for studies of language evolution. In cross-species comparative research, distinct homologies must be established for the perceptual, motor and higher-level (meaning-sensitive) components of animal and human communication systems.

6e. Language as a tool to augment thought

Last, differentiating the language network from the systems of knowledge and reasoning is likely to help understand the role of language in human cognition. As reviewed, several higher-level cognitive systems (implemented in the MD, ToM, and Default mode networks) that support different aspects of thought are important for many aspects of language use but are critically not language-selective. As a result, some individuals with even severe aphasia can nevertheless retain the ability to reason, plan and understand the world³⁴³. That said, under certain conditions, language might play a supportive role for certain cognitive functions. For example, the availability of linguistic labels can help people perform working memory tasks³⁴⁴, numerical tasks³⁴⁵, and concept-learning tasks^346,347,348, although a mechanistic understanding of how the language network helps during these tasks is currently lacking. For instance, Benn, Ivanova et al.¹¹⁰ showed that an object categorization task that was previously argued to rely on language resources³⁴⁹ did not actually recruit the language network. Exploring whether, when and why the language network might be involved in non-linguistic behaviors provides a critical test for theories that place language at the core of complex thought^{95,96,97,350,351}.

In conclusion, the language network constitutes a natural kind, distinct from both language-relevant perceptual and premotor areas and from cognitive systems that support real-life language use. These distinctions provide a critical foundation for both in-depth investigations of each individual component of language processing (“divide and conquer”¹⁴) and for probing inter-component interactions, paving the path forward for the 21^st-century neuroscience of language.

Supplementary Material

supplementary

NIHMS2142018-supplement-supplementary.pdf^{(562.1KB, pdf)}

References:

1.Simon HA The architecture of complexity. Proc. Am. Philos. Soc 106, 467–482 (1962). [Google Scholar]
2.Goldberg AE Constructions: A Construction Grammar Approach to Argument Structure. (University of Chicago Press, 1995). [Google Scholar]
3.Jackendoff R Foundations of Language: Brain, Meaning, Grammar, Evolution. (Oxford Univ. Press, 2009). [DOI] [PubMed] [Google Scholar]
4.Overath T, McDermott JH, Zarate JM & Poeppel D The cortical analysis of speech-specific temporal structure revealed by responses to sound quilts. Nat. Neurosci 18, 903–911 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Li J, Hiersche K & Saygin ZM Demystifying the Visual Word Form Area: precision fMRI of visual, linguistic, and attentional properties of ventral temporal cortex. Preprint at bioRxiv 10.1101/2023.06.15.544824 (2023). [DOI] [Google Scholar]
6.Guenther FH Neural Control of Speech. (MIT Press, 2016). [Google Scholar]
7.Duncan J The multiple-demand (MD) system of the primate brain: mental programs for intelligent behaviour. Trends Cogn. Sci 14, 172–179 (2010). [DOI] [PubMed] [Google Scholar]
8.Saxe R & Kanwisher N People thinking about thinking people: the role of the temporo-parietal junction in ‘theory of mind’. Neuroimage 19, 1835–1842 (2003). [DOI] [PubMed] [Google Scholar]
9.Buckner RL & DiNicola LM The brain’s default network: updated anatomy, physiology and evolving insights. Nat. Rev. Neurosci 20, 593–608 (2019). [DOI] [PubMed] [Google Scholar]
10.Tettamanti M & Weniger D Broca’s area: a supramodal hierarchical processor? Cortex 42, 491–494 (2006). [DOI] [PubMed] [Google Scholar]
11.Fadiga L Broca’s area in language, action, and music. Ann. NY Acad. Sci 1169, 448–458 (2009). [DOI] [PubMed] [Google Scholar]
12.Fitch WT & Martins MD Hierarchical processing in music, language, and action: Lashley revisited. Ann. NY Acad. Sci 1316, 87–104 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Fedorenko E & Shain C Similarity of computations across domains does not imply shared implementation: the case of language comprehension. Curr. Dir. Psychol. Sci 30, 526–534 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Saxe R, Brett M & Kanwisher N Divide and conquer: a defense of functional localizers. Neuroimage 30, 1088–1096 (2006). [DOI] [PubMed] [Google Scholar]
15.Nieto-Castañon A & Fedorenko E Subject-specific functional localizers increase sensitivity and functional resolution of multi-subject analyses. Neuroimage 63, 1646–1669 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Poldrack RA et al. Long-term neural and physiological phenotyping of a single human. Nat. Commun 6, 8885 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Gratton C & Braga RM Editorial overview: deep imaging of the individual brain: past, practice, and promise. Curr. Opin. Behav. Sci 40, (2021). [Google Scholar]
18.Fedorenko E The early origins and the growing popularity of the individual-subject analytic approach in human neuroscience. Curr. Opin. Behav. Sci 40, 105–112 (2021). [Google Scholar]
19.Fedorenko E, Hsieh P-J, Nieto-Castañon A, Whitfield-Gabrieli S & Kanwisher N A new method for fMRI investigations of language: defining ROIs functionally in individual subjects. J. Neurophysiol 104, 1177–94 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Huth AG, Heer WA, Griffiths TL, Theunissen FE & Gallant JL Natural speech reveals the semantic maps that tile human cerebral cortex. Nature 532, 453–458 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Lane C, Kanjlia S, Omaki A & Bedny M "Visual” cortex of congenitally blind adults responds to syntactic movement. J. Neurosci 35, 12859–12868 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Amalric M & Dehaene S A distinct cortical network for mathematical knowledge in the human brain. Neuroimage 189, 19–31 (2019). [DOI] [PubMed] [Google Scholar]
23.Braga RM, DiNicola LM, Becker HC & Buckner RL Situating the left-lateralized language network in the broader organization of multiple specialized large-scale distributed networks. J. Neurophysiol 124, 1415–1448 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Quillen IA, Yen M & Wilson SM Distinct neural correlates of linguistic and non-linguistic demand. Neurobiol. Lang 2, 202–225 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Hiersche KJ, Schettini E, Li J & Saygin ZM Functional dissociation of the language network and other cognition in early childhood. Preprint at bioRxiv 10.1101/2022.08.11.503597 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Goodglass H Agrammatism. in Studies in Neurolinguistics (eds. Whitaker H & Whitaker HA) vol. 1 237–260, (Academic Press, New York, NY, 1976). [Google Scholar]
27.Knecht S et al. Language lateralization in healthy right-handers. Brain 123, 74–81 (2000). [DOI] [PubMed] [Google Scholar]
28.Baynes K & Long DL Three conundrums of language lateralization. Lang. Linguist. Compass 1, 48–70 (2007). [Google Scholar]
29.Corballis MC The Gestural Origins of Language. in The Origins of Language (ed. Masataka N) 11–23 (Springer; Japan, Tokyo, 2008). [Google Scholar]
30.Ocklenburg S, Ströckens F & Güntürkün O Lateralisation of conspecific vocalisation in non-human vertebrates. Laterality: Asymmetries of Body, Brain and Cognition 18, 1–31 (2013). [DOI] [PubMed] [Google Scholar]
31.Sha Z et al. The genetic architecture of structural left–right asymmetry of the human brain. Nat. Hum. Behav 5, 1226–1239 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Knecht S, Drager B, Floel A, Lohmann H & Breitenstein C Behavioral relevance of atypical language lateralization in healthy subjects. Brain 124, 1657–1665 (2001). [DOI] [PubMed] [Google Scholar]
33.Van der Haegen L & Brysbaert M The relationship between behavioral language laterality, face laterality, and language performance in left-handers. PLOS One 13, 0208696 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Asaridou SS, Demir-Lira ÖE, Goldin-Meadow S, Levine SC & Small SL Language development and brain reorganization in a child born without the left hemisphere. Cortex 127, 290–312 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Tuckute G et al. Frontal language areas do not emerge in the absence of temporal language areas: a case study of an individual born without a left temporal lobe. Neuropsychologia 169, 108184 (2022). [DOI] [PubMed] [Google Scholar]
36.Newport EL et al. Language and developmental plasticity after perinatal stroke. Proc. Natl Acad. Sci 119, 2207293119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Herbert MR, Harris GJ, Adrien KT, Ziegler DA & Makris N Abnormal asymmetry in language association cortex in autism. Ann. Neurol 52, 588–596 (2002). [DOI] [PubMed] [Google Scholar]
38.Yuan W et al. fMRI shows atypical language lateralization in pediatric epilepsy patients. Epilepsia 47, 593–600 (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.De Guibert C et al. Abnormal functional lateralization and activity of language brain areas in typical specific language impairment (developmental dysphasia). Brain 134, 3044–3058 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Jouravlev O et al. Reduced language lateralization in autism and the broader autism phenotype as assessed with robust individual-subjects analyses. Autism Research 13, 1746–1761 (2020). [DOI] [PubMed] [Google Scholar]
41.Lindell AK In your right mind: right hemisphere contributions to language processing and production. Neuropsychol. Rev 16, 131–148 (2006). [DOI] [PubMed] [Google Scholar]
42.Martin KC et al. A weak shadow of early life language processing persists in the right hemisphere of the mature brain. Neurobiol. Lang 3, 364–385 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Burnstine TH et al. Characterization of the basal temporal language area in patients with left temporal lobe epilepsy. Neurology 40, 966–966 (1990). [DOI] [PubMed] [Google Scholar]
44.Copland DA & Angwin AJ Subcortical Contributions to Language. in The Oxford Handbook of Neurolinguistics (eds. Zubicaray GI & Schiller NO) 851–876 (Oxford University Press, 2019). [Google Scholar]
45.De Smet HJ, Baillieux H, Deyn PP, Mariën P & Paquier P The cerebellum and language: the story so far. Folia Phoniatrica et Logopaedica 59, 165–170 (2007). [DOI] [PubMed] [Google Scholar]
46.LeBel A & D’Mello AM A seat at the (language) table: incorporating the cerebellum into frameworks for language processing. Curr. Opin. Behav. Sci 53, 101310 (2023). [Google Scholar]
47.Bedny M, Pascual-Leone A, Dodell-Feder D, Fedorenko E & Saxe R Language processing in the occipital cortex of congenitally blind adults. Proc. Natl Acad. Sci 108, 4429–4434 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Scott TL, Gallée J & Fedorenko E A new fun and robust version of an fMRI localizer for the frontotemporal language system. Cogn. Neurosci 8, 167–176 (2017). [DOI] [PubMed] [Google Scholar]
49.Lipkin B et al. Probabilistic atlas for the language network based on precision fMRI data from> 800 individuals. Sci. Data 9, 529 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Malik-Moraleda S et al. An investigation across 45 languages and 12 language families reveals a universal language network. Nat. Neurosci 25, 1014–1019 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Mahowald K & Fedorenko E Reliable individual-level neural markers of high-level language processing: a necessary precursor for relating neural variability to behavioral and genetic variability. Neuroimage 139, 74–93 (2016). [DOI] [PubMed] [Google Scholar]
52.Elliott ML et al. What is the test-retest reliability of common task-functional MRI measures? New empirical evidence and a meta-analysis. Psychol. Sci 31, 792–806 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Saur D et al. Ventral and dorsal pathways for language. Proc. Natl Acad. Sci 105, 18035–18040 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Friederici AD Pathways to language: fiber tracts in the human brain. Trends Cogn. Sci 13, 175–181 (2009). [DOI] [PubMed] [Google Scholar]
55.Dick AS & Tremblay P Beyond the arcuate fasciculus: consensus and controversy in the connectional anatomy of language. Brain 135, 3529–3550 (2012). [DOI] [PubMed] [Google Scholar]
56.Blank I, Kanwisher N & Fedorenko E A functional dissociation between language and multiple-demand systems revealed in patterns of BOLD signal fluctuations. J. Neurophysiol 112, 1105–1118 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Paunov AM, Blank IA & Fedorenko E Functionally distinct language and Theory of Mind networks are synchronized at rest and during language comprehension. J. Neurophysiol 121, 1244–1265 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Menenti L, Gierhan SM, Segaert K & Hagoort P Shared language: overlap and segregation of the neuronal infrastructure for speaking and listening revealed by functional MRI. Psychol. Sci 22, 1173–1182 (2011). [DOI] [PubMed] [Google Scholar]
59.Silbert LJ, Honey CJ, Simony E, Poeppel D & Hasson U Coupled neural systems underlie the production and comprehension of naturalistic narrative speech. Proc. Natl Acad. Sci 111, 4687–4696 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Hu J et al. Precision fMRI reveals that the language-selective network supports both phrase-structure building and lexical access during language production. Cerebral Cortex 33, 4384–4404 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Lichtheim L On aphasia. Brain 7, 433–484 (1885). [Google Scholar]
62.Geschwind N The organization of language and the brain: language disorders after brain damage help in elucidating the neural basis of verbal behavior. Science 170, 940–944 (1970). [DOI] [PubMed] [Google Scholar]
63.Wernicke C The symptom complex of aphasia. in Proceedings of the Boston Colloquium for the Philosophy of Science 1966/1968 (eds. Cohen RS & Wartofsky MW) 34–97 (Springer; Netherlands, Dordrecht, 1969). doi: 10.1007/978-94-010-3378-7_2. [DOI] [Google Scholar]
64.Neville HJ et al. Cerebral organization for language in deaf and hearing subjects: biological constraints and effects of experience. Proc. Natl Acad. Sci 95, 922–929 (1998). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Vagharchakian L, Dehaene-Lambertz G, Pallier C & Dehaene S A temporal bottleneck in the language comprehension network. J. Neurosci 32, 9089–9102 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Regev M, Honey CJ, Simony E & Hasson U Selective and invariant neural responses to spoken and written narratives. J. Neurosci 33, 15978–15988 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Deniz F, Nunez-Elizalde AO, Huth AG & Gallant JL The representation of semantic information across human cerebral cortex during listening versus reading is invariant to stimulus modality. J. Neurosci 39, 7722–7736 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Diachek E, Blank I, Siegelman M, Affourtit J & Fedorenko E The domain-general multiple demand (MD) network does not support core aspects of language comprehension: a large-scale fMRI investigation. J. Neurosci 40, 4536–4550 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Paunov AM et al. Differential tracking of linguistic vs. mental state content in naturalistic stimuli by language and Theory of Mind (ToM) brain networks. Neurobiol. Lang 3, 413–440 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Olson HA, Chen EM, Lydic KO & Saxe RR Left-hemisphere cortical language regions respond equally to observed dialogue and monologue. Neurobiol. Lang 4, 575–610 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Chee MWL et al. Processing of visually presented sentences in Mandarin and English studied with fMRI. Neuron 23, 127–137 (1999). [DOI] [PubMed] [Google Scholar]
72.Hernandez AE, Dapretto M, Mazziotta J & Bookheimer S Language switching and language representation in Spanish–English bilinguals: an fMRI Study. Neuroimage 14, 510–520 (2001). [DOI] [PubMed] [Google Scholar]
73.Malik-Moraleda S et al. Functional characterization of the language network of polyglots and hyperpolyglots with precision fMRI. Preprint at bioRxiv 10.1101/2023.01.19.524657 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
74.Binder JR et al. Human brain language areas identified by functional magnetic resonance imaging. J. Neurosci 17, 353–362 (1997). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Blank I, Balewski Z, Mahowald K & Fedorenko E Syntactic processing is distributed across the language system. Neuroimage 127, 307–323 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
76.Regev TI et al. High-level language brain regions are sensitive to sub-lexical regularities. Preprint at bioRxiv https://doi.org/2021.06.11.447786 (2021). [Google Scholar]
77.Naselaris T, Kay KN, Nishimoto S & Gallant JL Encoding and decoding in fMRI. Neuroimage 56, 400–410 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
78.Pereira F et al. Toward a universal decoder of linguistic meaning from brain activation. Nat. Commun 9, 963 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
79.Caucheteux C, Gramfort A & King J-R Disentangling syntax and semantics in the brain with deep networks. Preprint at arXiv 10.48550/arXiv.2103.01620 (2021). [DOI] [Google Scholar]
80.Toneva M, Mitchell TM & Wehbe L Combining computational controls with natural text reveals aspects of meaning composition. Nat. Comput. Sci 2, 745–757 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
81.Gong XL et al. Phonemic segmentation of narrative speech in human cerebral cortex. Nat. Commun 14, 4309 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
82.Pasquiou A, Lakretz Y, Thirion B & Pallier C Information-restricted neural language models reveal different brain regions’ sensitivity to semantics, syntax, and context. Neurobiol. Lang 4, 611–636 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
83.Luria AR Traumatic Aphasia: Its Syndromes, Psychology and Treatment. (Mouton, 1970). [Google Scholar]
84.Bates E et al. Voxel-based lesion–symptom mapping. Nat. Neurosci 6, 448–450 (2003). [DOI] [PubMed] [Google Scholar]
85.Gorno-Tempini ML et al. Classification of primary progressive aphasia and its variants. Neurology 76, 1006–1014 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
86.Wilson SM et al. Transient aphasias after left hemisphere resective surgery. J. Neurosurg 123, 581–593 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
87.Fridriksson J et al. Anatomy of aphasia revisited. Brain 141, 848–862 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
88.Blank IA, Kiran S & Fedorenko E Can neuroimaging help aphasia researchers? Addressing generalizability, variability, and interpretability. Cogn. Neuropsychol 34, 377–393 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
89.Kertesz A & McCabe P Recovery patterns and prognosis in aphasia. Brain 100 Pt 1, 1–18 (1977). [DOI] [PubMed] [Google Scholar]
90.Wilson SM et al. Recovery from aphasia in the first year after stroke. Brain 146, 1021–1039 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
91.Fedorenko E, Behr MK & Kanwisher N Functional specificity for high-level linguistic processing in the human brain. Proc. Natl Acad. Sci 108, 16428–16433 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
92.Just MA & Carpenter PA A capacity theory of comprehension: individual differences in working memory. Psychol. Rev 99, 122 (1992). [DOI] [PubMed] [Google Scholar]
93.Thompson-Schill SL, D’Esposito M, Aguirre GK & Farah MJ Role of left inferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation. Proc. Natl Acad. Sci 94, 14792–14797 (1997). [DOI] [PMC free article] [PubMed] [Google Scholar]
94.Novick JM, Trueswell JC & Thompson-Schill SL Cognitive control and parsing: reexamining the role of Broca’s area in sentence comprehension. Cogn. Affect. Behav. Neurosci 5, 263–281 (2005). [DOI] [PubMed] [Google Scholar]
95.Dennett DC The role of language in intelligence. in Sprache Und Denken / Language and Thought 42–55 (1997). doi: 10.1515/9783110806595.42. [DOI] [Google Scholar]
96.Gentner D Why we’re so smart. in Language in Mind: Advances in the Study of Language and Thought (eds. Gentner D & Goldin-Meadow S) 195–235 (MIT Press, 2003). [Google Scholar]
97.Berwick RC & Chomsky N Why Only Us: Language and Evolution. (MIT press, 2016). [Google Scholar]
98.Rizzolatti G & Arbib MA Language within our grasp. Trends. Neurosci 21, 188–194 (1998). [DOI] [PubMed] [Google Scholar]
99.Arbib MA Mirror system activity for action and language is embedded in the integration of dorsal and ventral pathways. Brain Lang. 112, 12–24 (2010). [DOI] [PubMed] [Google Scholar]
100.Fedorenko E, McDermott JH, Norman-Haignere S & Kanwisher N Sensitivity to musical structure in the human brain. J. Neurophysiol 108, 3289–3300 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
101.Deen B, Koldewyn K, Kanwisher N & Saxe R Functional organization of social perception and cognition in the superior temporal sulcus. Cereb. Cortex 25, 4596–4609 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
102.Chen X et al. The human language system, including its inferior frontal component in ‘Broca’s area,’ does not support music perception. Cereb. Cortex 33, 7904–7929 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
103.Monti MM, Parsons LM & Osherson DN Thought beyond language: Neural dissociation of algebra and natural language. Psychol. Sci 23, 914–922 (2012). [DOI] [PubMed] [Google Scholar]
104.Chai LR, Mattar MG, Blank IA, Fedorenko E & Bassett DS Functional network dynamics of the language system. Cereb. Cortex 26, 4148–4159 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
105.Monti MM, Parsons LM & Osherson DN The boundaries of language and thought in deductive inference. Proc. Natl Acad. Sci 106, 12554–12559 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
106.Ivanova AA et al. Comprehension of computer code relies primarily on domain-general executive brain regions. Elife 9, 58906 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
107.Liu YF, Kim J, Wilson C & Bedny M Computer code comprehension shares neural resources with formal logical inference in the fronto-parietal network. Elife 9, 59340 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
108.Fedorenko E, Duncan J & Kanwisher N Language-selective and domain-general regions lie side by side within Broca’s area. Curr. Biol 22, 2059–2062 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
109.Wolna A et al. Domain-general and language-specific contributions to speech production in a second language: an fMRI study using functional localizers. Sci. Rep 14, 57 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
110.Benn Y et al. No evidence for a special role of language in feature-based categorization. Cereb. Cortex 33, 10380–10400 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
111.Ivanova AA et al. The language network is recruited but not required for nonverbal event semantics. Neurobiol. Lang 2, 176–201 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
112.Sueoka Y et al. The language network reliably ‘tracks’ naturalistic meaningful non-verbal stimuli. Neurobiol. Lang 1–64 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
113.Pritchett BL, Hoeflin C, Koldewyn K, Dechter E & Fedorenko E High-level language processing regions are not engaged in action observation or imitation. J. Neurophysiol 120, 2555–2570 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
114.Deen B & Freiwald WA Parallel systems for social and spatial reasoning within the cortical apex. Preprint at bioRxiv 10.1101/2021.09.23.461550 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
115.Jouravlev O et al. Speech-accompanying gestures are not processed by the language-processing mechanisms. Neuropsychologia 132, 107132 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
116.Koster-Hale J & Saxe R Theory of mind brain regions are sensitive to the content, not the structural complexity, of belief attributions. Proc. of the Annual Meeting of the Cognitive Science Society 33, (2011). [Google Scholar]
117.Shain C, Paunov A, Chen X, Lipkin B & Fedorenko E No evidence of theory of mind reasoning in the human language network. Cereb. Cortex 33, 6299–6319 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
118.Petersen SE, Fox PT, Snyder AZ & Raichle ME Activation of extrastriate and frontal cortical areas by visual words and word-like stimuli. Science 249, 1041–1044 (1990). [DOI] [PubMed] [Google Scholar]
119.Noppeney U & Price CJ Retrieval of abstract semantics. Neuroimage 22, 164–170 (2004). [DOI] [PubMed] [Google Scholar]
120.Wolna A et al. Broca is alive and well: an articulation-selective area in the left inferior frontal gyrus, distinct from nearby language and Multiple Demand areas. Society for the Neurobiology of Language (Marseille, France, 2023). [Google Scholar]
121.Varley RA, Klessinger NJ, Romanowski CA & Siegal M Agrammatic but numerate. Proc. Natl Acad. Sci 102, 3519–3524 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
122.Benn Y et al. Differentiating core and co-opted mechanisms in calculation: the neuroimaging of calculation in aphasia. Brain Cogn. 82, 254–264 (2013). [DOI] [PubMed] [Google Scholar]
123.Varley R Aphasic language, aphasic thought: an investigation of propositional thinking in an a-propositional aphasic. in Language and Thought: Interdisciplinary Themes (eds. Boucher J & Carruthers P) 128–145 (Cambridge University Press, Cambridge, 1998). doi:DOI: 10.1017/CBO9780511597909.009. [DOI] [Google Scholar]
124.Varley R & Siegal M Evidence for cognition without grammar from causal reasoning and ‘theory of mind’ in an agrammatic aphasic patient. Curr. Biol 10, 723–726 (2000). [DOI] [PubMed] [Google Scholar]
125.Lecours A & Joanette Y Linguistic and other psychological aspects of paroxysmal aphasia. Brain Lang. 10, 1–23 (1980). [DOI] [PubMed] [Google Scholar]
126.Varley R, Siegal M & Want SC Severe impairment in grammar does not preclude theory of mind. Neurocase 7, 489–493 (2001). [DOI] [PubMed] [Google Scholar]
127.Apperly IA, Samson D, Carroll N, Hussain S & Humphreys G Intact first- and second-order false belief reasoning in a patient with severely impaired grammar. Soc. Neurosci 1, 334–348 (2006). [DOI] [PubMed] [Google Scholar]
128.Siegal M & Varley R Aphasia, language, and theory of mind. Soc. Neurosci 1, 167–174 (2006). [DOI] [PubMed] [Google Scholar]
129.Baldo J et al. Is problem solving dependent on language? Brain Lang. 92, 240–250 (2005). [DOI] [PubMed] [Google Scholar]
130.Baldo JV & Dronkers NF Neural correlates of arithmetic and language comprehension: a common substrate? Neuropsychologia 45, 229–235 (2007). [DOI] [PubMed] [Google Scholar]
131.Fedorenko E & Blank I Broca’s area is not a natural kind. Trends Cogn. Sci 24, 270–284 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
132.DiNicola LM, Sun W & Buckner RL Side-by-side regions in dorsolateral prefrontal cortex estimated within the individual respond differentially to domain-specific and domain-flexible processes. J. Neurophysiol 130, 1602–1615 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
133.Friederici AD The cortical language circuit: from auditory perception to sentence comprehension. Trends Cogn. Sci 16, 262–268 (2012). [DOI] [PubMed] [Google Scholar]
134.Price CJ A review and synthesis of the first 20 years of PET and fMRI studies of heard speech, spoken language and reading. Neuroimage 62, 816–847 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
135.Hagoort P The neurobiology of language beyond single-word processing. Science 366, 55–58 (2019). [DOI] [PubMed] [Google Scholar]
136.Pylkkänen L The neural basis of combinatory syntax and semantics. Science 366, 62–66 (2019). [DOI] [PubMed] [Google Scholar]
137.Hickok G & Poeppel D The cortical organization of speech processing. Nat. Rev. Neurosci 8, 393–402 (2007). [DOI] [PubMed] [Google Scholar]
138.Bornkessel-Schlesewsky I, Schlesewsky M, Small SL & Rauschecker JP Neurobiological roots of language in primate audition: common computational properties. Trends Cogn. Sci 19, 142–150 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
139.Bozic M, Tyler LK, Ives DT, Randall B & Marslen-Wilson WD Bihemispheric foundations for human speech comprehension. Proc. Natl Acad Sci 107, 17439–44 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
140.Snijders TM et al. Retrieval and unification of syntactic structure in sentence comprehension: an fMRI study using word-category ambiguity. Cereb. Cortex 19, 1493–1503 (2009). [DOI] [PubMed] [Google Scholar]
141.Pallier C, Devauchelle AD & Dehaene S Cortical representation of the constituent structure of sentences. Proc. Natl Acad. Sci 108, 2522–2527 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
142.Lopopolo A, Frank SL, Bosch A & Willems RM Using stochastic language models (SLM) to map lexical, syntactic, and phonological information processing in the brain. PLOS One 12, 0177794 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
143.Shain C, Blank I, Van Schijndel M, Schuler W & Fedorenko E fMRI reveals language-specific predictive coding during naturalistic sentence comprehension. Neuropsychologia 138, 107307 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
144.Heilbron M, Armeni K, Schoffelen JM, Hagoort P & Lange FP A hierarchy of linguistic predictions during natural language comprehension. Proc. Natl Acad. Sci 119, 2201968119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
145.Ben-Shachar M, Hendler T, Kahn I, Ben-Bashat D & Grodzinsky Y The neural reality of syntactic transformations: evidence from functional magnetic resonance imaging. Psychol. Sci 14, 433–440 (2003). [DOI] [PubMed] [Google Scholar]
146.Constable RT et al. Sentence complexity and input modality effects in sentence comprehension: an fMRI study. Neuroimage 22, 11–21 (2004). [DOI] [PubMed] [Google Scholar]
147.Fedorenko E, Blank IA, Siegelman M & Mineroff Z Lack of selectivity for syntax relative to word meanings throughout the language network. Cognition 203, 104348 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
148.Rodd JM, Longe OA, Randall B & Tyler LK The functional organisation of the fronto-temporal language system: evidence from syntactic and semantic ambiguity. Neuropsychologia 48, 1324–1335 (2010). [DOI] [PubMed] [Google Scholar]
149.Rodd JM, Vitello S, Woollams AM & Adank P Localising semantic and syntactic processing in spoken and written language comprehension: an activation likelihood estimation meta-analysis. Brain Lang. 141, 89–102 (2015). [DOI] [PubMed] [Google Scholar]
150.Shain C et al. Graded sensitivity to structure and meaning throughout the human language network. Preprint at bioRxiv 10.1101/2021.11.12.467812 (2023). [DOI] [Google Scholar]
151.Grodzinsky Y & Santi A The battle for Broca’s region. Trends Cogn. Sci 12, 474–480 (2008). [DOI] [PubMed] [Google Scholar]
152.Baggio G & Hagoort P The balance between memory and unification in semantics: a dynamic account of the N400. Lang. Cogn. Process 26, 1338–1367 (2011). [Google Scholar]
153.Tyler LK et al. Left inferior frontal cortex and syntax: function, structure and behaviour in patients with left hemisphere damage. Brain 134, 415–431 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
154.Fedorenko E, Nieto-Castañon A & Kanwisher N Lexical and syntactic representations in the brain: an fMRI investigation with multi-voxel pattern analyses. Neuropsychologia 50, 499–513 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
155.Fedorenko E et al. Neural correlate of the construction of sentence meaning. Proc. Natl Acad. Sci 113, 6256–6262 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
156.Woolnough O et al. Spatiotemporally distributed frontotemporal networks for sentence reading. Proc. Natl Acad. Sci 120, 2300252120 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
157.Sag IA, Wasow T, Bender EM & Sag IA Syntactic Theory: A Formal Introduction. vol. 92 (Center for the Study of Language and Information, Stanford, CA, 1999). [Google Scholar]
158.Schabes Y & Joshi AK Parsing with lexicalized tree adjoining grammar. in Current Issues in Parsing Technology 25–47 (Springer; US, 1991). doi: 10.1007/978-1-4615-3986-5_3. [DOI] [Google Scholar]
159.Lerner Y, Honey CJ, Silbert LJ & Hasson U Topographic mapping of a hierarchy of temporal receptive windows using a narrated story. J. Neurosci 31, 2906–2915 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
160.Yeshurun Y, Nguyen M & Hasson U Amplification of local changes along the timescale processing hierarchy. Proc. Natl Acad. Sci 114, 9475–9480 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
161.Blank IA & Fedorenko E No evidence for differences among language regions in their temporal receptive windows. Neuroimage 219, 116925 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
162.Jacoby N & Fedorenko E Discourse-level comprehension engages medial frontal Theory of Mind brain regions even for expository texts. Lang. Cogn. Neurosci 35, 780–796 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
163.Friederici AD The brain basis of language processing: from structure to function. Physiol. Rev 91, 1357–1392 (2011). [DOI] [PubMed] [Google Scholar]
164.Dapretto M & Bookheimer SY Form and content: dissociating syntax and semantics in sentence comprehension. Neuron 24, 427–432 (1999). [DOI] [PubMed] [Google Scholar]
165.Vigneau M et al. Meta-analyzing left hemisphere language areas: phonology, semantics, and sentence processing. Neuroimage 30, 1414–1432 (2006). [DOI] [PubMed] [Google Scholar]
166.Price CJ The anatomy of language: a review of 100 fMRI studies published in 2009. Ann. NY Acad. Sci 1191, 62–88 (2010). [DOI] [PubMed] [Google Scholar]
167.Hagoort P & Indefrey P The neurobiology of language beyond single words. Annu. Rev. Neurosci 37, 347–362 (2014). [DOI] [PubMed] [Google Scholar]
168.Dronkers NF, Wilkins DP, Van Valin RD, Redfern BB & Jaeger JJ Lesion analysis of the brain areas involved in language comprehension. Cognition 92, 145–177 (2004). [DOI] [PubMed] [Google Scholar]
169.Mesulam M-M et al. Primary progressive aphasia and the evolving neurology of the language network. Nat. Rev. Neurol 10, 554–569 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
170.Gorno-Tempini ML et al. Cognition and anatomy in three variants of primary progressive aphasia. Ann. Neurol 55, 335–346 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]
171.Gough PM, Nobre AC & Devlin JT Dissociating linguistic processes in the left inferior frontal cortex with transcranial magnetic stimulation. J. Neurosci 25, 8010–8016 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
172.Hartwigsen G et al. Dissociating parieto-frontal networks for phonological and semantic word decisions: a condition-and-perturb TMS study. Cereb. Cortex 26, 2590–2601 (2016). [DOI] [PubMed] [Google Scholar]
173.Havas V et al. Electrical stimulation mapping of nouns and verbs in Broca’s area. Brain Lang. 145–146, 53–63 (2015). [DOI] [PubMed] [Google Scholar]
174.Sierpowska J et al. Morphological derivation overflow as a result of disruption of the left frontal aslant white matter tract. Brain Lang. 142, 54–64 (2015). [DOI] [PubMed] [Google Scholar]
175.Chang EF, Kurteff G & Wilson SM Selective interference with syntactic encoding during sentence production by direct electrocortical stimulation of the inferior frontal gyrus. J. Cogn. Neurosci 30, 411–420 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
176.Dragoy O et al. Functional linguistic specificity of the left frontal aslant tract for spontaneous speech fluency: evidence from intraoperative language mapping. Brain Lang. 208, 104836 (2020). [DOI] [PubMed] [Google Scholar]
177.Kidd E, Donnelly S & Christiansen MH Individual differences in language acquisition and processing. Trends Cogn. Sci 22, 154–169 (2018). [DOI] [PubMed] [Google Scholar]
178.Caramazza A & Hillis AE Lexical organization of nouns and verbs in the brain. Nature 349, 788–90 (1991). [DOI] [PubMed] [Google Scholar]
179.Shapiro K, Shelton J & Caramazza A Grammatical class in lexical production and morphological processing: evidence from a case of fluent aphasia. Cogn. Neuropsychol 17, 665–682 (2000). [DOI] [PubMed] [Google Scholar]
180.Berndt RS & Caramazza A A redefinition of the syndrome of Broca’s aphasia: implications for a neuropsychological model of language. Appl. Psycholinguist 1, 225–278 (1980). [Google Scholar]
181.Friedmann N & Grodzinsky Y Tense and agreement in agrammatic production: pruning the syntactic tree. Brain Lang. 56, 397–425 (1997). [DOI] [PubMed] [Google Scholar]
182.Linebarger MC, Schwartz MF & Saffran EM Sensitivity to grammatical structure in so-called agrammatic aphasics. Cognition 13, 361–392 (1983). [DOI] [PubMed] [Google Scholar]
183.Bates JC & Goodman E On the inseparability of grammar and the lexicon: evidence from acquisition, aphasia and real-time processing. Lang. Cogn. Process 12, 507–584 (1997). [Google Scholar]
184.Caplan D, Hildebrandt N & Makris N Location of lesions in stroke patients with deficits in syntactic processing in sentence comprehension. Brain 119, 933–949 (1996). [DOI] [PubMed] [Google Scholar]
185.Fedorenko E, Ryskin R & Gibson E Agrammatic output in non-fluent, including Broca’s, aphasia as a rational behavior. Aphasiology 37, 1981–2000 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
186.Brodbeck C et al. Parallel processing in speech perception with local and global representations of linguistic context. Elife 11, 72056 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
187.Tuckute G et al. Driving and suppressing the human language network using large language models. Nat. Hum. Behav (2024) doi: 10.1038/s41562-023-01783-7. [DOI] [PubMed] [Google Scholar]
188.Amit E, Hoeflin C, Hamzah N & Fedorenko E An asymmetrical relationship between verbal and visual thinking: Converging evidence from behavior and fMRI. Neuroimage 152, 619–627 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
189.Shain C, Blank IA, Fedorenko E, Gibson E & Schuler W Robust effects of working memory demand during naturalistic language comprehension in language-selective cortex. J. Neurosci 42, 7412–7430 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
190.Nieuwenhuis S, Forstmann BU & Wagenmakers E-J Erroneous analyses of interactions in neuroscience: a problem of significance. Nat. Neurosci 14, 1105–1107 (2011). [DOI] [PubMed] [Google Scholar]
191.Caramazza A & Coltheart M Cognitive Neuropsychology twenty years on. Cogn. Neuropsychol 23, 3–12 (2006). [DOI] [PubMed] [Google Scholar]
192.Poldrack RA et al. Scanning the horizon: towards transparent and reproducible neuroimaging research. Nat. Rev. Neurosci 18, 115–126 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
193.Jain S & Huth AG Incorporating context into language encoding models for fMRI. in Advances in Neural Information Processing Systems (eds. Bengio S et al. ) vol. 31 (2018). [Google Scholar]
194.Schrimpf M et al. The neural architecture of langauge: integrative modeling converges on predictive processing. Proc. Natl Acad. Sci 118, 2105646118 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
195.Caucheteux C, Gramfort A & King J-R Deep language algorithms predict semantic comprehension from brain activity. Sci. Rep 12, 16327 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
196.Goldstein A et al. Shared computational principles for language processing in humans and deep language models. Nat. Neurosci 25, 369–380 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
197.Regev TI et al. Intracranial recordings reveal three distinct neural response patterns in the language network. Preprint at bioRxiv 10.1101/2022.12.30.522216 (2022). [DOI] [Google Scholar]
198.Hasson U, Yang E, Vallines I, Heeger DJ & Rubin N A hierarchy of temporal receptive windows in human cortex. J. Neurosci 28, 2539–2550 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
199.Jain S et al. Interpretable multi-timescale models for predicting fMRI responses to continuous natural speech. Adv. Neural Inf. Process. Syst 33, 13738–13749 (2020). [Google Scholar]
200.Desbordes T et al. Dimensionality and ramping: signatures of sentence integration in the dynamics of brains and deep language models. J. Neurosci (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
201.Caucheteux C, Gramfort A & King JR Evidence of a predictive coding hierarchy in the human brain listening to speech. Nat. Hum. Behav 7, 430–441 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
202.Langacker RW Foundations of Cognitive Grammar. (Stanford University Press, 1987). [Google Scholar]
203.Bybee J Language, Usage and Cognition. (Cambridge University Press, 2010). doi: 10.1017/CBO9780511750526. [DOI] [Google Scholar]
204.Norman-Haignere S, Kanwisher NG & McDermott JH Distinct cortical pathways for music and speech revealed by hypothesis-free voxel decomposition. Neuron 88, 1281–1296 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
205.Boebinger D, Norman-Haignere SV, McDermott JH & Kanwisher N Music-selective neural populations arise without musical training. J. Neurophysiol 125, 2237–2263 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
206.Baker CI et al. Visual word processing and experiential origins of functional selectivity in human extrastriate cortex. Proc. Natl Acad. Sci 104, 9087–9092 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
207.Hamamé CM et al. Dejerine’s reading area revisited with intracranial EEG: selective responses to letter strings. Neurology 80, 602–603 (2013). [DOI] [PubMed] [Google Scholar]
208.Flinker A, Chang EF, Barbaro NM, Berger MS & Knight RT Sub-centimeter language organization in the human temporal lobe. Brain Lang. 117, 103–109 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
209.Cibelli ES, Leonard MK, Johnson K & Chang EF The influence of lexical statistics on temporal lobe cortical dynamics during spoken word listening. Brain Lang. 147, 66–75 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
210.Leonard MK, Baud MO, Sjerps MJ & Chang EF Perceptual restoration of masked speech in human cortex. Nat. Commun 7, 13619 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
211.Keshishian M et al. Joint, distributed and hierarchically organized encoding of linguistic features in the human auditory cortex. Nat. Hum. Behav 7, 740–753 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
212.Scott SK, Blank CC, Rosen S & Wise RJ Identification of a pathway for intelligible speech in the left temporal lobe. Brain 123, 2400–2406 (2000). [DOI] [PMC free article] [PubMed] [Google Scholar]
213.Mesgarani N, Cheung C, Johnson K & Chang EF Phonetic feature encoding in human superior temporal gyrus. Science 343, 1006–1010 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
214.Price C, Thierry G & Griffiths T Speech-specific auditory processing: where is it? Trends Cogn. Sci 9, 271–276 (2005). [DOI] [PubMed] [Google Scholar]
215.Leaver AM & Rauschecker JP Cortical representation of natural complex sounds: effects of acoustic features and auditory object category. J. Neurosci 30, 7604–7612 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
216.Norman-Haignere SV et al. A neural population selective for song in human auditory cortex. Curr. Biol 32, 1470–1484.e12 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
217.Leonard MK et al. Large-scale single-neuron speech sound encoding across the depth of human cortex. Nature (2023) doi: 10.1038/s41586-023-06839-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
218.Oganian Y, Bhaya-Grossman I, Johnson K & Chang EF Vowel and formant representation in the human auditory speech cortex. Neuron 111, 2105–2118.e4 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
219.Broca P Remarques sur le siège de la faculté du langage articulé, suivies d’une observation d’aphémie (perte de la parole) [French]. Bulletins de la Société d’anatomie 6, 330–357 (1861). [Google Scholar]
220.Tremblay P & Dick AS Broca and Wernicke are dead, or moving past the classic model of language neurobiology. Brain Lang. 162, 60–71 (2016). [DOI] [PubMed] [Google Scholar]
221.Hillis AE et al. Re-examining the brain regions crucial for orchestrating speech articulation. Brain 127, 1479–1487 (2004). [DOI] [PubMed] [Google Scholar]
222.Flinker A et al. Redefining the role of Broca’s area in speech. Proc. Natl Acad. Sci 112, 2871–2875 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
223.Basilakos A, Smith KG, Fillmore P, Fridriksson J & Fedorenko E Functional characterization of the human speech articulation network. Cereb. Cortex 28, 1816–1830 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
224.Long MA et al. Functional segregation of cortical regions underlying speech timing and articulation. Neuron 89, 1187–1193 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
225.Dronkers NF A new brain region for coordinating speech articulation. Nature 384, 159–161 (1996). [DOI] [PubMed] [Google Scholar]
226.Richardson JD, Fillmore P, Rorden C, Lapointe LL & Fridriksson J Re-establishing Broca’s initial findings. Brain Lang. 123, 125–130 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
227.Fedorenko E, Fillmore P, Smith K, Bonilha L & Fridriksson J The superior precentral gyrus of the insula does not appear to be functionally specialized for articulation. J. Neurophysiol 113, 2376–82 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
228.Wolpert DM, Goodbody SJ & Husain M Maintaining internal representations: the role of the human superior parietal lobe. Nat. Neurosci 1, 529–533 (1998). [DOI] [PubMed] [Google Scholar]
229.Gale DJ, Flanagan JR & Gallivan JP Human somatosensory cortex is modulated during motor planning. J. Neurosci 41, 5909–5922 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
230.Ariani G, Pruszynski JA & Diedrichsen J Motor planning brings human primary somatosensory cortex into action-specific preparatory states. Elife 11, 69517 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
231.Segawa JA, Tourville JA, Beal DS & Guenther FH The neural correlates of speech motor sequence learning. J. Cogn. Neurosci 27, 819–831 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
232.Penfield W & Rasmussen T The Cerebral Cortex of Man; a Clinical Study of Localization of Function. (Macmillan, Oxford, England, 1950). [Google Scholar]
233.Bouchard KE, Mesgarani N, Johnson K & Chang EF Functional organization of human sensorimotor cortex for speech articulation. Nature 495, 327–332 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
234.Darley FL, Aronson AE & Brown JR Differential diagnostic patterns of dysarthria. J. Speech Hear. Res 12, 246–269 (1969). [DOI] [PubMed] [Google Scholar]
235.Mackenzie C Dysarthria in stroke: a narrative review of its description and the outcome of intervention. Int. J. Speech Lang. Pathol 13, 125–136 (2011). [DOI] [PubMed] [Google Scholar]
236.Eberhard DM, Simons GF & Fennig CD Ethnologue: Languages of the World. (SIL International, 2023). [Google Scholar]
237.Dejerine J Sur un cas de cécité verbale avec agraphie suivi d’autopsie [French]. Mémoires de la Société de Biologie 3, 197–201 (1891). [Google Scholar]
238.McCandliss BD, Cohen L & Dehaene S The visual word form area: expertise for reading in the fusiform gyrus. Trends Cogn. Sci 7, 293–299 (2003). [DOI] [PubMed] [Google Scholar]
239.Glezer LS & Riesenhuber M Individual variability in location impacts orthographic selectivity in the “visual word form area”. J. Neurosci 33, 11221–11226 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
240.Saygin ZM et al. Connectivity precedes function in the development of the visual word form area. Nat. Neurosci 19, 1250–1255 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
241.Hirshorn EA et al. Decoding and disrupting left midfusiform gyrus activity during word reading. Proc. Natl Acad. Sci 113, 8162–8167 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
242.Tainturier MJ & Rapp B The spelling process. in Handbook of Cognitive Neuropsychology 263–289 (2015). [Google Scholar]
243.Longcamp M et al. Functional specificity in the motor system: evidence from coupled fMRI and kinematic recordings during letter and digit writing. Hum. Brain Mapp 35, 6077–6087 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
244.Baumann A et al. Neural correlates of executed compared to imagined writing and drawing movements: a functional magnetic resonance imaging study. Front. Hum. Neurosci 16, 829576 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
245.Nespoulous JL et al. Agrammatism in sentence production without comprehension deficits: reduced availability of syntactic structures and/or of grammatical morphemes? a case study. Brain Lang. 33, 273–295 (1988). [DOI] [PubMed] [Google Scholar]
246.Levy DF et al. Apraxia of speech with phonological alexia and agraphia following resection of the left middle precentral gyrus: illustrative case. Journal of Neurosurgery: Case Lessons 5, (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
247.Roux FE et al. The graphemic/motor frontal area Exner’s area revisited. Ann. Neurol 66, 537–545 (2009). [DOI] [PubMed] [Google Scholar]
248.Roux FE, Draper L, Köpke B & Démonet JF Who actually read Exner? Returning to the source of the frontal ‘writing centre’ hypothesis. Cortex 46, 1204–1210 (2010). [DOI] [PubMed] [Google Scholar]
249.Warren RM Perceptual restoration of missing speech sounds. Science 167, 392–393 (1970). [DOI] [PubMed] [Google Scholar]
250.Ganong WF Phonetic categorization in auditory word perception. J. Exp. Psychol. Hum. Percept. Perform 6, 110–125 (1980). [DOI] [PubMed] [Google Scholar]
251.Samuel AG Phonemic restoration: insights from a new methodology. J. Exp. Psychol. Gen 110, 474–494 (1981). [DOI] [PubMed] [Google Scholar]
252.Cogan GB et al. Sensory-motor transformations for speech occur bilaterally. Nature 507, 94–98 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
253.Mahowald K et al. Dissociating language and thought in large language models. Trends in Cogn. Neurosci (in press). [DOI] [PMC free article] [PubMed] [Google Scholar]
254.Duncan J & Owen AM Common regions of the human frontal lobe recruited by diverse cognitive demands. Trends. Neurosci 23, 475–483 (2000). [DOI] [PubMed] [Google Scholar]
255.Cole MW & Schneider W The cognitive control network: Integrated cortical regions with dissociable functions. Neuroimage 37, 343–360 (2007). [DOI] [PubMed] [Google Scholar]
256.Fedorenko E, Duncan J & Kanwisher N Broad domain generality in focal regions of frontal and parietal cortex. Proc. Natl Acad. Sci 110, 16616–16621 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
257.Hugdahl K, Raichle ME, Mitra A & Specht K On the existence of a generalized non-specific task-dependent network. Front. Hum. Neurosci 9, 430 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
258.Shashidhara S, Mitchell DJ, Erez Y & Duncan J Progressive recruitment of the frontoparietal multiple-demand system with increased task complexity, time pressure, and reward. J. Cogn. Neurosci 31, 1617–1630 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
259.Assem M, Glasser MF, Essen DC & Duncan J A domain-general cognitive core defined in multimodally parcellated human cortex. Cereb. Cortex 30, 4361–4380 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
260.Assem M, Blank IA, Mineroff Z, Ademoğlu A & Fedorenko E Activity in the fronto-parietal multiple-demand network is robustly associated with individual differences in working memory and fluid intelligence. Cortex 131, 1–16 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
261.Duncan J Disorganisation of behaviour after frontal lobe damage. Cogn. Neuropsychol 3, 271–290 (1986). [Google Scholar]
262.Miller EK & Cohen JD An integrative theory of prefrontal cortex function. Annu. Rev. Neurosci 24, 167–202 (2001). [DOI] [PubMed] [Google Scholar]
263.Woolgar A et al. Fluid intelligence loss linked to restricted regions of damage within frontal and parietal cortex. Proc. Natl Acad. Sci 107, 14899–14902 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
264.Gläscher J et al. Distributed neural system for general intelligence revealed by lesion mapping. Proc. Natl Acad. Sci 107, 4705–4709 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
265.Mineroff Z, Blank IA, Mahowald K & Fedorenko E A robust dissociation among the language, multiple demand, and default mode networks: evidence from inter-region correlations in effect size. Neuropsychologia 119, 501–511 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
266.Woolgar A, Duncan J, Manes F & Fedorenko E The multiple-demand system but not the language system supports fluid intelligence. Nat. Hum. Behav 2, 200–204 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
267.MacGregor LJ et al. Causal contributions of the domain-general (Multiple Demand) and the language-selective brain networks to perceptual and semantic challenges in speech comprehension. Neurobiol. Lang 3, 665–698 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
268.Wehbe L et al. Incremental language comprehension difficulty predicts activity in the language network but not the Multiple Demand network. Cereb. Cortex 31, 4006–4023 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
269.Monti MM, Osherson DN, Martinez MJ & Parsons LM Functional neuroanatomy of deductive inference: a language-independent distributed network. Neuroimage 37, 1005–1016 (2007). [DOI] [PubMed] [Google Scholar]
270.Wild CJ et al. Effortful listening: the processing of degraded speech depends critically on attention. J. Neurosci 32, 14010–14021 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
271.Sherafati A et al. Prefrontal cortex supports speech perception in listeners with cochlear implants. Elife 11, 75323 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
272.Brownsett SLE et al. Cognitive control and its impact on recovery from aphasic stroke. Brain 137, 242–254 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
273.Geranmayeh F, Wise RJS, Mehta A & Leech R Overlapping networks engaged during spoken language production and its cognitive control. J. Neurosci 34, 8728–8740 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
274.Hartwigsen G & Saur D Neuroimaging of stroke recovery from aphasia – insights into plasticity of the human language network. Neuroimage 190, 14–31 (2019). [DOI] [PubMed] [Google Scholar]
275.Turkeltaub PE A taxonomy of brain–behavior relationships after stroke. Journal of Speech, Language, and Hearing Research 62, 3907–3922 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
276.Stefaniak JD, Halai AD & Lambon Ralph MA The neural and neurocomputational bases of recovery from post-stroke aphasia. Nat. Rev. Neurol 16, 43–55 (2020). [DOI] [PubMed] [Google Scholar]
277.Wilson SM & Schneck SM Neuroplasticity in post-stroke aphasia: a systematic review and meta-analysis of functional imaging studies of reorganization of language processing. Neurobiol. Lang 2, 22–82 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
278.De Clercq P, Gonsalves AR, Gerrits R & Vandermosten M Individualized functional localization of the language and multiple demand network in chronic post-stroke aphasia. Preprint at bioRxiv 10.1101/2024.01.12.575350 (2024). [DOI] [Google Scholar]
279.Billot A Ph.D. Thesis. (Boston University, 2023). [Google Scholar]
280.Billot A et al. Robust dissociation between the language and Multiple Demand networks in aging and after a stroke. in Organization of Human Brain Mapping (Montreal, Canada, 2023). [Google Scholar]
281.Gallagher HL et al. Reading the mind in cartoons and stories: an fMRI study of ‘theory of mind’ in verbal and nonverbal tasks. Neuropsychologia 38, 11–21 (2000). [DOI] [PubMed] [Google Scholar]
282.Saxe R & Powell LJ It’s the thought that counts: specific brain regions for one component of theory of mind. Psychol. Sci 17, 692–699 (2006). [DOI] [PubMed] [Google Scholar]
283.Jacoby N, Bruneau E, Koster-Hale J & Saxe R Localizing Pain Matrix and Theory of Mind networks with both verbal and non-verbal stimuli. Neuroimage 126, 39–48 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
284.Vogeley K et al. Mind reading: neural mechanisms of theory of mind and self-perspective. Neuroimage 14, 170–181 (2001). [DOI] [PubMed] [Google Scholar]
285.Ruby P & Decety J What you believe versus what you think they believe: a neuroimaging study of conceptual perspective-taking. Eur. J. Neurosci 17, 2475–2480 (2003). [DOI] [PubMed] [Google Scholar]
286.Saxe R, Schulz LE & Jiang YV Reading minds versus following rules: dissociating theory of mind and executive control in the brain. Soc. Neurosci 1, 284–298 (2006). [DOI] [PubMed] [Google Scholar]
287.Koster-Hale J & Saxe R Theory of mind: a neural prediction problem. Neuron 79, 836–848 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
288.Bruneau EG, Pluta A & Saxe R Distinct roles of the ‘shared pain’ and ‘theory of mind’ networks in processing others’ emotional suffering. Neuropsychologia 50, 219–231 (2012). [DOI] [PubMed] [Google Scholar]
289.Willems RM, Benn Y, Hagoort P, Toni I & Varley R Communicating without a functioning language system: implications for the role of language in mentalizing. Neuropsychologia 49, 3130–3135 (2011). [DOI] [PubMed] [Google Scholar]
290.Dronkers NF, Ludy CA & Redfern BB Pragmatics in the absence of verbal language: descriptions of a severe aphasic and a language-deprived adult. J Neurolinguistics 11, 179–190 (1998). [Google Scholar]
291.Spotorno N, Koun E, Prado J, Henst JB & Noveck IA Neural evidence that utterance-processing entails mentalizing: The case of irony. Neuroimage 63, 25–39 (2012). [DOI] [PubMed] [Google Scholar]
292.van Ackeren MJ, Casasanto D, Bekkering H, Hagoort P & Rueschemeyer S-A Pragmatics in action: indirect requests engage Theory of Mind areas and the cortical motor network. J. Cogn. Neurosci 24, 2237–2247 (2012). [DOI] [PubMed] [Google Scholar]
293.Jang G et al. Everyday conversation requires cognitive inference: neural bases of comprehending implicated meanings in conversations. Neuroimage 81, 61–72 (2013). [DOI] [PubMed] [Google Scholar]
294.Feng W et al. Effects of contextual relevance on pragmatic inference during conversation: an fMRI study. Brain Lang. 171, 52–61 (2017). [DOI] [PubMed] [Google Scholar]
295.Hagoort P & Levinson SC Neuropragmatics. in The Cognitive Neurosciences (eds. Gazzaniga MS & Mangun GR) 667–674 (The MIT Press, 2014). doi: 10.7551/mitpress/9504.003.0073. [DOI] [Google Scholar]
296.Hauptman M, Blank I & Fedorenko E Non-literal language processing is jointly supported by the language and theory of mind networks: evidence from a novel meta-analytic fMRI approach. Cortex 162, 96–114 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
297.Ferstl EC & von Cramon DY What does the frontomedian cortex contribute to language processing: coherence or theory of mind? Neuroimage 17, 1599–1612 (2002). [DOI] [PubMed] [Google Scholar]
298.Buckner RL, Andrews-Hanna JR & Schacter DL The brain’s default network: anatomy, function, and relevance to disease. Ann. NY Acad. Sci 1124, 1–38 (2008). [DOI] [PubMed] [Google Scholar]
299.Spreng RN & Grady CL Patterns of brain activity supporting autobiographical memory, prospection, and theory of mind, and their relationship to the default mode network. J. Cogn. Neurosci 22, 1112–1123 (2010). [DOI] [PubMed] [Google Scholar]
300.Mars RB et al. On the relationship between the ‘default mode network’ and the ‘social brain’. Front. Hum. Neurosci 6, 189 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
301.Braga RM & Buckner RL Parallel interdigitated distributed networks within the individual estimated by intrinsic functional connectivity. Neuron 95, 457–471 5 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
302.DiNicola LM, Braga RM & Buckner RL Parallel distributed networks dissociate episodic and social functions within the individual. J. Neurophysiol 123, 1144–1179 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
303.Gusnard DA, Akbudak E, Shulman GL & Raichle ME Medial prefrontal cortex and self-referential mental activity: relation to a default mode of brain function. Proc. Natl Acad. Sci 98, 4259–4264 (2001). [DOI] [PMC free article] [PubMed] [Google Scholar]
304.Buckner RL & Carroll DC Self-projection and the brain. Trends Cogn. Sci 11, 49–57 (2007). [DOI] [PubMed] [Google Scholar]
305.Hassabis D & Maguire EA Deconstructing episodic memory with construction. Trends Cogn. Sci 11, 299–306 (2007). [DOI] [PubMed] [Google Scholar]
306.Chang CH, Nastase SA & Hasson U Information flow across the cortical timescale hierarchy during narrative construction. Proc. Natl Acad. Sci 119, 2209307119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
307.Zadbood A, Nastase S, Chen J, Norman KA & Hasson U Neural representations of naturalistic events are updated as our understanding of the past changes. Elife 11, (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
308.Baldassano C, Hasson U & Norman KA Representation of real-world event schemas during narrative perception. J. Neurosci 38, 9689–9699 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
309.Baldassano C et al. Discovering event structure in continuous narrative perception and memory. Neuron 95, 709–721 5 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
310.Parvizi J et al. Altered sense of self during seizures in the posteromedial cortex. Proc. Natl Acad. Sci 118, 2100522118 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
311.Philippi CL et al. Lesion network mapping demonstrates that mind-wandering is associated with the default mode network. J. Neurosci. Res 99, 361–373 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
312.Ferstl EC, Neumann J, Bogler C & von Cramon DY The extended language network: a meta-analysis of neuroimaging studies on text comprehension. Hum. Brain Mapp 29, 581–593 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
313.Simony E et al. Dynamic reconfiguration of the default mode network during narrative comprehension. Nat. Commun 7, 12141 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
314.Lenartowicz A, Kalar DJ, Congdon E & Poldrack RA Towards an ontology of cognitive control. Top. Cogn. Sci 2, 678–692 (2010). [DOI] [PubMed] [Google Scholar]
315.Poldrack RA et al. The cognitive atlas: toward a knowledge foundation for cognitive neuroscience. Front. Neuroinform 5, 17 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
316.Kanwisher N Functional specificity in the human brain: a window into the functional architecture of the mind. Proc. Natl Acad. Sci 107, 11163–11170 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
317.Humphries C, Binder JR, Medler DA & Liebenthal E Time course of semantic processes during sentence comprehension: an fMRI study. Neuroimage 36, 924–932 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
318.Lambon Ralph MA, Jefferies E, Patterson K & Rogers TT The neural and computational bases of semantic cognition. Nat. Rev. Neurosci 18, 42–55 (2017). [DOI] [PubMed] [Google Scholar]
319.Matheson HE & Barsalou LW Embodiment and grounding in cognitive neuroscience. Stevens’ Handbook of Experimental Psychology and Cognitive Neuroscience 3, 1–27 (2018). [Google Scholar]
320.Tanenhaus MK, Spivey-Knowlton MJ, Eberhard KM & Sedivy JC Integration of visual and linguistic information in spoken language comprehension. Science 268, 1632–1634 (1995). [DOI] [PubMed] [Google Scholar]
321.Ferreira F The misinterpretation of noncanonical sentences. Cogn. Psychol 47, 164–203 (2003). [DOI] [PubMed] [Google Scholar]
322.Hagoort P, Hald L, Bastiaansen M & Petersson KM Integration of word meaning and world knowledge in language comprehension. Science 304, 438–441 (2004). [DOI] [PubMed] [Google Scholar]
323.Van Berkum JJ, Brink D, Tesink CM, Kos M & Hagoort P The neural integration of speaker and message. J. Cogn. Neurosci 20, 580–591 (2008). [DOI] [PubMed] [Google Scholar]
324.Gibson E, Bergen L & Piantadosi S Rational integration of noisy evidence and prior semantic expectations in sentence interpretation. Proc. Natl Acad. Sci 110, 8051–8056 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
325.Zhang Y, Frassinelli D, Tuomainen J, Skipper JI & Vigliocco G More than words: word predictability, prosody, gesture and mouth movements in natural language comprehension. Proc.R. Soc. B Biol. Sci 288, 20210500 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
326.Elman JL On the meaning of words and dinosaur bones: lexical knowledge without a lexicon. Cogn. Sci 33, 547–582 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
327.Hasson U, Egidi G, Marelli M & Willems RM Grounding the neurobiology of language in first principles: the necessity of non-language-centric explanations for language comprehension. Cognition 180, 135–157 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
328.Bedny M & Caramazza A Perception, action, and word meanings in the human brain: the case from action verbs. Ann. NY Acad. Sci 1224, 81–95 (2011). [DOI] [PubMed] [Google Scholar]
329.Pereira F, Gershman S, Ritter S & Botvinick M A comparative evaluation of off-the-shelf distributed semantic representations for modelling behavioural data. Cogn. Neuropsychol 33, 175–190 (2016). [DOI] [PubMed] [Google Scholar]
330.Grand G, Blank IA, Pereira F & Fedorenko E Semantic projection recovers rich human knowledge of multiple object features from word embeddings. Nat. Hum. Behav 6, 975–987 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
331.Paulk AC et al. Large-scale neural recordings with single neuron resolution using Neuropixels probes in human cortex. Nat. Neurosci 25, 252–263 (2022). [DOI] [PubMed] [Google Scholar]
332.Yamins DL & DiCarlo JJ Using goal-driven deep learning models to understand sensory cortex. Nat. Neurosci 19, 356–365 (2016). [DOI] [PubMed] [Google Scholar]
333.Sussillo D, Churchland MM, Kaufman MT & Shenoy KV A neural network that finds a naturalistic solution for the production of muscle activity. Nat. Neurosci 18, 1025–1033 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
334.Tuckute G, Feather J, Boebinger D & McDermott JH Many but not all deep neural network audio models capture brain responses and exhibit correspondence between model stages and brain regions. PLOS Biol. 21, e3002366 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
335.Tuckute G, Kanwisher N & Fedorenko E Language in brains, minds, and machines. Annu. Rev. Neurosci 47, (2024). [DOI] [PubMed] [Google Scholar]
336.Wong L et al. From word models to world models: translating from natural language to the probabilistic language of thought. Preprint at arXiv 10.48550/arXiv.2306.12672 (2023). [DOI] [Google Scholar]
337.Stager CL & Werker JF Infants listen for more phonetic detail in speech perception than in word-learning tasks. Nature 388, 381–382 (1997). [DOI] [PubMed] [Google Scholar]
338.Werker JF & Tees RC Influences on infant speech processing: toward a new synthesis. Annu. Rev. Psychol 50, 509–535 (1999). [DOI] [PubMed] [Google Scholar]
339.Oller DK, Wieman LA, Doyle WJ & Ross C Infant babbling and speech. J. Child Lang 3, 1–11 (1976). [Google Scholar]
340.Oller DK The Emergence of the Sounds of Speech in Infancy. in Child Phonology 1: Production (eds. Yeni-Komshian G, Kavanagh J & Ferguson CA) 93–1123 (NY: Academic Press, New York, 1980). [Google Scholar]
341.Wellman HM, Cross D & Watson J Meta-analysis of theory-of-mind development: the truth about false belief. Child. Dev 72, 655–684 (2001). [DOI] [PubMed] [Google Scholar]
342.Best JR & Miller P H A developmental perspective on executive function. Child. Dev 81, 1641–1660 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
343.Fedorenko E & Varley R Language and thought are not the same thing: evidence from neuroimaging and neurological patients. Ann. NY Acad. Sci 1369, 132–153 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
344.Pontillo D, Salverda AP & Tanenhaus MK Flexible Use of Phonological and Visual Memory in Language-mediated Visual Search. in Proceedings of the 38th Annual Meeting of the Cognitive Science Society (2015). [Google Scholar]
345.Frank MC, Everett DL, Fedorenko E & Gibson E Number as a cognitive technology: evidence from Pirahã language and cognition. Cognition 108, 819–824 (2008). [DOI] [PubMed] [Google Scholar]
346.Lupyan G, Rakison DH & McClelland JL Language is not just for talking: redundant labels facilitate learning of novel categories. Psychol. Sci 18, 1077–1083 (2007). [DOI] [PubMed] [Google Scholar]
347.Brojde CL, Porter C & Colunga E Words can slow down category learning. Psychon. Bull. Rev 18, 798–804 (2011). [DOI] [PubMed] [Google Scholar]
348.Lupyan G & Casasanto D Meaningless words promote meaningful categorization. Lang. Cogn 7, 167–193 (2015). [Google Scholar]
349.Lupyan G & Mirman D Linking language and categorization: evidence from aphasia. Cortex 49, 1187–1194 (2013). [DOI] [PubMed] [Google Scholar]
350.Sapir E Language: An Introduction to the Study of Speech. (Harcourt, Brace & World Inc, 1921). [Google Scholar]
351.Carruthers P The cognitive functions of language. Behav. Brain Sci 25, 657–725 (2002). [DOI] [PubMed] [Google Scholar]
352.Brett M, Johnsrude IS & Owen AM The problem of functional localization in the human brain. Nat. Rev. Neurosci 3, 243–249 (2002). [DOI] [PubMed] [Google Scholar]
353.Kanwisher N, McDermott J & Chun MM The fusiform face area: a module in human extrastriate cortex specialized for face perception. J. Neurosci 17, 4302–4311 (1997). [DOI] [PMC free article] [PubMed] [Google Scholar]
354.Tootell RB et al. Functional analysis of human MT and related visual cortical areas using magnetic resonance imaging. J. Neurosci 15, 3215–3230 (1995). [DOI] [PMC free article] [PubMed] [Google Scholar]
355.Frost MA & Goebel R Measuring structural–functional correspondence: spatial variability of specialised brain regions after macro-anatomical alignment. Neuroimage 59, 1369–1381 (2012). [DOI] [PubMed] [Google Scholar]
356.Mueller S et al. Individual variability in functional connectivity architecture of the human brain. Neuron 77, 586–595 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
357.Haxby JV, Guntupalli JS, Nastase SA & Feilong M Hyperalignment: modeling shared information encoded in idiosyncratic cortical topographies. Elife 9, (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
358.Salvo JJ, Holubecki AM & Braga RM Correspondence between functional connectivity and task-related activity patterns within the individual. Curr. Opin. Behav. Sci 40, 178–188 (2021). [Google Scholar]
359.Tsao DY, Freiwald WA, Tootell RB & Livingstone MS A cortical region consisting entirely of face-selective cells. Science 311, 670–674 (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]
360.Bruffaerts R et al. Functional identification of language-responsive channels in individual participants in MEG investigations. Preprint at bioRxiv 10.1101/2023.03.23.533424 (2023). [DOI] [PubMed] [Google Scholar]
361.Liu J, Harris A & Kanwisher N Stages of processing in face perception: an MEG study. Nat. Neurosci 5, 910–916 (2002). [DOI] [PubMed] [Google Scholar]
362.Poeppel D, Emmorey K, Hickok G & Pylkkänen L Towards a new neurobiology of language. J. Neurosci 32, 14125–14131 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
363.Biswal B, Zerrin Yetkin F, Haughton VM & Hyde JS Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn. Reson. Med 34, 537–541 (1995). [DOI] [PubMed] [Google Scholar]
364.Poldrack RA & Yarkoni T From brain maps to cognitive ontologies: informatics and the search for mental structure. Annu. Rev. Psychol 67, 587–612 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
365.Tannen D, Hamilton HE & Schiffrin D The Handbook of Discourse Analysis. (John Wiley & Sons, 2015). [Google Scholar]
366.Pessoa L The Entangled Brain. (MIT Press, 2022). [Google Scholar]
367.Buzsáki G The Brain from Inside Out. (Oxford University Press, 2019). [Google Scholar]
368.Buzsáki G The brain–cognitive behavior problem: a retrospective. eNeuro 7, 0069–20 2020 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supplementary

NIHMS2142018-supplement-supplementary.pdf^{(562.1KB, pdf)}

PERMALINK

The language network as a natural kind within the broader landscape of the human brain

Evelina Fedorenko

Anna A Ivanova

Tamar I Regev

Abstract

1. Introduction

Box 1∣. Beyond coarse anatomy: The importance of functional localization.

Fig. 1 ∣. Brain systems supporting language comprehension and language production.

2. The language network as a natural kind

Fig. 2 ∣. The topography of the language network, its variability across individuals, its stability within individuals over time, and its selectivity for language over non-linguistic inputs and tasks.

2a. Anatomy and inter-connectivity

2b. Responsiveness to different kinds of language

2c. Causal role in language ability

2d. Selectivity for language versus non-linguistic inputs and tasks

Box 2∣. The language network versus language: ontological kinds at different levels.

3. The internal structure and core computations of the language network

3a. Sensitivity to regularities at sub-lexical, word, and phrase and sentence level, but not discourse level

Fig. 3 ∣. Sensitivity of the language network to linguistic structure at multiple information scales, and functional similarity of its different areas.

3b. Functional differences among the language areas?

Past claims about dissociations within the language network.

Similar functional profiles for the different language network areas.

Making meaningful progress in understanding the internal architecture of the language network.

4. The language network and language-relevant perceptual and premotor areas

Fig. 4 ∣. Sensitivity to meaning in the language network, but not in specialized perceptual and premotor areas.

4a. Speech

Speech perception.

Speech articulation.

4b. Reading and writing

Reading.

Writing.

5. The cognitive networks the language network interacts with to support real-life language use

Fig. 5 ∣. Selectivity for language in the language network, but not in systems of knowledge and reasoning.

5a. Multiple demand network: task demands beyond comprehension and production

Network properties.

Dissociation from the language network.

Contributions to language.

5b. Theory of mind network: pragmatics and social reasoning

Network properties.

Dissociation from the language network.

Contributions to language.

5c. The Default Mode network: narratives and situation modeling

Network properties.

Dissociation from the language network.

Contributions to language.

6. Open questions and a way forward

6a. The problem of meaning

6b. Sharp edges versus functional gradients

6c. A mechanistic understanding of language processing

6d. The role of language in development, evolution and cognition

6e. Language as a tool to augment thought

Supplementary Material

References:

Associated Data

Supplementary Materials

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases