Functional specificity for high-level linguistic processing in the human brain

“…patching across 46 tasks spanning four cognitive domains (Figure 1A), each corresponding to a distinct network in the human brain: language (Fedorenko et al., 2011, 2024a), multiple demand (Duncan, 2010; Assem et al., 2020), physical reasoning (Fischer et al., 2016), and theory of mind…”

“Studies now routinely compare model features with fMRI, ECoG, MEG, eye tracking, and behavioral data in order to ask whether model representations capture brain-relevant structure [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18].”

“This could explain why language processing recruits substantially different circuitry [Fedorenko et al., 2011].”

“…left-hemisphere regions, this system responds when we listen to or read language [27-28] or when we speak [29], but not when we engage in apparently similar but nonlinguistic processes, such as arithmetic, music perception, understanding gestures, or reasoning [30-32] (for reviews, see [10,33]).”

“…regions during easy and hard math localizer trials (adding three consecutive addends to a two-digit number to compute a final sum) relative to rest, as well as a whole brain analysis that showed frontoparietal activation in a hard > easy arithmetic contrast (Fedorenko, Behr, & Kanwisher, 2011).”

“Future studies using nonsensical Braille words as tactile controls, as done in previous research (Baldassano et al., 2018; Fedorenko et al., 2011; Kim et al., 2017; Pallier et al., 2011; Reich et al., 2011) are warranted.”

“While responses were recorded for the whole brain, we focus on two specific regions, involved in either linguistic or visual processing: the left-hemisphere Language network (Fedorenko et al., 2011) and the Visual network (Power et al., 2011; Buckner et al., 2008).”

“It is known to be crucial to higher-level linguistic processing, e.g., syntactic and semantic processing on the sentence level (Fedorenko et al., 2011; Hickok, 2022; Mollica et al., 2020; Stockert et al., 2023; Tyler et al., 2010; Vlooswijk et al., 2010; Malik-Moraleda et al., 2022).”

“…particular, therefore, have often relied on alternative visual presentation techniques which standardise temporal exposure to written sentence content across participants, such as phrase-by-phrase or word-by-word presentation (e.g. Fedorenko et al., 2011; Hoenig & Scheef, 2009; Zhu et al., 2012).”

“These regions were in accord with the high-level 276 linguistic network (Fedorenko et al., 2011; Friederici, 2011) and are considered to underly language277 supported conceptual processing (Xu et al.”

“This network consists of left lateral frontal and lateral temporal areas and selectively supports language comprehension and production (e.g., Fedorenko et al., 2011).”

“However, this network shows virtually no engagement in non-linguistic tasks [11-14] and is therefore languageselective.”

“Much recent work aims to characterize the functional specificity of sub-regions both within in the IFG (Fiebach et al., 2006; Fedorenko et al., 2011) and the STG (Indefrey and Levelt, 2004) and to more carefully specify the functional anatomy of language (Poeppel and Hickok, 2004).”

“The language network responds in a highly selective manner to linguistic inputs (e.g., Fedorenko et al. 2011; Monti et al. 2012; Chen et al. 2023; see Fedorenko et al. 2024 for a recent review) and shows sensitivity to linguistic structure at different levels, from sound patterns to sentences…”

“The visuospatial working memory experiment w as adopted from Fedorenko et al. (2011), with an added perceptual contr ol (color) condition (Fig.”

“…52 , and several aspects of cognition (executive control supported by the multiple demand network or MD, intuitive physical reasoning, language, theory of mind or TOM, episodic memory and prospection or episodic projection, and functions supported by the default mode network or DMN) 3–5,53– 56 .”

“In neurolinguistics, growing evidence suggests that both syntactic and semantic combinatorial processes recruit a left-lateralized fronto-temporal network, including inferior frontal and superior temporal regions [FBK11].”

“Subject-specific fROIs were defined using a functionally constrained, anatomically guided approach (Fedorenko et al., 2011; Nieto-Casta ˜ n ´ on and Fedorenko, 2012; Abassi and Papeo, 2024; Munin et al., 2025).”

“…factors may explain this: First, commonly employed cognitive assessments (e.g., verbal fluency, naming tests) engage hybrid processes involving both executive control and semantic retrieval — the latter typically reliant on the LAN (Benn et al., 2023; Fedorenko et al., 2011; Blank et al., 2014).”

“…directly comparing the network that supports phonological processing of written words and arithmetic concurrently has been severely lacking, with the exception of only a handful of studies (Andin et al. 2015; Evans et al. 2016; Fedorenko et al. 2011; Prado et al. 2011; Simon et al. 2002).”

“As children become more proficient readers, brain responses to complex language processing within the bilateral frontotemporal language network become increasingly more left-lateralized which is thought to reflect increasing specialisation to language-related tasks 7,8,9,10,11 .”

“…memory tasks Bilingualism: Language and Cognition 13 https://doi.org/10.1017/S1366728925100138 Published online by Cambridge University Press (Fedorenko et al., 2011), particularly in representing and maintaining the complex structures (see also Hickok & Poeppel, 2007), the connection from…”

“First, LLMs can be directly trained on the specific task of thematic role assignment (also called “semantic role labeling”) via supervised learning, wherein training sentences are paired with correct role labels for different words.”

“These tasks include the ability to differentiate between the pitch of two neighboring piano keys and appreciate whether a melody is melodious when processing external auditory stim‑ uli (Giuliano et al., 2011; Fedorenko et al., 2011).”

“The neuroscience field has recently provided some clues to answer this question of language-thought dissociation (Fedorenko et al., 2011; Blank et al., 2014; Fedorenko et al., 2024a); specifically, brain activities during processing linguistic and non-linguistic reasoning stimuli is different…”

“Functional connectivity studies have shown that other language-related brain regions are predominantly located in the left temporal lobe, left inferior frontal gyrus, and left middle frontal gyrus, as well as areas extending into the angular gyrus [4,5].”

“Furthermore, this network has been shown to be strongly selective for linguistic input (Fedorenko et al., 2011)(see Fedorenko and Blank (2020) for a review).”

“…et al., 2008; Binder et al., 2009), Language Network (related to language processing, understanding, word meaning, and sentence comprehension) (Fedorenko et al., 2011), Task Positive Network (related to attention, salience information) (Binder et al., 2009; Duncan, 2010; Power et al., 2011),…”

“…consistent results, showing that other nodes within the language network (e.g., superior temporal and inferior parietal cortices) do not show any sensitivity to cognitive demand/difficulty (Fedorenko et al., 2011), whereas nodes within the MDS do exhibit such a sensitivity (Fedorenko et al., 2013).”

“To further investigate the relation of math statements to language, we performed a sensitive analysis in 7 regions of interest associated with syntactical processing in previous studies: temporal pole (TP), anterior and posterior superior temporal sulcus (aSTS and pSTS), temporo-parietal junction (TPJ), inferior front…”

“These include word recall; a sentence-versus-pseudoword-list task adapted from the human language localizer (Fedorenko et al., 2011); multi-step deduction over invented categories; MMLU questions (Hendrycks et al., 2020); a Stroop variant (Stroop, 1935) and an adaptation of the Multi-Source…”

“The spatial WM task was introduced in Fedorenko et al. (2011) and has been used in many subsequent studies as a localizer for the multiple demand system (Blank et al., 2014; Shashidhara et al., 2019, 2020, 2024; Diachek et al., 2020; Malik-556 Moraleda et al., 2022).”

“In particular, the multiple-demand (MD) network has been linked to domain-general executive control (Duncan, 2010; Duncan & Owen, 2000; Duncan, 2013), while the language network has been associated with linguistic processing (Fedorenko et al., 2011; Fedorenko et al., 2012).”

“We first examined the adults’ responses to these conditions, expecting to find higher responses to language than music or foreign speech in individually-defined language regions, as demonstrated previously (Chen et al., 2023; Deen et al., 2015; Fedorenko et al., 2011).”

“In individual participants, this network is functionally distinct from brain regions that support language ( Fedorenko et al., 2011; Paunov et al., 2022), attention ( Mars et al., 2012; Scholz et al., 2009), and executive functions ( Saxe, Schulz, et al., 2006).”

“…the HCP dataset, the ‘language’ contrast, story-math, likely recruits two different functional networks: the language network, which is sensitive to sentence-level meaning (Fedorenko et al., 2011) and the default network, which tracks story-level meanings (Braga et al., 2020; Simony et al., 2016).”

“Furthermore, work by Fedorenko et al. indicates that the MD network is not highly engaged during language processing when task demands remain low and are linguistic in nature (Diachek et al., 2020; Fedorenko et al., 2011, 2024; MacGregor et al., 2022), a distinction we return to in the section.”

“Participants completed three in-scanner tasks: a language localizer (Scott et al., 2017), a verbal working memory task (Fedorenko et al, 2011), and a spatial working memory task (Fedorenko et al, 2011; Lee et al., 2024).”

“In adults, language processing is characterized by a high degree of functional specialization and neural efficiency, primarily localized within a left-lateralized fronto-temporal network (Fedorenko et al., 2011; Mesulam, 2016; Vigneau et al., 2006).”

“…recruit left-hemisphere (and, to a lesser extent, right-hemisphere) temporal regions, as well as cerebellar regions associated with language, more robustly during higher-level speech processing, consistent with models of language processing (Fedorenko et al., 2011; Malik-Moraleda et al., 2022).”

“Language comprehension and production are predominantly supported by the core left-lateralized language-related network (LRN) , which encompasses the lateral frontal and temporal brain regions (Tyler et al. 2011; Fedorenko et al. 2011; for review, see Price 2012; Fedorenko et al. 2024).”

“We operationalize this as: the IFG (associated with syntactic processing; Fedorenko et al. 2011) should show a steeper alignment gradient with typological distance than the PTL (associated with lexico-semantic processing; Hickok and Poeppel 2007).”

“Evidence suggests that contrasting Sentence > Nonwords blocks reliably identifies high-level language processing regions, including in the LPFC. 37,52 Auditory language task: The auditory language task was adapted from Scott et al. 86 using publicly available code and stimuli…”

“Encompassing Broca’s area, the left IFG is involved in semantic processing of word reading (Fedorenko et al., 2011; Sahin et al., 2009), as well as language comprehension, production, and complex higher-order linguistic processes (Bitan et al., 2009; Friederici, 2011; Pollack et al., 2015).”

“This normalization ensures that the LPI captures relative language specificity rather than absolute effect size differences [34].”

“Through stimulus-evoked or task-evoked functional magnetic resonance imaging (fMRI), many studies have characterized localized neural responses across domains such as visual, language, and motor controls (Brett et al., 2002; Fedorenko et al., 2011; Genon et al., 2018; Tootell et al., 1995).”

“The human language network (LN) comprises a set of brain regions that collectively support the core components of language processing, including comprehension, production, and integration of linguistic information (Binder et al., 1997; Fedorenko et al., 2011).”

“Language is one of the most extensively studied brain functions, in both clinical (Bookheimer 2002; Saur and Hartwigsen 2012) and basic research settings (Fedorenko et al. 2011; Hickok 2012), owing to its central importance for the human experience (Tomasello 2009; Fitch 2010).”

“We were especially struck by the hemispheric dissociation of the LANG network in both studies, as this network is the most consistently anatomically left-lateralized association network (e.g., Braga et al., 2020; Fedorenko, Behr, and Kanwisher, 2011; Lipkin et al., 2022).”

“Dorsolateral prefrontal cortex (DLPFC) has been suggested to be an additional substrate for semantic processing (Bottini et al., 1994; Fedorenko et al., 2011), in a subregion partially distinct from the one serving general executive functions (Fedorenko et al., 2011).”

“Other networks potentially involved underlie language (language network) [124,125] and momentary task control (the fronto-parietal network) [126–129], the latter of which may be engaged, on a fast timescale, when confronted with alcohol “in the moment”.”

“Functional mapping methods demonstrate that language-specific regions exhibit minimal activation during tasks such as mental arithmetic, working-memory maintenance, or cognitive control 72–74 .”

“Neuroimaging studies corroborate this dissociation, showing limited fMRI activation in core language-processing regions (e.g. Broca ’ s area) in response to musical stimuli (X. Chen et al., 2023; Fedorenko et al., 2011).”

“For instance, studies utilizing fMRI have shown synchronous activation patterns between Broca’s and Wernicke’s areas, thus highlighting a functionally interconnected network essential for fluency and comprehension (Tripathi, 2024; Fedorenko et al., 2011).”

“Language processing recruits a domain-specific language-selective network, which includes a set of left-lateralized frontal and temporal regions (Binder et al, 1997; Fedorenko et al, 2010, 2011; Fedorenko and Thompson Schill, 2014; Friederici, 2011; Scott et al, 2017; Vigneau et al, 2006).”

“For example, the bilateral intraparietal cortex (IPS) appears to be sensitive to the processing of numerical (e.g., magnitude) and linguistic content [12] (however, see [13,14] for data against a potential overlap between the language network and mathematics).”

“Asp explains that, for instance, Evelina Fedorenko’s work on linguistic processing in humans (Fedorenko et al. 2011) suggests a domain-specific language network with modality-independent responses, which challenges traditional domain-general cognitive models.”

“It has recently been demonstrated that this language network exhibits selective activation for high-level linguistic processes in adults (Fedorenko et al., 2011) across 12 45 languages and 12 language families (Malik-Moraleda et al., 2022), thus representing a universal neural template for language…”

“These studies have revealed neural territory specific to language 5,73 that closely interacts with other networks involved in cognitive control and theory of mind 4,74,75 .”

“The dATL belongs to the high-level linguistic system in the left perisylvian network, encompassing the inferior frontal gyrus, the lateral temporal cortex, and the inferior parietal cortex (for instance, Fedorenko et al., 2011; Friederici, 2011).”