Neurobiology of Language

Syntax processing impairment is a defining feature of primary progressive aphasia (PPA), particularly in the nonfluent/agrammatic variant. However, current models are derived largely from Indo-European languages, and it remains unclear whether the cognitive and neuroanatomical mechanisms of syntax processing generalize to Chinese, an analytic, pro-drop language that encodes grammatical relationships primarily through word order and function words rather than inflectional morphology. Participants were recruited from the international Chinese Language Assessment in PPA cohort and included 64 individuals with PPA (19 logopenic, 15 nonfluent/agrammatic, 15 semantic, and 15 unclassifiable/mixed) and 60 cognitively healthy controls. Participants completed sentence comprehension and production tasks with systematically manipulated syntactic features. Item-level linguistic determinants were examined using ridge regression, diagnostic performance using receiver operating characteristic analyses, and neuroanatomical correlates using voxel-based morphometry of structural MRI. Performance on sentence comprehension and production was influenced by grammatical and lexical features. Subject-relative clauses, passive constructions, a higher proportion of function words, longer character length, and greater lexical content were associated with poorer performance among participants with PPA. Sentence comprehension was impaired in the logopenic, nonfluent/agrammatic, and semantic variants, even when isolated to the syntactic component of the sentence comprehension task. Similarly, sentence production was impaired across all PPA variants (all P < 0.01 versus controls). No significant differences were observed among PPA variants. Adjusted receiver operating characteristic analyses demonstrated high discrimination between each PPA variant and controls, with areas under the curve ranging from 0.87 to 0.97 for sentence comprehension-syntax and from 0.91 to 0.97 for sentence production. However, pairwise comparisons showed no significant differences in discriminative performance among PPA variants after correction for multiple comparisons. Both sentence comprehension-syntax and sentence production showed associations with semantic and repetition measures in addition to motor speech measures. Voxel-based morphometry showed that higher sentence comprehension-syntax scores were associated with greater grey matter volume in left frontotemporal and cingulate regions, whereas higher sentence production scores were additionally associated with greater gray matter volume in frontostriatal structures, including the caudate, putamen, and thalamus. Syntactic processing impairment in Chinese-speaking individuals with PPA extends beyond the nonfluent/agrammatic variant and is associated with distributed frontotemporal and frontostriatal regions, as well as lexical-semantic and auditory verbal short term memory mechanisms. These findings suggest that the cognitive and neural manifestations of syntactic impairment may be shaped by language-specific typological properties, underscoring the need for linguistically informed approaches to the assessment and interpretation of PPA.

Language can be experienced in highly serial form, as in speech, or in fully parallel form, as in text, yet how our brains adapt to differences in seriality is not understood. Here, we held modality constant (visual) and examined whether processing mechanisms underlying grammaticality detection are similar under Rapid Serial Visual Presentation (RSVP) and Rapid Parallel Visual Presentation (RPVP). To identify shared neural representations of grammaticality, we trained a classifier on magnetoencephalography (MEG) responses to grammatical and ungrammatical three-word phrases in one presentation mode and tested its generalization to the other. Behaviorally, accuracy was higher and reaction times faster for grammatical vs. ungrammatical phrases across presentation modes. Neurally, cross-presentation decoding revealed shared representations of grammaticality at word 2 across train-test directions--albeit with different latencies and durations--and an early shared representation at word 3 when trained on RPVP. We take the bidirectional representations at word 2 to capture distinct but presentation-invariant combinatorial processes and the asymmetry at word 3 as reflecting stronger global structure encoding in RPVP.

Developmental dyslexia is associated with impaired cortical tracking of continuous speech and difficulties in perceiving within-word prosodic structure in children and adults. For adults and continuous speech, both low-frequency delta- and theta-band impairments and high-frequency gamma- and beta-band differences are found. To date, perceptual difficulties with intra-word prosody have not been studied in direct relation to neural speech encoding. Here neurotypical adults and adults with dyslexia listened to multisyllabic words. EEG was recorded, and cortical encoding was quantified using temporal response function (TRF) modelling. Four speech representations were explored: lexical stress pattern (intra-word prosodic prominence), the low-frequency speech envelope, word onsets and syllable onsets. Adults with dyslexia showed significantly reduced encoding of intra-word prosodic prominence (delta band) and of the speech envelope (delta and theta bands). In contrast, encoding of word and syllable onsets was preserved. Pinpointing delta-band cortical tracking as central to encoding intra-word prosodic prominence opens new avenues for research.

Humans routinely encounter speech in languages they have never heard, yet how the brain responds to such input and whether bilingual experience shapes this response remains unknown. Here, we used electroencephalography (EEG) and temporal response function (TRF) modeling to examine cortical tracking of the speech envelope in 24 English-monolingual and 24 English-Mandarin bilingual adults. Participants listened to naturally produced continuous speech in three languages: English (familiar to all), Mandarin (familiar to bilinguals only), and Vietnamese (unfamiliar to all). We report two main findings. First, both monolinguals and bilinguals showed enhanced cortical tracking for unfamiliar relative to familiar languages, evidenced by higher EEG prediction accuracy (PA). Monolinguals showed enhanced tracking for both Mandarin and Vietnamese, whereas bilinguals showed enhancement only for Vietnamese, consistent with Mandarin being a familiar language for this group. This finding suggests that enhanced cortical encoding of unfamiliar speech is a general property of the listening brain, not a signature of listening to a non-native language or reduced language proficiency. Second, bilinguals strikingly showed stronger cortical tracking than monolinguals overall, in both PA and TRF peak weights, with the TRF peak weight advantage present across all three languages, suggesting a difference in how bilingual experience shapes the neural encoding of speech. These findings have implications for understanding how the brain navigates the linguistic diversity of everyday life in an increasingly global, multilingual world.

Comprehending spoken language requires extracting syntactic structure and meaning from a continuous acoustic signal. Psycholinguistic evidence points to a key role for prosody in this process. However, because interactions between prosody and syntax are typically demonstrated using carefully controlled sentence constructions, it remains unclear whether such interactions constitute a general, default property of spoken language processing. Addressing this question requires naturalistic, broad-coverage data. Here, we combined experimental psycholinguistics with naturalistic neuroscience to investigate the joint neural encoding of syntax and prosody in MEG recordings of 26 participants (19 female) listening to narrative speech. Linguistic annotations revealed that strong prosodic boundaries tend to coincide with syntactic phrase closures, and that the strength of prosodic boundaries scales with the number of phrases closed. Using time-resolved regression with temporal response functions, we modeled how these features relate to delta-band neural activity during continuous speech processing. Our results reveal a bidirectional interaction between syntax and prosody. Prosodic marking modulates the neural processing of syntax, with phrase closures being encoded more strongly when accompanied by strong prosodic boundaries. Conversely, syntactic structure shapes prosodic encoding, such that boundary strength is neurally encoded primarily at points of syntactic phrase closure. These findings demonstrate that syntax and prosody are encoded not independently, but in a context-dependent and mutually constraining manner, highlighting their joint role in guiding structure building during naturalistic spoken language comprehension.

Chinese is one of the living languages with the highest homophone densities, making homophone processing a critical component of language comprehension and orthographic decoding, as well as an ideal model for examining its cognitive mechanisms. This study assessed mono- and bi-character homophone selection function in 129 Mandarin- and Cantonese-speaking participants, including 66 cognitively healthy individuals, 17 with semantic variant (sv) primary progressive aphasia (PPA), 23 with the nonfluent/agrammatic variant (nfv) PPA, and 23 with the logopenic variant (lv) PPA. All PPA variants showed impairment in homophone selection, with most significant deficits in svPPA and lvPPA. ROC analyses demonstrated its strong diagnostic utility for distinguishing PPA from controls. Voxel-based morphometry revealed that homophone performance correlated with gray-matter volume in left anterior and inferior temporal regions, with variant-specific patterns observed. These findings described the potential role of homophone selection in diagnosing PPA syndrome in Chinese speakers, illuminate crosslinguistic influences on language impairment in PPA, and provide novel insights into the cognitive processing of homophones.

Free recall of naturalistic stimuli, such as films and stories, reveals what people remember and how they organize it. Scoring such recall requires matching each recalled utterance to the encoded stimulus, and doing this manually often makes large studies impractical. Existing automated methods mostly return similarity or aggregate scores rather than explicit matches between recalled and stimulus content. Here we introduce an open-source, model-agnostic pipeline that uses a large language model (LLM) to divide the stimulus and the recall into information units and to match units between the two texts. The output is a table of unit-to-unit matches with accuracy labels, from which standard analysis scripts compute recall measures. We validated the pipeline with four LLMs (Claude Sonnet 4.5, GPT 5.4, Gemini 2.5 Pro, and Gemini 2.5 Flash) on three datasets. On typed recall of four short stories, agreement between each model and two trained raters ({phi} = .73-.79) was close to the agreement between the raters themselves ({phi} = .77). A second run reproduced each model's matches ({phi} = .91-.97) more closely than the raters agreed with each other. On spoken recall of 10 films, every model attributed recall to the correct film with high agreement. On uncorrected speech-to-text recall of a television episode, agreement varied more across models. The median cost was under $1 per participant for every model. These results suggest that the pipeline can serve as a reliable, low-cost alternative to manual scoring of free recall.

Language processing does not occur in a vacuum. Language comprehension and production have its own temporal organization, and it unfolds against ongoing neural. We examined whether the base organization persists from rest into sentence processing, and we further examined how it relates to evoked responses and semantic decisions. Sentences crossed linguistic argument versus adjunct completions with semantic plausibility. Anomalies elicited a centroparietal late positivity from 600 to 900 ms (+0.628 microvolt, 95% CI 0.349-0.907). Channel-wise autocorrelation timescales formed person-specific topographies: own-person rest-task similarity exceeded mismatched-person similarity (Fisher-z advantage = 0.463, permutation P = 5.00 x 10^-6), identifying 78.6% of participants against 7.1% chance. The advantage survived Cz re-referencing and surface-Laplacian transformation. In separate ERP predictor tests, longer absolute task timescales accompanied a smaller anomaly effect (cross-fitted Spearman rho = -0.486, Holm-adjusted P = 0.030). Stronger topographic continuity also accompanied greater anomaly-related decision slowing, although the behavioral association fell short of correction across six endpoints (rho = 0.547, familywise P = 0.054). Supporting analyses evaluated larger eligible samples and an eyes-closed subset. The findings place persistent, person-specific temporal organization within sentence processing and link absolute task timescale to the neural response to semantic anomaly.

Reading behaviour is sensitive to contextual demands. For example, when reading to study or answer questions, readers strategically allocate more attention to goal-relevant sections, read slower, and regress more frequently to extract specific information based on task demands. This suggests that readers flexibly adapt their strategies according to goals and situational constraints. However, the neurocognitive mechanisms supporting this flexibility remain unclear. To test this, we combined eye-tracking with EEG during naturalistic reading and examined how short-term reading goals modulated the neural tracking of hierarchical linguistic features at the orthographic (orthographic neighbourhood size), lexical (word frequency), and contextual-semantic (Surprisal) levels. Forty native English speakers read narratives in three goal conditions: Baseline (no goal), Proofreading (detect spelling errors), and Comprehension (attend to content). Fixation-aligned word-level EEG activity was modelled using multivariate temporal response function to estimate the strength of feature-specific neural tracking. Orthographic tracking was significantly enhanced during Proofreading relative to Baseline, and predicted better Proofreading accuracy. Conversely, stronger frequency tracking predicted better Comprehension accuracy. Contextual-semantic tracking was not modulated by task goal. These findings suggest that reading goals selectively bias neural processing towards linguistic features (sublexical vs. lexical) that support successful task performance, providing evidence for short-term functional plasticity in linguistic processing and supporting a flexible predictive processing account of goal-directed reading.

Visual orthographic knowledge plays a critical role in early literacy development and may provide a shared foundation for reading and spelling, yet the nature of its underlying representations remains debated. Behavioral measures offer indirect evidence and may be influenced by other metalinguistic processes. It remains unclear whether beginning readers encode larger orthographic structures, such as onset-rime units, beyond individual letters and smaller letter combinations but below the whole-word level. It is also unknown whether neural sensitivity to these structures helps explain the relationship between spelling and reading. To address these questions, we used a fast periodic visual stimulation (FPVS) paradigm in first- and second-grade children to contrast pseudowords constructed with the typical ordering of onset followed by rime elements (i.e., onset-rime) against matched strings in which the same elements appeared in the reverse order (i.e., rime-onset). Neural sensitivity to onset-rime orthographic structure was associated with both spelling ability and word-reading efficiency and, critically, statistically mediated the relationship between these two literacy skills. These relationships remained significant after controlling for age and removing influential observations, and sensitivity to onset-rime orthographic structure explained additional variance in word-reading efficiency beyond spelling ability and age. Together, these findings provide neurophysiological evidence for onset-rime structure as an important sublexical orthographic unit in beginning readers, highlighting its potential role in linking spelling with efficient word recognition during early reading development.

Communication is inherently interactive, yet most neuroimaging studies examine individuals in isolation. Here, we used functional near-infrared spectroscopy (fNIRS) hyperscanning to investigate interbrain synchrony (IBS) during naturalistic communication. In Study 1, 13 healthy familiar dyads completed Joint Singing, Joint Reading, and Picture Guessing tasks. IBS was quantified using wavelet transform coherence across bilateral frontal and temporoparietal regions. Joint Singing produced higher IBS than the other tasks in several regions, including the dorsolateral prefrontal cortex and temporoparietal junction, and selected effects were greater in true than pseudo-dyads. In Study 2, we applied the same paradigm to five individuals with chronic post-stroke aphasia interacting with trained speech-language pathologists. IBS could be estimated in all dyads after lesion-informed channel exclusion, although regional coverage varied with lesion location. These findings demonstrate that fNIRS hyperscanning can capture task-dependent interpersonal neural coupling during spoken interaction and can be extended to communication involving individuals with aphasia. This approach provides a foundation for studying how neural coordination between communication partners relates to functional communication and partner-supported interaction.

Effective verbal communication requires individuals to align their communicative intentions through the precise use of utterances, yet how this is achieved in dynamic, open-ended communication remains unclear. Combining predictive deep language models (DLMs) with cognitive computational modeling, we tested a computational division of labor between context-sensitive global constraint on the utterance space and partner-sensitive local adjustment of selection priorities. We formalized local adjustment using an active inference inspired updating rule within a statistical-pragmatic hybrid model. Thirty dyads performed a dynamic, open-ended collaborative word-generation task during functional near infrared spectroscopy (fNIRS) hyperscanning. Interlocutors converged in their lexical choices, and lexical alignment was best explained by a hybrid model integrating contextual selection priors with interaction-specific local adjustment among the models tested. This hybrid process formed a dynamic hypothesis landscape comprising a global semantic search region and local selection priorities. Critically, interaction did not simply reduce uncertainty: within individuals, the hypothesis space of global semantic search progressively narrowed while local uncertainty among plausible alternatives increased; between partners, both global hypothesis space and local selection became increasingly aligned, with alignment of local selection priorities most strongly predicting subsequent lexical alignment. The contribution of local adjustment was stronger during precise verbal communication than in the control tasks. Neural analyses further identified correlates of the local adjustment process: temporoparietal and anterior temporal regions supported partner-related inference and semantic prediction error, respectively, and their activity during thinking predicted subsequent inferior frontal activity associated with self-updating during speaking. The participant's own hypothesis landscape was also represented in the left inferior parietal-supramarginal cortex. The main findings generalized across language models and an independent Czech dataset. Together, these findings reveal that alignment in open-ended communication emerges from complementary statistical and pragmatic computations, providing a framework for understanding human and human-AI communication.

Inhibitory control deficits have long been proposed as a central neurocognitive mechanism underlying childhood Attention Deficit/Hyperactivity Disorder (ADHD). However, neuroimaging findings remain heterogeneous, partly due to methodological variability across developmental samples and experimental paradigms. The present study aimed to characterize the neural architecture of motor response inhibition in typically developing (TD) children and identify reproducible functional abnormalities associated with pediatric ADHD using stringent, developmentally specific inclusion criteria. We conducted two coordinate-based meta-analyses using Activation Likelihood Estimation (ALE). Following a systematic literature search (up to August 2026), we identified 23 task-based fMRI studies reporting inhibitory-control-related activation in TD children (237 foci) and 17 studies comparing children with ADHD and TD peers (114 foci) during the Stop-Signal Task or Go/No-Go Task. All participants were younger than 18 years of age. Statistical significance was determined using cluster-level family-wise error (FWE) correction. In TD children, ALE revealed a robust inhibitory-control network encompassing the bilateral anterior insula (AI), inferior frontal gyrus (IFG), and the midline pre-supplementary motor area (preSMA) and anterior cingulate cortex (ACC). This spatial organization closely resembled the canonical inhibitory-control network consistently observed in adults, suggesting that the core neural architecture supporting motor inhibition is already detectably organized during development. In the between-group analysis, children with ADHD exhibited significant hypoactivation primarily within the preSMA. Notably, this region demonstrated striking spatial overlap with the preSMA cluster reliably recruited in TD children during motor inhibition. These findings provide convergent support for inhibitory-control models of ADHD and identify altered preSMA recruitment as one of the most consistent functional abnormalities associated with pediatric ADHD. While the neural infrastructure supporting inhibitory control is broadly established during childhood, children with ADHD exhibit reduced recruitment of a key cortical node within this network during motor inhibition demands.

ObjectiveRecovery from post-stroke aphasia is traditionally thought to plateau within a year after stroke. As such, no studies have characterized long-term recovery using comprehensive behavioral assessment and longitudinal functional MRI. We sought to characterize behavioral and neuroplastic changes over nearly six years in a patient with chronic post-stroke aphasia. MethodsA patient with chronic left hemisphere stroke underwent behavioral assessment and task-based fMRI at 9, 22, 33 and 71 months after stroke. Testing assessed auditory comprehension, semantic and phonological processing, naming, reading, spelling, motor speech, and executive function. Task-related activation during an adaptive semantic matching paradigm was quantified within language and semantic regions of interest. ResultsBehavior improved across multiple language domains. WAB-R Aphasia Quotient increased from 49.2 to 69.2, along with the large gains in word finding, connected speech, reading and spelling, whereas auditory comprehension remained relatively stable. fMRI demonstrated progressively increasing activation within left-hemisphere language regions, reaching normative range by the final assessment, along with increasing activation in semantic/executive regions. DiscussionRecovery of language production abilities can continue for years after stroke, supported by neuroplastic changes in the brain. Neurologists should inform stroke survivors that aphasia recovery can continue indefinitely after stroke.

In aphasia, variability in lesion location and post-stroke network reorganization complicates efforts to identify the regions that support semantic versus phonological processing. In prior work, researchers have often examined these domains in separate patient cohorts, further limiting direct comparison of their neural substrates within aphasia. To address this gap, we investigated the degree of overlap and specialization within semantic and phonological networks in adults with and without aphasia and examined how neural activity relates to task performance and lesion profiles. Sixteen neurologically healthy adults and 15 participants with aphasia following left hemisphere stroke completed adaptive Semantic Matching and Rhyme Judgment tasks during functional near-infrared spectroscopy (fNIRS) recording. Changes in oxyhemoglobin and deoxyhemoglobin were measured within regions of interest, and activation was related to standardized task performance and damage to left ventral and dorsal language pathways in participants with aphasia. Controls showed largely distinct activation patterns consistent with prior fMRI findings: Semantic Matching preferentially recruited ventral stream regions, particularly left temporal cortex, whereas Rhyme Judgment more strongly recruited dorsal regions, including left inferior frontal and inferior parietal cortex. Participants with aphasia showed greater inter-individual variability and no clear group-level segregation of dorsal and ventral activity. Better Semantic Matching performance was associated with less left ventral stream damage and greater left dorsal activation, whereas better Rhyme Judgment performance was associated with lower right hemisphere ventral and dorsal activation. These findings suggest that post-stroke language performance reflects residual specialization, lesion-dependent constraints, and flexible recruitment of surviving tissue. Overall, fNIRS shows promise as an alternative to fMRI for language mapping in post-stroke aphasia, although replication in larger samples with more extensive cortical coverage is warranted.

Neuroimaging studies identify the superior temporal sulcus (STS) as an anatomical and functional target in autism spectrum disorder (ASD). Yet, the STS is a heterogeneous structure containing unique subregions, such as the superior temporal asymmetrical pit (STAP), which is a recently identified human-specific landmark. To date, no study has compared the STAP between ASD individuals and neurotypical controls (NTs) in a large sample. Here, we analyzed the mean intra-hemispheric sulcal depth and inter-hemispheric laterality of the STAP, as well as the anterior and posterior portions of the STS (aSTS, pSTS), in 100 NTs and 100 ASD male individuals (ages 5-18) using a discovery-replication design. Across samples, the STAP and pSTS showed rightward asymmetry, while the aSTS showed leftward asymmetry within groups. All features of the STAP, aSTS, or pSTS did not differ significantly between groups. However, pSTS laterality related to individual differences in social cognition, which was further supported by a meta-analysis across over 200 studies showing that neurocognitive functional differences associated with social cognition and ASD are restricted to portions of the STS outside of the STAP. Together, these findings provide fine-grained insight into prior studies showing coarse morphological differences of the STS between individuals with ASD and NTs, as well as open up new questions and new neuroanatomical targets of the heterogeneous STS in future studies of brain structure, brain function, and cognition in ASD, as well as other neurodevelopmental disorders targeting the superior temporal cortex.

Naming pictures from the same semantic category can interfere with subsequent word retrieval. Although the left inferior frontal gyrus (LIFG) has been implicated in resolving this semantic interference, its causal role in this process remains unclear. Across three experiments, we investigated the contribution of the LIFG to interference resolution during blocked-cyclic naming. Experiment 1 tested whether continuous theta-burst stimulation over the LIFG increased semantic interference relative to vertex stimulation. Although the overall interference effect did not differ significantly between stimulation sites, exploratory analyses revealed an unexpected pattern based on the initial context (related vs. unrelated) with which items were named, suggesting that LIFG disruption affected retrieval of previously related items even when they were subsequently named in unrelated contexts. This observation motivated two further experiments examining the consequences of prior naming context more directly. Experiment 2 established that, without stimulation, interference observed in related blocks was no longer evident after the items were redistributed into unrelated blocks. Experiment 3 combined this design with cathodal or sham transcranial direct current stimulation (tDCS) over the LIFG. Planned comparisons showed that cathodal stimulation slowed naming of initially related items both within related blocks and after redistribution into unrelated blocks; corresponding comparisons for items named consistently in unrelated blocks were not significant. The similar patterns across two stimulation methods suggest that LIFG-mediated control may influence both selection under current competition and subsequent retrieval of previously competing words. These findings could be accommodated within a hybrid account in which working memory supports context-guided selection and selection-related control helps recalibrate lexical accessibility, limiting the carryover of interference into subsequent naming.

Comprehending a narrative requires continuously updating a model of the discourse, and beta-band activity is thought to signal when that model should be held versus changed. Whether beta also grades how much to update, and whether that index is carried by discrete bursts beyond sustained power, is unknown for naturalistic comprehension. We recorded MEG from 61 human participants (22 men, 39 women) while they listened to naturalistic Hebrew narratives, detected beta bursts in the continuous signal, and quantified the semantic shift between successive sentences from sentence embeddings. Beta burst rate decreased after sentence boundaries, and regression-based deconvolution separated a pre-onset maintenance phase from a graded post-onset response whose magnitude scaled with semantic shift (peak ~0.9 s). This response was independent of lexical surprisal, the acoustic envelope, and referent introduction, largely robust to dependency structure, though it shared variance with propositional quantity. Burst occurrence carried a component of the effect beyond the accompanying amplitude change: applying that change as a pure gain to each participant's envelope and re-running the identical detection predicts only 75% of the observed burst-rate reduction (excess p = .038), the effect survived an amplitude-invariant threshold, and a Poisson model reproduced it as a graded reduction in burst-initiation rate. The effect held in both participant groups, which heard differently framed narratives, and was decisively supported (BF10 > 250). Beta bursts thus provide a graded signal of semantic updating during continuous comprehension, expressed in the rate of burst occurrence over and above the accompanying change in sustained power.

Resting-state connectivity can predict task-evoked fMRI activation, but correspondence with an individual task map may partly reflect a shared population pattern. We evaluated the Variational Resting-state-to-Task Prediction TransformeR (VRPTR), a three-dimensional encoder-decoder combining a compressed Transformer bottleneck, variational latent sampling, and multiscale skip connections. Models were trained in 360 healthy adults from the WU-Minn Human Connectome Project and evaluated in 40 held-out participants from the same dataset for the story-versus-math language contrast. VRPTR achieved mean voxel-map Pearson r=0.642 and Dice AUC=0.519, exceeding compact volumetric BrainSurfCNN-like and SWIFUN-like comparators by {Delta}r=0.0376 and 0.0335, respectively. A template fixed from the 360 training maps yielded residual correlation 0.364; VRPTR improved correlation but not mean absolute or root mean squared error over this template. A matched factorial ablation associated Transformer processing with improved map correlation ({Delta}r=0.0168); the variational main effect did not survive multiple-comparison correction. Component swaps indicated that subject-specific activation patterns depended more on multiscale skip features than on the bottleneck representation. Latent sampling produced a modest association between posterior standard deviation and voxelwise absolute error (mean Spearman {rho}=0.091). Raw 95% intervals covered only 4.7% of observed values. Five-fold calibration within the held-out cohort increased coverage to 94.8%, and spatially varying uncertainty modestly improved probabilistic scores over a constant-width model. Uncertainty also increased under input degradation. These findings support the contribution of Transformer processing to language-map prediction and distinguish it from subject-specific spatial conditioning and relative uncertainty estimation. Independent validation of calibrated uncertainty and evaluation in clinical cohorts remain necessary.

Lesion Network Mapping (LNM) is a method used to map symptom-related circuits in the human brain. Recent work has raised concerns about the method and the specificity of LNM-derived networks, which often show high spatial overlap across unrelated conditions. Null models have been proposed to address the raised limitations1. Useful statistical tools in network neuroscience, they are reference procedures; in LNM, the normative connectome matrix C remains the source of the derived networks also under null modelling. Their broader role as a general solution for LNM therefore warrants careful evaluation. Here, we examine the proposed null models step-by-step and test them empirically across multiple datasets. We show that the proposed null-model correction effectively only replaces C with a transformed matrix C' (e.g., C' ~ C - mean(C)), one that is directly derived from C and retains much of its low-dimensional structure. Consequently, null-model-corrected LNM maps remain shaped by the same standard set of connectome-derived patterns and continue to show repetition and non-specificity across unrelated conditions. We discuss the scope and limitations of null models for LNM, showing that they can statistically benchmark LNMs, but that their application does not resolve the core limitation of non-specificity of LNM networks.