This study investigates the effect of different types of stimuli on narrative discourse production in healthy aging. 162 adults, aged 20–89 years, produced narratives prompted by two single images (“Picnic” and “Cookie Theft”) and three vignettes (“Nest”, “Quarrel”, and “Flowerpot”). The study addressed three key questions: (1) how aging impacts discourse production, (2) differences in narratives elicited using single versus sequential images, and (3) task-specific effects on linguistic performance. The results showed that aging affects narrative quality more than length. Older adults in their late 70s and 80s produced fewer grammatically complete sentences, more coherence errors, and less informative narratives, particularly with single-image tasks requiring higher cognitive inference. Sequential images elicited longer narratives with fewer local coherence errors, highlighting their scaffolding benefits. The “Cookie Theft” picture elicited more semantic and local coherence errors in older adults, highlighting its potential clinical utility for detecting subtle linguistic difficulties. Overall, task design significantly influenced narrative outcomes, with sequential images supporting coherence and elaboration, while single images challenged inference and lexical retrieval. The clinical implications of these findings are discussed.
Pragmatics has often been viewed as a post-semantic layer that enriches literal meaning through inference. This modular view neglects the cognitive continuity linking syntax, memory, and interpretation. We propose a neurocognitively grounded model of pragmatics within the Neo-UnCartesian Linguistic Program (NULP), where syntactic phases serve as retrieval checkpoints binding linguistic forms to event representations indexed in episodic memory. Core pragmatic phenomena-reference, anaphora, implicature, presupposition, and speech acts-are recast as dynamic operations on a structured event model rather than detached inferential steps. Psycholinguistic, neuroimaging, and clinical evidence supports this account: hippocampal reinstatement tracks pronoun resolution and narrative coherence, while disruptions in hippocampal-prefrontal circuits selectively impair pragmatic integration despite intact syntax. This event-binding architecture clarifies why large language models (LLMs), though compositionally fluent, often hallucinate or lose referential consistency: lacking episodic grounding, they fail in context-sensitive reasoning. Incorporating hippocampus-inspired event indexing and phase-based retrieval could yield AI systems with genuine pragmatic competence and narrative coherence. Philosophically, this framework dissolves the rigid syntax-semantics-pragmatics divide, treating language as a temporal binding mechanism that constructs shared models of reality. NULP thus unites formal theory, cognitive neuroscience, and AI design, offering a comprehensive program for understanding-and engineering-meaning in both biological and artificial minds.
The human brain rapidly extracts meaning from compound words, yet it remains debated whether this occurs through early morphological decomposition or direct access to whole-word representations. While previous research has examined factors such as homophone density, frequency, and semantic transparency, the temporal dynamics of these effects, particularly in the auditory domain, remain unclear. Here, we recorded scalp electroencephalography (EEG) from 40 native Mandarin speakers listening to Chinese disyllabic compound words. Using multivariate pattern analysis (MVPA) and representational similarity analysis (RSA), we traced the temporal evolution of acoustic and linguistic representations during auditory word processing. Results revealed a clear temporal hierarchy: acoustic features emerged first (88 - 190 ms), followed by effects of first-character frequency (144 - 174 ms), homophone density (154 - 215 ms), and semantic transparency (305 - 349 ms). Both constituent-level and whole-word representations were observed, indicating parallel engagement of decompositional and holistic mechanisms. Together, these findings impose critical temporal constraints on dual-route accounts of compound recognition, indicating that decompositional and holistic mechanisms are not simply co-activated in parallel, but are dynamically organized across syllable-locked temporal windows, requiring a temporally structured interpretation of dual-route processing in spoken language comprehension.
We investigated differences in event-related potentials (ERPs) during semantic priming in delayed reading aloud (RA) and lexical decision (LD) tasks. Two different types of analyses were used, a standard time-locked ERP analysis (including item-based regression) and representational similarity analysis (RSA). Across the analyses, the results showed that the RA and LD tasks differed in important ways. With the standard ERP analysis, these included differences in the effect of both word length and word frequency on priming. This suggests task related differences in psycholinguistic variables modulate both the extent and nature of semantic priming. These results pose a challenge for current models of reading. The RSA showed that related primes in the LD task produced greater across-item similarity than related primes in the RA task; a small portion of this effect may have reflected target prediction rather than prime-target integration. This suggests that semantic information was used to a different extent in the RA and LD tasks. Interestingly, results in both the RA and LD tasks based on the pattern of ERPs did not correlate with a semantic model based on simple association scores that the prime-target pairs were chosen on. Overall, the results show that even in reading tasks with relatively similar task demands, there are strong task-related differences that affect how priming occurs and the extent to which different psycholinguistic variables modulate it.
Reading is a complex cognitive process that integrates orthographic, phonological and semantic systems distributed across large-scale brain networks. While prior research has delineated key regions involved in reading, such as the Visual Word Form Area (VWFA), Inferior Frontal Gyrus (IFG), and Angular Gyrus, the extent to which individual variability in intrinsic functional connectivity shapes the architecture of reading networks remains unresearched. This study adopts graph autoencoder (GAE), a topology-preserving deep learning model, to resting-state fMRI data from 69 neurologically typical adults to identify latent neurofunctional subtypes of reading-related brain organisation. Unlike conventional dimensionality reduction or clustering methods that overlook network topology or assume linearity, GAEs preserve the graph structure of functional connectivity, enabling the discovery of non-linear patterns embedded in brain networks. The study extends beyond limited region-of-interest approaches by modelling connectivity across a broader, theoretically informed set of cortical regions implicated in reading. The low-dimensional latent embeddings derived from each functional connectome of each participant revealed three robust and biologically distinct subtypes. These subtypes were characterised by differential engagement of dorsal (phonological) and ventral (semantic) pathways, as well as differences in graph-theoretical metrics such as global efficiency and modularity. Crucially, these subtypes were validated through bootstrapped clustering and exploratory behavioural correlations in a subset of participants, suggesting that individual differences in intrinsic brain connectivity reflect distinct neurocognitive profiles. These findings offer a data-driven framework for understanding individual variability in reading networks, with potential implications for personalized educational approaches pending validation in intervention studies.
Dysarthria, a neuro motor speech disorder, presents significant challenges in developing effective speech command recognizers (SCRs) for people with cerebral palsy (CP). The speech variability in terms of articulatory deficits, speaking rates, and hoarseness, makes accurate recognition difficult. Robust SCRs can boost communication in educational settings and empower people with CP by providing access to smart systems, improving independence and overall quality of life. This study explores the influence of dysarthric severity levels on the recognition accuracy of a wav2vec 2.0-based SCR on 15 people with CP. Analyzing the common characteristics among individuals with the same severity level is a key aspect studied in this work. Additionally, to address the scarcity of large dysarthric speech databases, the usage of multi-task learning (MTL) with severity identification as auxiliary task is proposed. This would further examine the utility of severity information in developing efficient SCRs. It was found that the severity level of dysarthria greatly influences the SCR performance, as articulatory errors tend to increase with severity. Results show that the approach of MTL outperforms the baseline system, with a relative improvement of 8.13%, and achieving an accuracy of 92.24%. The t-SNE analysis revealed phonetically similar words, and these findings can be used to refine the vocabulary used, and to improve the training strategies and rehabilitation practices. The study underlines the importance of severity-aware or severity-adaptive SCR systems and validates the utility of wav2vec 2.0 models for building assistive applications for people with CP.
This systematic review, guided by the PRISMA-P framework, aims to synthesize empirical findings on age-related language decline across lexical, syntactic, semantic, and phonological domains, evaluate neural and cognitive compensatory mechanisms, and assess intervention efficacy. Adhering to PRISMA-P guidelines, we identified 30 eligible studies from 447 initial records. The results indicate that semantic processing exhibits better resilience than syntactic processing, with right-hemisphere activation rising with age, but its compensatory function remains debated. Bilingualism provides protective benefits while also resulting in greater frequency effects. The loss of neuromodulation may contribute to heightened neural noise. The effectiveness of interventions is significantly influenced by cognitive reserve. This study highlights that language decline in aging demonstrates varying susceptibility among linguistic subsystems through a comprehensive assessment of converging and diverging evidence.
For bilinguals, lexical knowledge in the native language (L1) influences word recognition in the second language (L2). When words share both form and meaning across languages, L2 visual word recognition is typically faster, a phenomenon known as the cognate facilitation effect. Such crosslinguistic lexical interaction effects have informed bilingual word recognition models that posit nonselective lexical activation across languages. However, debate remains about the automaticity of this L2-L1 coactivation. The present study examines how the intelligibility of the orthographic input, manipulated through the addition of noise or a difficult font, affects crosslinguistic lexical interactions in Spanish English (L1 L2) bilinguals. The extent to which task difficulty modulates L2-L1 crosslinguistic interactions speaks to different models of bilingual visual word recognition. Experiments 1 shows that stimulus intelligibility attenuates cognate effects. Experiment 2 shows that modulation of cognate effects are partly due to crosslinguistic lexical competitors that is exacerbated by noise. Taken together, Experiments 1 and 2 highlight the importance of simultaneously considering multiple crosslinguistic lexical effects to assess language co-activation. Overall, the findings demonstrate that even under conditions of reduced intelligibility, there is evidence for automatic L2-L1 lexical co-activation.
This special issue contribution highlights new research in the field of bilingualism, with a particular emphasis on translation equivalents, while also reflecting on the key foundational work from the past. Over 30 years ago, the seminal book Cognitive Processing in Bilinguals was published and significantly shaped the study of bilingualism, especially in relation to how bilingual individuals process translation equivalents based on Altarriba's (1992) original paper with that very focus. This paper will discuss the literature that has circulated since 1992 and how this previous research on translation equivalents has provided the foundation for new research in the growing literature on bilingualism. More recent research has continued to build on the findings from 1992 through neuroimaging techniques and technological advancements. The focus of the current paper is to showcase how present research harkens back to the findings and theories presented over thirty years ago. This paper highlights the findings from 1992 and demonstrates that while we have come far, there are still findings that were known by the researchers in Cognitive Processing in Bilinguals back in 1992 that remain both valid and reliable.
This event-related potentials (ERP) study investigated the neural processing of narrative voice, operationalized as grammatical person, in short, externally focalized narratives. Thirty-six participants read three-sentence stories written in the first or third person, in which discourse congruence was manipulated by means of action verbs that were either congruent or incongruent with the preceding discourse context. ERPs were time-locked to the target verbs. Results revealed an effect of semantic congruence on the N400 component (350-550 ms), with incongruent verbs eliciting larger centro-parietal negativities than congruent verbs, indicating increased semantic integration demands. No reliable modulation of the N400 by narrative voice was found. In a later time window (550-700 ms), an exploratory interaction between Narrative voice x Hemisphere was observed over temporal electrodes, although this effect was not supported by significant post hoc comparisons and should therefore be interpreted cautiously. Taken together, these findings suggest that while discourse-level coherence strongly constrains online semantic processing, effects of narrative voice are attenuated when grammatical person is manipulated under conditions of external focalization. These results highlight the importance of focalization and methodological control when investigating neural correlates of narrative voice.
Whether bilinguals’ native and second languages share common conceptual representations has been a focal point of debate in the fields of linguistics and psychology. However, most previous research has been based on traditional language processing tasks, which are heavily influenced by L1 or L2. This study employed the Transformer-based language brain decoding model to decode the brain visual cortex activities of five unbalanced Chinese-English bilinguals separately into Chinese and English sentences while viewing images of five categories (horses, buildings, flowers, fruits, landscapes). The results demonstrated that, firstly, the average decoding scores of all participants in the entire visual cortex for both L1 and L2 significantly exceed the corresponding baseline scores under the three evaluation methods of ROUGE, METEOR, and BCS in natural language processing. Secondly, the decoding accuracy of L2 was higher than that of L1 for all participants across the visual cortex. Lastly, whether decoding L1 or L2, the language decoding accuracy of the higher visual cortex brain signals was higher than that of the lower visual cortex. These findings suggest that the brain signals evoked by natural images in bilingual individuals can simultaneously be decoded into both L1 and L2 languages, yielding results consistent with the shared conceptual representation hypothesis. The factors contributing to decoding differences between L1 and L2, as well as across different brain regions, are also discussed. The findings provide new evidence for bilingual conceptual representation mechanisms, and lay groundwork for the future development of brain-computer interface system and clinical applications of brain language decoding technology.
The effects of bilateral subthalamic nucleus (STN) stimulation on spontaneous language production are poorly understood. Using a pre-post surgery design, semantic and morphosyntactic parameters were evaluated according to the “Analysis of Spontaneous Speech in Aphasia” in seven people with Parkinson's disease (PD). Relative to normative data, the number of nouns and percentage of correct sentences produced were most subjective to changes over time and effects of stimulation. Twelve months after surgery, three patients with predominant left hemispheric dopamine depletion showed a normalization of the number of nouns during STN stimulation, compared to results below the norm in the bilateral OFF condition. Moreover, the effect on the percentage of correct sentences was variable, but the results remained below the norm when pre-operative deficits were present. These outcomes suggest that motor symptom laterality and pre-operative language functioning are among the variables that mediate STN stimulation effects on spontaneous language production, and emphasize the relevance to include the number of nouns and percentage of correct sentences produced as parameters in future larger-scale research.
Examining specific linguistic aspects in isolation, traditional language tests fail to capture the dynamic interactions that support discourse production abilities. The accurate assessment of discourse production is therefore crucial for identifying language difficulties and procedures of discourse analysis have emerged as a valid methodological solution. Despite this, heterogeneity in discourse measures limits comparability across studies, and the lack of normative data across the adult lifespan complicates the differentiation between healthy and pathological aging. The present study addresses this issue by providing the first standardization of linguistic measures extracted using a MultiLevel procedure of discourse Analysis (MLA). Narrative samples from 717 healthy Italian-speaking adults (aged 20 - 94) were elicited through a picture description task using two single images and three vignettes. Speech samples were transcribed and analyzed using a semi-automatic pipeline. Normative data, adjusted for age and education, and data-driven age bands were calculated for linguistic measures assessing productivity, lexical difficulties, grammatical construction, macrolinguistic difficulties, and lexical informativeness. Results provide standardized norms across multiple linguistic measures and reveal distinct age-related shifts in performance. Together, these findings offer the most comprehensive adult lifespan framework to date for narrative discourse production and highlight the importance of data-driven age bands for research and clinical assessment in healthy aging. Furthermore, they show that healthy aging disproportionately affects higher-order integrative discourse mechanisms rather than core lexical and morphosyntactic encoding processes.
The study of distributed cortical circuits in the neural representation of lexical categories of the sense of “olfaction” leads to a greater understanding of the semantic connections of words in perception related to this sense. For this purpose, this research was conducted on Persian vocabulary, especially conceptual metaphors. In this study, the neural representation of olfactory vocabulary was compared between olfactory metaphor and olfactory literal sentences. A database of 48 sentences related to the conceptual metaphors of the sense of smell, 48 literal sentences, and 48 literal sentences of the sense of smell was formed. The processing of these sentences in the brain was investigated using functional magnetic resonance imaging. Twenty-six participants participated, and their brain activity was recorded while they read sentences presented on a monitor (in silence). Statistical comparisons between brain activity during exposure to both metaphorical and literal olfactory sentences showed that secondary olfactory regions were active in both types of sentences. However, literal olfactory sentences showed greater activity. That is, they represent activities in sensory-motor and perceptual representations. The orbitofrontal cortex (secondary olfactory area) integrates information from olfactory, gustatory, and visual modalities, supporting multimodal sensory processing. During the processing of literal olfactory sentences, limited activation was observed in emotion-related regions. These findings can pave the way for a better understanding of the brain mechanisms related to processing Persian words.
The Slow Prediction Hypothesis suggests that predictions based on the thematic roles of sentence arguments unfold relatively slowly and may not influence verb processing at early time points. The current ERP study tested whether such predictions could become available at verb onset when arguments exhibit contrastive animacy features and appear in an order that either aligns with or violates animacy hierarchy. Participants read Mandarin SOV ba-construction sentences in which the arguments were presented in either Animate-first order (e.g., “the dancer ba the ribbon threw out”) or Inanimate-first order (e.g., “the alarm clock ba the roommate woke up”), with verb compatibility manipulated by reversing argument order. We measured N400 responses to the verb, and found that when argument order violated the animacy hierarchy, retrieving the target verb was easier. A verb-type distribution analysis further revealed that Animate-first and Inanimate-first configurations led to predictions of different types of verbs. Together, these findings suggest that animacy-based thematic information can guide both when and what comprehenders predict during sentence processing. The results refine the Slow Prediction Hypothesis by demonstrating that predictions based on thematic role information are not uniformly slow.
The present study combined masked priming with event-related potentials (ERPs) to examine the automaticity and time course of metaphor comprehension in Finnish-English learners. Participants performed a sensicality judgment task on familiar metaphors in Finnish (L1) and English (L2), with target words preceded by masked metaphorical, literal, or control primes. Behaviorally, higher accuracy was observed in L1 than in L2, with no reliable effects of prime type. By contrast, ERP results revealed clear modulation by language and prime. In L1, metaphorical primes elicited enhanced P200 responses relative to literal primes, indicating early perceptualattentional facilitation, while unrelated primes produced the largest N400, reflecting increased semantic integration difficulty. Both metaphorical and literal primes elicited larger late positive components (LPCs) than control primes, suggesting parallel late-stage integration of literal and metaphorical meanings. In L2, both metaphorical and literal primes enhanced early P200 amplitudes relative to control primes, but metaphorical primes elicited larger N400 responses than literal primes, indicating weaker semantic facilitation. Conversely, literal primes elicited larger LPCs, reflecting a bias toward literal semantic integration. These findings suggest that masked priming modulated metaphor processing differently in L1 and L2, with L1 showing smaller P200, larger N400, and smaller LPC amplitudes than L2. These patterns indicate more automatic semantic activation and flexible integration in L1, whereas L2 metaphor comprehension shows a stronger bias toward literal processing. The distinct neurocognitive mechanism between L1 and L2 underscores the importance of L1 background and language experience in shaping the dynamics of metaphor comprehension across both languages.
Event-related potentials (ERPs) offer a powerful method for investigating neural responses to metaphors. However, findings concerning ERP correlates of metaphor processing have been inconsistent across studies. The present meta-analysis aimed to determine whether metaphors elicit different ERP responses compared with literal expressions and to clarify the influence of conventionality, with conventionality rating method assessed as a potential moderator. We synthesized findings from 22 studies (N = 627) examining N400 and/or LPC/P600 ERP responses to metaphorical and literal expressions. The results of meta-analyses revealed 1) more negative N400 and less positive LPC responses to metaphors compared to literal expressions, with no LPC differences between conventional metaphors and literal expressions 2) more negative N400 and less positive LPC responses to novel metaphors compared to conventional ones, 3) no significant influence of the moderator conventionality rating method on differences in N400 and LPC. These findings highlight the need for further investigation, particularly regarding the LPC/P600 window, to assess the role of late sustained negativity in later processing stages of metaphors, and underscore the importance of larger sample sizes.
Pediatric Brain Tumors (PBT) vary in type, grade, and location, and lead to different clinical outcomes and treatments. Standard interventions often include neurosurgery followed by chemotherapy and/or radiotherapy, which improve survival rates and reduce the risk of relapse, but can also lead to neurocognitive impairments. Although there has been increasing interest in the neuropsychological effects of PBT, pragmatic ability (i.e. the way people use verbal and non-verbal communication in a given context) is still under-researched. This study aimed to help fill the gap in identifying and characterizing pragmatic ability in children following PBT. In particular, children's performance has been compared with that of children with Typical Development (TD). The study involved 20 PBT survivors (aged 6-16 years) who were matched by age and sex with 23 TD children. Pragmatic ability was assessed using the Assessment Battery for Communication (ABaCo), which evaluates several pragmatic phenomena in both comprehension and production using various means of expression. The results showed that children after PBT performed significantly worse than the TD group in both pragmatic comprehension and production. In particular, children who underwent both radiotherapy and chemotherapy showed the greatest difficulties. These results emphasize the importance of a comprehensive and multimodal communicative assessment, as provided by the ABaCo, and highlight the need to focus on less commonly studied areas, such as the use of gestures and paralinguistic cues, contextual sensitivity and conversational skills, to gain a better understanding of communication outcomes following PBT.
Employing behavioral experimental methods and event-related potential (ERP) technology, this study explored whether a semantically related P200 amplitude existed when participants analyzed semantic relationships and orthographic structures in Chinese words. Two groups of participants were recruited to perform a semantic relatedness judgment task (Experiment 1) and an orthographic structure judgment task (Experiment 2), respectively. Results of Experiment 1 showed that semantically related word pairs elicited shorter reaction times and larger P200 amplitudes within the 150-300 ms time window compared to unrelated pairs. This evidence suggests that semantic processing of Chinese words commences within the early P200 time window. Results of Experiment 2 showed that the P200 amplitude differences between semantically related and unrelated word pairs were significant in a narrower and later time window of 200-250 ms. This evidence suggests that the brain will automatically process the semantics of Chinese words even if the orthographic task requires no semantic activation. Further conjunction analyses suggested that in the time window of 200-250 ms, the P200 amplitude differences between semantically related and unrelated word pairs were significantly larger in the semantic task than those in the orthographic task. Thus, we speculate that participants’ semantics activation will be weakened by orthographic task demands. Overall, this study provides a new perspective for understanding the automatic processing of Chinese word semantics from an early processing time window of P200 and also semantic/non-semantic task nature.
Using event-related potentials (ERPs), this study investigated how the vertical position of verbal anchoring modulates pictorial metaphor processing through metaphoric congruency effect. A 2 (verbal anchoring position: top vs. bottom) & times; 2 (product involvement level: high vs. low) within-subjects design was employed with 34 participants. Participants completed a verbo-pictorial matching task in which they judged whether the verbally presented metaphor features (shared vs. unrelated) appropriately matched the fusion-structured pictorial metaphors. Shared features were intrinsically positive, whereas unrelated features served as fillers. ERPs were time-locked to the onset of verbal anchoring. The results revealed attenuated frontally distributed N200 and N400 responses, as well as weakened parietal LPC activities, accompanied by enhanced parietally biased P300 amplitudes when verbal anchoring was positioned at the top, suggesting that spatial metaphoric congruency elicits stronger neural engagement for promoting smoother semantic integration and reanalysis. Moreover, an interaction between anchoring position and involvement level was observed. In high-involvement conditions with top-positioned anchoring and low-involvement conditions with bottom-positioned anchoring, larger P300 amplitudes and smaller N400 amplitudes emerged, indicating increased endogenous attention and enhanced semantic integration fluency under involvement-congruent spatial configurations. These results deepen our understanding of the dynamic interaction between pictorial metaphor and verbal anchoring, and offer neural evidence for the design of multimodal promotional discourse that incorporates metaphors.
