Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
People in depression tend to retrieve overly general memories instead of specific events (Overgeneral Memory, Williams et al., 2007). The most commonly used measures in overgeneral memory (OGM) research, the Autobiographical Memory Test (AMT), consists of five positive and five negative cue words. OGM studies often fail to find differences in specificity between positive and negative cues and this could partly result from a methodological issue. An examination of 61 studies using AMT revealed that more than 85 cues were chosen without using a standardized selection method. We explored the characteristics of these AMT cues used in OGM research. Forty university students (20 female) rated the cues for valence, arousal and dominance and filled out depression measures. We found that valence, arousal and dominance ratings were modulated by participants’ depression level. Our results suggest a strong need to have a standardized selection method for words used in OGM research.
This article analyzes the transformation of the Uzbek language in the digital environment acrossthree main areas. Digital technologies are driving rapid changes in the language: lexical units borrowed fromEnglish, graphic abbreviations, symbols expressing emotions, and hybrid forms have become actively used onthe internet and social media. At the same time, natural language processing (NLP) systems working with theUzbek language, such as text analysis, machine translation, and speech recognition models, are insufficientlyequipped with corpus and linguistic databases
Abstract This article empirically investigates the semantic meaning of ‘wh-exclamatives’ (e.g., What a view!). Previous literature has identified three distinct meaning components of wh-exclamatives: ±mirativity (speaker’s surprise), ±noteworthiness (degree of a referent’s characteristic), and ±valence (speaker’s strong positive or negative evaluation of the referent). In Study 1, participants (n = 50) rated the felicity of wh-exclamatives in contexts manipulated by these three meaning components. The findings indicate that ±noteworthiness and ±valence, but not ±mirativity, influence the felicitous use of wh-exclamatives. Study 2 (n = 80) examines the role of wh-exclamatives as linguistic valence setters, investigating how the construction interacts with the inherent valence of discourse referents (valence of a referent in isolation) to determine the overall valence of embedded referents in a valence rating study. The results reveal that wh-exclamatives have a polarity-conforming valence-strengthening function, enhancing the positivity of positive referents and the negativity of negative ones. Neutrally valenced referents are preferentially positivized, indicating a constructional positivity bias.
The rise of Artificial Intelligence (AI)-based tools is transforming language education, offering adaptive innovations for both language learners and instructors. However, concerns remain about their ability to represent linguistic diversity. Shaped by the data they process and the priorities of their creators, AI systems risk reinforcing dominant linguistic norms. This study explores these issues using Austrian Standard German (ASG), a distinct variety of German, as a case study. By analysing the behaviour of four AI tools—two models of a chatbot, a text-feedback system, and a grammar-correction tool—we assessed whether they recognised ASG as a legitimate standard or altered it to align with German Standard German (GSG). The evaluation, informed by structured testing, revealed significant shortcomings in how these systems handle linguistic variation. Our findings underscore the risks of erasing linguistic particularities and emphasise the need for AI tools to serve as a means of fostering linguistic identity—particularly in educational contexts.
This paper presents new resources and baselines for Dependency Parsing in Pomak, an endangered Eastern South Slavic language with substantial dialectal variation and no widely adopted standard.We focus on the variety spoken in Turkey (Uzunköprü) and ask how well a dependency parser trained on the existing Pomak Universal Dependencies treebank, which was built primarily from the variety that is spoken in Greece, transfers across dialects.We run two experimental phases.First, we train a parser on the Greek-variety UD data and evaluate zero-shot transfer to Turkish-variety Pomak, quantifying the impact of phonological and morphosyntactic differences.Second, we introduce a new manually annotated Turkish-variety Pomak corpus of 650 sentences and show that, despite its small size, targeted fine-tuning substantially improves accuracy; performance is further boosted by cross-variety transfer learning that combines the two dialects.
Replication package for "CEI: A Benchmark for Evaluating Pragmatic Reasoning in Language Models" (DMLR 2026). The Contextual Emotional Inference (CEI) Benchmark is a dataset of 300 expert-authored scenarios for evaluating how well language models interpret pragmatically complex utterances in social contexts. Each scenario presents a communicative exchange involving indirect speech (sarcasm, mixed signals, strategic politeness, passive aggression, or deflection) where the speaker's literal words diverge from their actual emotional state. Three trained annotators independently labeled every scenario using Plutchik's 8 basic emotions and Valence-Arousal-Dominance ratings. This archive contains: • data/human-gold/ — 5 merged annotation CSVs (300 scenarios, 3 annotators each) • scripts/ — Pipeline, analysis, and HuggingFace upload scripts • config/ — Model definitions and pricing configuration • papers/dmlr2026/ — Paper source (LaTeX), bibliography, and figures • reports/dmlr2026/ — Baseline results (JSON) and LaTeX tables • LICENSE (MIT for code) and README.md The dataset is released under CC-BY-4.0. Code is released under MIT. GitHub: https://github.com/jon-chun/cei-tom-dataset-base HuggingFace: https://huggingface.co/datasets/jonc/cei-benchmark
To foster investigations of noun phrase (NP) inflection in German at scale, this paper introduces DiNoS (Distributional Noun Structure), a data-driven lexicon of NP heads, which includes statistical information on the dependents and the morphosyntactic features of their original in-context appearances.We make available the source code for the extraction of NPs from CoNLL-U treebanks, which includes rule-based heuristics to improve feature annotation coverage and ensures a homogeneous lemmatisation strategy across treebanks.While the resulting JSON-based lexicon is suitable for no-code interaction for non-experts, it is further supported by a toolkit for the automatic calculation of, and access to, various statistical overviews.In this paper, we present the heuristics employed to extract NP datasets from the German Universal Dependencies' Hamburg Dependency and GSD treebanks.In addition, we provide a preview of the emerging DiNoS lexica's properties and discuss some implications of noun and determiner word form ambiguity for NP complexity.
In a multilingual and multicultural society like Malaysia, the spelling of names serves as a personal identifier and as a reflection of sociocultural and linguistic norms. While prior studies have examined spelling variation in educational and digital contexts, less is known about public perceptions of spelling variations of names. Against this backdrop, this study investigates how Malaysians perceive the acceptability of spelling variations. Using a mixed-methods design, 355 participants responded to a questionnaire comprising Likert-scale evaluations of ten real names, along with open-ended questions. Quantitative analysis revealed significant variability in participants’ acceptability ratings, though names that retained phonological clarity (e.g., Adrianah) were generally more accepted than others (e.g., Frrdy). No significant differences were found across gender, age, ethnicity, or professional background. Thematic analysis of qualitative responses highlighted two key influences on naming decisions: individual sociocultural factors (e.g., religious beliefs, family tradition) and sociolinguistic-aesthetic factors (e.g., pronunciation, trends, media influence). These findings suggest a growing tolerance towards spelling variations, possibly indicative the value of distinguishing one’s identity through names.
Large language models (LLMs) perform strongly on many NLP tasks, but their ability to produce explicit linguistic structure remains unclear. We evaluate instruction-tuned LLMs on two structured prediction tasks for Standard Arabic: morphosyntactic tagging and labeled dependency parsing. Arabic provides a challenging testbed due to its rich morphology and orthographic ambiguity, which create strong morphology-syntax interactions. We compare zero-shot prompting with retrieval-based in-context learning (ICL) using examples from Arabic treebanks. Results show that prompt design and demonstration selection strongly affect performance: proprietary models approach supervised baselines for feature-level tagging and become competitive with specialized dependency parsers. In raw-text settings, tokenization remains challenging, though retrieval-based ICL improves both parsing and tokenization. Our analysis highlights which aspects of Arabic morphosyntax and syntax LLMs capture reliably and which remain difficult.
Abstract Part-of-Speech (POS) tagging is a core task in natural language processing (NLP) and a crucial building block for higher-level applications such as parsing and machine translation. For low-resource and morphologically rich languages such as Kangri, POS tagging remains challenging due to scarce annotated corpora and limited linguistic resources. This paper presents a comparative study of three POS tagging approaches for Kangri: a feature-based Conditional Random Field (CRF), an untuned Bidirectional Long Short-Term Memory (BiLSTM) baseline, and a hyperparameter-tuned BiLSTM. All models are trained and evaluated on the Universal Dependencies (UD) Kangri Treebank. The tuned CRF achieves strong test-set performance (70.4\% accuracy and weighted F1 0.695), the untuned BiLSTM provides a robust neural baseline (66.0\% accuracy), and a hyperparameter-optimized BiLSTM reaches higher validation accuracy during tuning. We analyze per-tag strengths and weaknesses, training dynamics, and provide recommendations for future improvements in low-resource POS tagging.
Misconceptions in physics are persistent explanatory frameworks that develop in early infancy and continue throughout primary, secondary, and university education, even among pre-service educators. Instead of being eradicated through formal education, these intuitive notions often coexist with scientific principles, resulting in a disjointed or contextually dependent comprehension. This review consolidates recent studies on the origin, endurance, and alteration of misunderstandings across several educational stages and physics disciplines, encompassing mechanics, thermal phenomena, energy, and electromagnetism. It emphasizes that misconceptions are influenced by perceptual experiences, linguistic norms, and evaluative methods that favour procedural fluency over explanatory reasoning. The paper analyzes instructional methods that facilitate conceptual change, highlighting hands-on and virtual experimentation, guided inquiry, diagnostic evaluation, and innovative digital and AI-driven technologies. The research identifies cross-cutting themes that emphasize the significance of cognitive conflict, explicit interaction with learners' concepts, and coherent conceptual advancement in curriculum. Significant study deficiencies encompass the necessity for longitudinal studies, examinations of transfer and durability, and a more profound inquiry into teacher education as a domain for conceptual advancement. The findings emphasize the importance of intentional conceptual education and comprehensive teacher preparation to foster lasting and significant comprehension in physics.
Multisensory virtual nature immersion is emerging as a tool for supporting mental health management. In this study, we evaluate the effectiveness of two types of multisensory immersion systems for patients diagnosed with Post Traumatic Stress Disorder: (1) a PORTABLE system, which delivers olfactory stimulation directly through a head-mounted display (HMD), and (2) a POD system, in which participants are seated in a multisensory pod that augments HMD audio-visual content with synchronized olfactory cues and haptic feedback (wind and vibration). Beyond comparing the two modalities, this work focuses on how neural activity changes throughout the experience. Specifically, we analyze electroencephalography (EEG) responses in relation to participants’ subjective affective ratings, clinical questionnaires, and cognitive faculties.
The study is devoted to the analysis of the impact of globalization and digital technologies on the transformation of language norms in modern digital discourse. The relevance of the work is due to the growing role of digital communication, in which language becomes a key factor of cultural identity and social integration. The goal is to identify interdependencies between the level of digital maturity of states, the intensity of language hybridization and the dynamics of sociolinguistic variability. The object of the study is the digital language space of the global communication environment. The methodology is based on a combination of systemic, institutional, econometric and corpus-linguistic approaches using official statistical databases and corpora of digital texts for 2015–2024. The results show that during this period ICT Development Index increased from 5.32 to 7.41 points, DESI Index – from 48.7 to 70.4 points, and the share of Internet users – from 58.2% to 84.3%. At the same time, the number of languages with digital representation increased from 312 to 387, and the share of English-language content decreased from 55.1% to 49.3%. The hybridization index increased from 0.42 to 0.61, clearly indicating the establishment of a polycentric mode or multi-node digital discourse. Hybridization is both method and degree: different languages or their structural elements used in the same message (verbal or symbolic), as well as fully/partially mixed code messages; hashtags expressed in different linguistic forms/fonts, etc. (multimedia message). This means how various linguistic components are arranged within one structural unit up to multimedia messages comprising codes written with different fonts‐forms on various levels of hybridity. The econometric model registered a very high positive correlation between digital maturity and hybridization (r=0.82).
Introduction: Blood flow restriction (BFR) walking elicits improved fitness, but participants often report higher ratings of perceived exertion (RPE) and pain during BFR walking compared to non-BFR walking. The primary aim was to investigate how multiple BFR walking exposures might affect RPE and pain. Methods: 14 healthy, trained participants completed three BFR walking sessions on separate days. The treadmill speed that elicited 3/10 RPE while BFR was not applied was determined and that same speed was used during all BFR walking sessions. Participants walked for 15 minutes at the predetermined speed while 60% limb occlusion pressure was applied bilaterally to the thighs. RPE (0-10) and pain (0-10) were recorded during each minute of exercise. Two-way repeated measures analysis of variance determined if session (1-3) and/or time (1-15 minutes) affected RPE or pain. Statistical significance was established at p<0.05. Results: RPE was higher during session 1 compared to session 2 (minutes 8-15, 4.4±1.4 versus 3.6±0.8). RPE was higher during session 1 compared to session 3 (minutes 3-5, 3.5±0.7 versus 2.9±0.6; minutes 7-15, 4.3±1.3 versus 3.5±1.1). No significant differences were observed for pain. Conclusion: Participants might tolerate BFR walking better after completing two BFR walking sessions as lowered RPE responses were observed.
To achieve deep natural language understanding, syntactic constituent parsing plays a crucial role and is widely required by many artificial intelligence systems for processing both text and speech. A recent approach involves using standard sequence-to-sequence models to handle constituent parsing as a machine translation problem, moving away from traditional task-specific parsers. These models are typically initialized with pre-trained encoder-only language models like BERT or RoBERTa. However, the use of pre-trained encoder-decoder language models for constituency parsing has not been thoroughly explored. To bridge this gap, we extend the sequence-to-sequence framework by investigating parsers built on pre-trained encoder-decoder architectures, including BART, mBART, and T5. We fine-tune them to generate linearized parse trees and extensively evaluate them on different linearization strategies across both continuous treebanks and more complex discontinuous benchmarks. Our results demonstrate that our approach outperforms all prior sequence-to-sequence models and performs competitively with leading task-specific constituent parsers on continuous constituent parsing.
This study examines the linguistic landscape (LL) of food stall signage in Ruteng by analyzing the linguistic forms, typologies, and communicative functions embedded within bottom-up commercial signs. Grounded in LL theory and enriched by geosemiotic, mobility, and language ideology perspectives, the research employed a descriptive qualitative design to explore how language operates symbolically and informationally in everyday public spaces. Thirty-six food stall signs were collected through field observation and analyzed for their linguistic composition, spatial emplacement, and semiotic features. The findings reveal that Indonesian overwhelmingly dominates monolingual signage, reflecting national language policy and shared linguistic norms, while bilingual and multilingual signs incorporate regional languages, Arabic, and English to index ethnic identity, religious affiliation, cultural heritage, or aspirations toward modernity. All signs constitute bottom-up signage, highlighting individual agency and cultural creativity in shaping the local LL. Functionally, the signs perform both informational roles—such as conveying menu offerings and halal status—and symbolic roles, expressing identity markers, migration histories, and ideological values. Overall, the study demonstrates that food stall signage in Ruteng functions as a meaningful semiotic artifact that materializes the intersection of language, identity, and mobility in the public sphere, offering valuable insights into the sociolinguistic dynamics of multilingual Indonesian communities.
These data contain three rows of data: 1. The unique identifier of the video and its order shown; 2. The subject's Likert rating of engagement, where 1-5 from least to most; 3. The subject's Likert rating of familiarity with the video's subject matter, also likert 1-5 from least to most.
ABSTRACT The widespread use of TikTok among elementary school students has brought noticeable changes to the way children communicate in their daily lives. The platform is no longer used merely as a source of digital entertainment, but has also begun to shape students’ word choices, speaking styles, and language habits. This condition can be observed among students at MIS Al-Khairaat Pombewe, who have become increasingly familiar with viral expressions, popular abbreviations, slang, and the mixing of Indonesian with foreign languages in everyday conversations. Such circumstances have raised concerns regarding the declining use of proper and standard Indonesian within the school environment. This study employed a descriptive qualitative approach involving the principal, teachers, and students selected purposively as research informants. Data were collected through observations, interviews, and documentation, then analyzed through the stages of data reduction, data presentation, and conclusion drawing. The findings reveal that TikTok exerts a dual influence on children’s language development. On the one hand, the platform contributes to vocabulary expansion, enhances students’ creativity in language use, and broadens their digital knowledge. On the other hand, the intensity of TikTok usage encourages the frequent use of informal language in formal situations, leading to a gradual decline in the use of proper Indonesian according to linguistic norms. Therefore, the involvement of teachers and parents is necessary to guide children toward wiser social media use without neglecting the development of their language abilities. ABSTRAK Fenomena penggunaan TikTok di lingkungan sekolah dasar memperlihatkan perubahan yang cukup nyata pada cara siswa berkomunikasi sehari-hari. Platform ini tidak lagi sekadar dimanfaatkan sebagai hiburan digital, tetapi turut membentuk pilihan kata, gaya berbicara, hingga kebiasaan berbahasa anak. Kondisi tersebut terlihat pada siswa MIS Al-Khairaat Pombewe yang semakin akrab dengan istilah viral, singkatan populer, bahasa gaul, serta pencampuran bahasa Indonesia dengan bahasa asing dalam percakapan mereka. Situasi ini memunculkan perhatian terhadap menurunnya penggunaan bahasa Indonesia yang baik dan benar di lingkungan sekolah. Kajian ini memanfaatkan pendekatan deskriptif kualitatif dengan melibatkan kepala sekolah, guru, dan siswa sebagai informan yang dipilih secara purposive. Informasi penelitian diperoleh melalui observasi, wawancara, dan dokumentasi, kemudian dipahami melalui tahapan reduksi data, penyajian data, dan penarikan kesimpulan. Temuan penelitian memperlihatkan bahwa TikTok memberi pengaruh ganda terhadap perkembangan bahasa anak. Di satu sisi, media sosial tersebut membantu siswa memperluas kosakata, meningkatkan kreativitas dalam berbahasa, dan memperkaya wawasan digital mereka. Di sisi lain, intensitas penggunaan TikTok ikut mendorong penggunaan bahasa informal dalam situasi formal sehingga kebiasaan menggunakan bahasa Indonesia sesuai kaidah menjadi semakin berkurang. Karena itu, keterlibatan guru dan orang tua dibutuhkan agar penggunaan media sosial dapat diarahkan secara lebih bijak tanpa mengabaikan perkembangan kemampuan berbahasa siswa.
Over the past six decades, a variety of systems have been developed for representing sign language forms, from Stokoe Notation (Stokoe, 1960) to SignWriting (Sutton, 1999) and lexical database schemas.Each was designed with specific goals and applications, leading to a fragmented landscape of representations.To enable greater interoperability and data sharing among sign language users and researchers, we propose a robust approach to translating between notation systems.As a first step in this direction, we introduce a formal mapping framework between HamNoSys and the SL CatForm coding schema, describe its implementation, and present empirical evidence of its performance.An extensive evaluation of mapping mismatches revealed improvements to the mapping logic needed to further advance the HNS2CF mapping tool.However, the initial version of the system already achieves an overall accuracy of 76.7% and an in-depth analysis reveals that many apparent mismatches stem from annotator disagreement rather than mapping errors, indicating that the tool's actual accuracy is even higher.These results demonstrate the feasibility and promise of establishing mapping mechanisms across sign representation systems.
Abstract This study examines the use of third-person singular inflections -th and -s in Early Modern English sermons, adopting a micro-sociolinguistic approach. Focusing on linguistic data from the Corpus of Sermons in Early Modern English (CoSEME), it aims to explain how and why the -th form was retained longer in religious language despite the broader linguistic shift to the -s form. The paper employs third-wave sociolinguistic theories and analyses the data within two frameworks: discourse community and community of practice. By investigating sermons from three generations of preachers, the study identifies a notable divergence in the use of -th forms, linking these patterns to religious identity and practice. The findings reveal that -th forms gained social meaning as markers of solemnity and authenticity within the religious domain, particularly among certain religious communities. The analysis highlights a complex interplay between language, community, and identity, demonstrating that the use of -th forms was influenced by sociopolitical factors, individual agency, and communal linguistic norms. The paper contributes to understanding the role of social meaning in historical language variation and change, offering insights into how specific linguistic features may be maintained within genres to project identity and align with community values. It also assesses the utility of sociolinguistic community frameworks for analysing the social significance of linguistic variation in historical contexts.
The article examines „Georgian“ jokes as a speech genre in which the linguistic norm is violated through connotative reduplication and other forms of language play. Based on Bakhtin’s theory of speech genres and the study of humor, an anecdote is interpreted as a metalanguage practice reflecting social attitudes towards normality, deviation, and linguistic authority. The analysis is carried out in the context of the linguistically polyphonic urban space of Old Tiflis, where Russian historically served as the language of the empire, comparable to the role of French in Southeastern Europe. This sociolinguistic situation contributed to active language contact and the formation of hybrid speech forms, which later became a source of anecdotal texts. Special attention is paid to reduplication as a key mechanism for creating a comic effect in „Georgian“ jokes. Echo constructions such as shashlik-mashlyk, salad-malat, culture-multur do not have an independent denotation. However, they perform pragmatic and semiotic functions: they enhance the expressiveness of the utterance, mark irony, and refer to a conscious deviation from the norm. An analysis of Soviet and post-Soviet anecdotal empirical material, including the „Georgian“ typical characters Gogi and Givi, shows that reduplication serves to carnivalize the norm and symbolically explore the center-periphery relationship. At the end of the article, the author concludes that reduplication in „Georgian“ jokes is a key mechanism of the comic effect. The „echoes“ of main parts of words, such as multur, malat, and mashlik, lack independent denotation and serve to carnivalize linguistic norms and increase the expressiveness of speech. They assume knowledge of the normative code and turn deviation from the norm into a conscious language game and a form of collective language memory.
This paper presents the new Universal Dependencies tree bank for the Macedonian language, marking a significant step towards the comprehensive linguistic analysis of Macedonian within the UD framework.It briefly addresses dependency grammar from a theoretical perspective and moves on to describing the treebank development process, from sentence selection, word segmentation and lemmatization, to POS-tagging, morphological features, and dependency tagging.Given the mostly manual labor invested in developing the treebank, semiautomatic NLP tools specifically designed for this purpose have also been presented and commented.Based on the defined tagset, the paper provides examples and visualizations of annotated sentences.The creation of the first Macedonian UD treebank enhances Macedonian linguistic resources and provides valuable insights into the morphological and syntactic structures of this Balkan Sprachbund language.It also contributes to the broader understanding of language-specific challenges within the UD framework and facilitates cross-linguistic comparisons in the Balkan and broader region.
Understanding how memories of past experiences shape subjective feelings is complicated by the fact that we constantly update our memories. These updates are particularly impactful when individuals are reminded of emotionally positive or negative attributes of the original event. Yet, it remains unclear how such memory updating influences subjective feelings. Here, we investigated how the reactivation of emotional information affects episodic memory, subjective feelings, and their interaction. Across three experiments, participants first learned both positive and negative attributes associated with unfamiliar individuals. Then, they were reminded of a single positive or negative attribute for each individual to reactivate the memory partially. Finally, we reassessed memory for and subjective feelings about each individual’s attributes. In Experiments 1 and 2, these procedures were distributed across three days, while in Experiment 3, they occurred on a single day. Across these three experiments, reminding with negative attributes shifted subjective feelings in a negative direction. Reminded attributes were also better remembered, particularly for negative ones, and changes in subjective feelings were more strongly associated with reminded attributes. However, positive reminders only influenced subjective feelings to change positively when all procedures occurred on the same day. Together, these findings support a model in which memory updating shapes both episodic memory and emotional experience in a valence-dependent manner.
Virtual reality (VR) offers new opportunities to promote active behaviors by enhancing engagement and allowing controlled modifications of urban environments. This study investigates whether virtual environments (VEs) can evoke affective responses comparable with real environments (REs), both psychologically and physiologically, by using an immersive VE combined with a walking simulator that replicates walking motion. Forty-nine healthy adults, Luxembourg residents or cross-border commuters, aged 18–65, including students, university staff, and the general public, walked two contrasting street segments, walking-friendly and car-friendly, in both RE and VE in a crossover design. Affective responses were assessed through questions on aesthetics, safety, enjoyment, comfort, relaxation, momentary stress, and real-time physiological data collected using E4 wristband. Significant differences emerged between the RE and VE across all affective measurements, except for nonspecific skin conductance responses, with the RE consistently eliciting more positive affective responses. Nevertheless, similar affective trends were observed in both the RE and VE across the two segments. Moreover, environmental characteristics significantly influenced affective responses in both the RE and VE, with the walking-friendly segment yielding more positive affective ratings than the car-friendly one. The interactions between environment type (RE vs. VE) and segment type (car-friendly vs. walking-friendly) were not significant for most measurements, indicating that the effect of environment type on affective responses remained consistent across segments. These findings emphasize that VEs can mimic the overall patterns of affective responses observed in REs. This research highlights VR’s potential in planning healthier cities, offering insights into its benefits and limitations for future research. • VR walks showed lower affective ratings than real-world walks. • VR replicated the overall affective trends observed in the real-world environment. • Car-friendly street segment triggered higher arousal in the VR and real environment. • VR is a valid tool for assessing human-environment interactions.
Abstract This paper examines the term lot, a lexical innovation associated with the Celtic root lue, and specific to the early medieval Latin documentation of the Catalan linguistic domain. In the context of the texts and charters from this region domestic livestock is frequently mentioned, primarily in wills, contracts of sale, donations, and ecclesiastical dowries. Drawing on the documentary corpus that underpins the Glossarium Mediae Latinitatis Cataloniae (GMLC), our research identifies the word lot, a term which, to date, is absent both from this glossary and from other lexicographical works specializing in the vocabulary of medieval Latin and early Catalan. On the basis of the evidence gathered in the lexical database of the Corpus Documentale Latinum Cataloniae (CODOLCAT), we propose an etymological and semantic interpretation of the term. Our hypothesis is that this word designates the young bull (sp. novillo ), the offspring of the bull over two years of age, and that is semantically distinct from bos ‘ox’ and uitellus ‘calf’. We also draw on the etymological dictionaries of Wartburg (FEW) and Meyer-Lübke (REW) within the framework of a comparison with related terms in other Romance and Celtic languages, such as Old French and Breton. This study highlights the importance of the medieval Latin corpus as a tool to identify lexical items of pre-literary Catalan that have not survived.
The study aim was to evaluate differences in ratings of valence made for a set of nonspeech sounds varying in spectrotemporal modulation. For 17 auditory-frequency bands (from 125 to 9474 Hz), modulation index values were extracted at five rates (from 2 to 32 Hz). Higher ratings of valence were associated with lower modulation values in the frequency region below 500 Hz and higher modulation values in the 1620-4000 Hz region. Listeners with hearing loss demonstrated significantly shallower relationships between modulation index and ratings of valence compared to adults with normal hearing. This could partially explain previously demonstrated compressed valence associated with hearing loss.
This article examines the role and significance of language corpora and linguistic databases in linguistic expertise. It discusses the possibilities of conducting objective semantic, pragmatic, and stylistic analyses of texts through corpus-based methods. The study also highlights the contribution of linguistic databases and artificial intelligence technologies to improving the accuracy, reliability, and efficiency of expert conclusions. Furthermore, the relevance of developing specialized corpora and databases for forensic linguistics in Uzbekistan is substantiated.
This paper presents a novel treebank-driven approach to comparing syntactic structures in speech and writing using dependency-parsed corpora. Adopting a fully inductive, bottom-up method, we define syntactic structures as delexicalized dependency (sub)trees and extract them from spoken and written Universal Dependencies (UD) treebanks in two syntactically distinct languages, English and Slovenian. For each corpus, we analyze the size, diversity, and distribution of syntactic inventories, their overlap across modalities, and the structures most characteristic of speech. Results show that, across both languages, spoken corpora contain fewer and less diverse syntactic structures than their written counterparts, with consistent cross-linguistic preferences for certain structural types across modalities. Strikingly, the overlap between spoken and written syntactic inventories is very limited: most structures attested in speech do not occur in writing, pointing to modality-specific preferences in syntactic organization that reflect the distinct demands of real-time interaction and elaborated writing. This contrast is further supported by a keyness analysis of the most frequent speech-specific structures, which highlights patterns associated with interactivity, context-grounding, and economy of expression. We argue that this scalable, language-independent framework offers a useful general method for systematically studying syntactic variation across corpora, laying the groundwork for more comprehensive data-driven theories of grammar in use.
This repository contains GSD-NP and GSD-DiNoS, both derived from Universal Dependencies' (UD) GSD Treebank. GSD-NP (.conllu) is a subset of UD-GSD and comprises its simplex noun phrases (NP): Common nouns (NN/NOUN) and their direct dependents (determiners, adnominal adjectives, nmods, adpositions, adverbs). It consists of 49,425 NPs (119.0k tokens) and has an improved feature annotation coverage (gender, case, number). Breaking with UD annotation, a total of 3,649 APPRART tokens were reconstructed in GSD-NP to restore the original orthographic forms. GSD-DiNoS (.json) is a custom data-driven lexion-like data structure built on GSD-NP, which aggregates NPs with the same head lemma. For each lemma, absolute frequencies of the lemma and its word forms are captured. Moreover, each occurrence feeds into three areas of interest within the word form entry: morphosyntactic features in isolation (gender, case, number), in combination with groups of dependents (collocations), and in combination with the syntactic function (dependency relations). GSD-DiNoS spans 17,433 unique lemmas and 20,190 unique word forms, stemming from 49,416 NPs. Lemmas were relemmatised to assign unique lemmas to nominal compounds, a highly productive and often lexicalised construction in German.
Everyday experiences can evoke positive feelings that differ among individuals and guide their behavior. Although reward processing is often linked to positive feelings, the mechanisms underlying subjective positive affect and whether they are shared for varied positive experiences is unclear. Here, we used fMRI and predictive modeling to investigate how dynamic, personalized positive experiences are encoded in the brain in humans of both sexes. Neural representations and functional integration during experiences of monetary reward, social media, music, and positive autobiographical memories were used to predict participants' affect ratings of each experience. Across experiences, positive affect was encoded in multivariate neural patterns and functional coupling of distributed cortical and subcortical brain areas, including some value-linked brain regions, like the orbitofrontal cortex. Restricted sets of brain representations and functional connections linked to sensory processing were involved in encoding stimulus-specific positive affect. Our findings suggest that positive affect may be computed and communicated throughout the brain.
Large language models (LLMs) offer a promising approach to machine translation (MT) for extremely low-resource languages by incorporating linguistic resources through in-context learning. However, LLMs often struggle to apply grammatical information effectively during translation. Inspired by recent progress in chain-of-thought reasoning, we investigate whether low-resource MT can benefit from structured intermediate steps of linguistic analysis and grammatical reasoning. We propose a pipeline for automatically generating step-by-step linguistic reasoning traces from Universal Dependencies treebanks, dictionaries, and grammar-rule banks. We evaluate these traces in three settings: in-context learning (ICL), supervised fine-tuning (SFT), and reinforcement fine-tuning (RFT), on Xibe and Chintang as test cases. Our results show that linguistic reasoning traces are most effective as inference-time guidance: in ICL, reliable sentence-specific traces substantially improve translation performance across most models, languages, and metrics. In contrast, using the linguistic reasoning traces as training data yields smaller and less consistent gains, as models learn the trace format but often generate erroneous content. These findings suggest that LLMs can leverage grammatical information for low-resource MT when given reliable linguistic analyses, while learning to generate such analyses remains a major bottleneck.
Implicit discourse relation recognition (IDRR) addresses the classification of discourse relations between text segments without explicit connectives. Existing prompt-based methods for IDRR often rely heavily on predicting surface connectives as an indicator for the discourse relation, which is inherently limited by the capacity of pre-trained language models. Meanwhile, standard attention mechanisms in these models are easily distracted by task-irrelevant tokens. This paper proposes a Role-Focused Prompt Framework that addresses these limitations by introducing a role-centric perspective to IDRR. Our approach is built on two core innovations: (1) the incorporation of linguistically grounded semantic roles (e.g., Cause/Effect for Contingency relation) into IDRR, which directly captures the underlying argument structure that determines discourse relations, reducing reliance on connectives; (2) a focused prompt structure that condenses the input to its core semantic concepts (argument summaries, connective, and semantic roles), creating a high signal-to-noise environment for attention-based reasoning. Extensive experiments on Penn Discourse TreeBank 2.0 (PDTB 2.0) demonstrate that our framework achieves competitive results, providing complementary direction for IDRR research. Ablation studies validate that both innovations are essential to the framework. Our work demonstrates that incorporating linguistically grounded semantic roles and focusing on task-relevant concepts can effectively specialize pre-trained models for IDRR.
Gender equality remains a cornerstone of sustainable social development, yet in Zimbabwe, the persistent use of sexist language in meetings, workshops, and public gatherings continues to reinforce gender hierarchies and marginalize the girl child. This study is of critical importance as it interrogates the linguistic and cultural foundations that perpetuate gender bias, aiming to advance gender-neutral communication as a pathway to empowerment and equity. Despite numerous gender mainstreaming policies, a significant research gap exists in understanding how everyday sexist discourse sustains systemic inequalities and limits girls’ participation and agency in social, educational, and professional spaces. The primary objective of this study is to examine the prevalence, forms, and socio-cultural implications of sexist language in Zimbabwean institutional and community contexts, and to propose strategies for promoting inclusive and gender-sensitive communication. Adopting a mixed-methods design, the research combines qualitative interviews, focus group discussions, and document analysis with quantitative surveys conducted across government, educational, and civic institutions. Data will be analysed thematically and statistically to capture both the depth and breadth of linguistic gender bias. Preliminary findings indicate that sexist language is deeply embedded in traditional communication practices, institutional norms, and even policy discourse, resulting in subtle but pervasive disempowerment of girls and women. However, evidence also points to a growing awareness and readiness among educators, policymakers, and community leaders to adopt gender-neutral communication frameworks. The study’s implications are profound: by reshaping linguistic norms, Zimbabwe can cultivate environments that affirm equality, inclusivity, and respect. The research contributes to global discourse on gender and language while providing actionable recommendations for policymakers, educators, and advocacy groups to dismantle linguistic barriers and foster the full empowerment of the girl child.
In this paper we introduce an example-based method for exploring dependency treebanks that is based on principles of vector symbolic architectures.It leverages key properties of this framework to provide fast and flexible search capabilities, since all combinations of query parameters can be compared with a given parse tree in parallel via a single vector operation.The framework also allows for graded similarity and the natural integration of various kinds of information, such as word embeddings.After some background on the framework and an explanation of our implementation, we provide a few examples of the system's output and draw comparisons to similar applications.
This paper introduces the first steps towards the creation of a novel resource for contemporary Sardinian within the Universal Dependencies framework.Sardinian is a Romance language spoken in Sardinia, an island belonging to the Italian Republic and located in the center of the western Mediterranean.It is a minority and endangered language, traditionally transmitted mainly orally, and characterized by a multiplicity of varieties (usually grouped into two macro-varieties Logudorese and Campidanese), all recognized as part of the Sardinian linguistic continuum.These varieties share basic morphosyntactic features, while presenting differences at the lexical level and in the realization of specific constructions.This internal variation can be particularly challenging with regard to the normalization of lemmas and the linguistic characterization of certain phenomena.The development of the treebank therefore aims to provide an annotated resource for contemporary Sardinian that takes into account the specificities of the different varieties, using Universal Dependencies to represent them within a unified theoretical framework, in order to facilitate both linguistic analysis and automatic processing.The present paper thus describes some linguistic characteristics of Sardinian and the attempts to encode them within the UD framework.Finally, we present the results of our evaluation of an NLP pipeline for Sardinian, trained on our corpus, for the Stanford Stanza parser.
This thesis examines how affective relations to mathematics emerge, become visible, and are sustained in Swedish early childhood education (preschool and preschool class). Integrating psychological, physiological, and interactional perspectives, it explores how children’s orientations toward mathematics are shaped through teachers’ affective dispositions, children’s real-time engagement, and the multimodal practices of everyday preschool interaction. Three empirical studies address two overarching questions: How teachers and children co-construct affective relations to mathematics, and What different methodological approaches contribute to understanding these relations. Study I used a municipal survey of 357 educators, employing an adapted Math Anxiety Scale for Teachers of Preschool. A two-factor structure (general vs. teaching-related anxiety) was confirmed. Results showed that certified preschool teachers with higher math anxiety reported significantly lower frequency of math teaching and math-related talk—particularly in structured situations. This association did not appear for caregivers, revealing a profession-specific pathway through which teachers’ affective orientations may shape children’s opportunities to encounter mathematics. Study II combined pupillometry, child-friendly affect ratings, and visuospatial working memory measures with preschool class children (mean age 5.67) and one parent each. Task difficulty reliably increased pupil dilation during problem processing, but anticipatory activation before math tasks was minimal. No associations emerged between children’s affect ratings and physiological measures, nor between children’s and parents’ math anxiety. These findings suggest that physiological components typically associated with math anxiety are not yet consolidated at this age. Study III analyzed nearly six hours of video-recorded everyday mathematics activities in a preschool using multimodal conversation analysis. Teachers’ responses to children’s affective stances—through prosody, gaze, body orientation, gesture, and material action—organized participation in counting, problem-solving, and emergent math play. Response practices accomplished re‑engagement, repair, explanation, and moral work, demonstrating that affect is central—not peripheral—to the organization of early mathematics activities. Across studies, affective relations to mathematics appear relational, situated, and developmentally emergent. The thesis shows the value of mixed methods for capturing both intrapersonal and interactional dimensions of affect. Practically, it highlights the need to support teachers’ own affective orientations to mathematics and to design early math environments that combine cognitive challenge with interactional practices that recognize and build on children’s affective engagement.
In recent years, Natural Language Processing has improved significantly, but most developments favor high-resource languages. Due to the absence of annotated corpora, lexical databases, and computational tools many languages are under-resourced. Morphosyntactic analysis is a fundamental component of several NLP tasks, including part-of-speech tagging, syntactic parsing, and machine translation. It focuses on understanding words forms and sentence structure in a language. The under-resources languages have limited linguistic resources and contain complex morphological structures which makes the development of computational approaches for morphosyntactic analysis challenging. This study reviews different computational methods used for morphosyntactic analysis in under-resourced languages. 19 studies were reviewed to examine the technique used, research areas and performance trends in the literature. The results show that the percentage of neural network-based is about 42% of the analyzed literature, then statistical ML methods 32% and the rule-based approaches 26%. In Performance comparison the neural network models (82%) achieve higher accuracy compared to statistical models (75%) and rule-based models (70%). Also, it shows that the morphological processing and language resource development is the most investigated research areas. These developments and performance of the models are affected by limited annotated datasets and linguistic diversity. The results show the need for better linguistic resources and hybrid computational approaches for morphosyntactic analysis in under-resourced languages. These can help guide future research to develop better NLP tools for these languages.
This study has investigated how teachers at a rural school and an urban school perceive the interplay between dialect, dialect levelling and linguistic norms in teaching. The researcher also observed four Swedish language lessons in grade 4 and 6 at each school. In addition, one lesson in each grade was observed in other subjects, such as mathematics, history, geography and home and consumer studies, resulting in a total of eight observed lessons. To explore the teachers’ perceptions, four semi-structured interviews were conducted with teachers who teach Swedish. The semi-structured interviews, in combination with participant observations, complemented each other well and contributed to strengthening, problematizing, and nuancing the findings. The results show that students’ spoken language varies between a Västerbotten dialect, a more standard variety of Swedish, and a form of language influenced by social media, including slang, abbreviations, informal chat language, and English words and expressions. It emerged that the teachers perceived the influence of social media on language as relatively strong, and that words becoming popular on social media spread quickly and become a natural part of students’ spoken language. The observations also revealed variations in how students’ spoken language was expressed depending on the teaching situation and context. Furthermore, the results show that different forms of dialect are still present in students’ speech through dialectal words and expressions, as well as features such as stress, prosody, and pronunciation. Signs of dialect levelling were also identified, as some students used a more standardized form of language in different situations and contexts.
This examines the specifics of linguistic tools functioning as instruments for shaping and expressing cultural identity in the works of the prominent Quebec writer Jacques Godbout. The study primarily focuses on a comparative analysis of the author’s landmark works: the early novel “Hail Galarneau!” (1967) and the later work “The Concierge of the Pantheon” (2006). The theoretical framework of the analysis is based on Lise Gauvin’s concept of “linguistic hyperconsciousness” (surconscience linguistique), which views the language of a Francophone writer in a different cultural environment as an object of constant reflection and conflict. The analysis traces the evolution of Jacques Godbout’s authorial strategy: from the active use of “joual” and neologisms as instruments of socio-political resistance during the Quiet Revolution to a subtle, ironic play with linguistic norms in the heart of the metropolis — Paris. The study demonstrates that in the novel The Concierge of the Pantheon, language transforms into an independent character, embodying the author's linguistic hyperconsciousness. Through the prism of Pierre Léon’s semiotic model of phonostylistic analysis, the research identifies the role of linguistic “indicators” (involuntary markers of origin) and “signals” (conscious markers of otherness) in the speech of the protagonist, Julien Mackay. Special emphasis is placed on the interweaving of professional meteorological discourse (“des tempêtes d’amour”, “spécialiste des orages domestiques”), Quebec dialectaisms (“bout d’criss”, “l’air cheap”), and anglicisms-calques (“tomba amoureux fou”, “ces cahiers made in France”, “Gibert Jeune est un self”). It is established that by employing markers of the hero’s geographical and mental affiliation within the context of “flawless” Parisian French, the author creates an effect of “detonation.” The scientific novelty of the paper lies in determining the mechanisms of transforming “linguistic hyperconsciousness” into an act of linguistic resistance, where irony serves as a key tool for deconstructing the myth of the monolithic academic norm. The study concludes that Jacques Godbout’s oeuvre represents the complex trajectory of Quebec literature: from the struggle for self-identification to integration into the global Francophone context through the preservation of its own linguacultural uniqueness.
This paper introduces the first steps towards the creation of a novel resource for contemporary Sardinian within the Universal Dependencies framework. Sardinian is a Romance language spoken in Sardinia, an island belonging to the Italian Republic and located in the center of the western Mediterranean. It is a minority and endangered language, traditionally transmitted mainly orally, and characterized by a multiplicity of varieties (usually grouped into two macro-varieties Logudorese and Campidanese), all recognized as part of the Sardinian linguistic continuum. These varieties share basic morphosyntactic features, while presenting differences at the lexical level and in the realization of specific constructions. This internal variation can be particularly challenging with regard to the normalization of lemmas and the linguistic characterization of certain phenomena. The development of the treebank therefore aims to provide an annotated resource for contemporary Sardinian that takes into account the specificities of the different varieties, using Universal Dependencies to represent them within a unified theoretical framework, in order to facilitate both linguistic analysis and automatic processing. The present paper thus describes some linguistic characteristics of Sardinian and the attempts to encode them within the UD framework. Finally, we present the results of our evaluation of an NLP pipeline for Sardinian, trained on our corpus, for the Stanford Stanza parser.
Language has historically functioned as both a communicative system and a repository of cultural identity, social memory, and collective belonging. In the contemporary digital era, Artificial Intelligence (AI) technologies increasingly mediate communication practices through predictive text systems, speech recognition software, algorithmic moderation, and Large Language Models (LLMs). Although such technologies have improved efficiency, accessibility, and multilingual interaction, emerging scholarship suggests that AI systems may also reproduce existing linguistic hierarchies by privileging dominant and standardized language forms while marginalizing dialectal and non-standard varieties (Helm et al., 2024; Hofmann et al., 2024). This theoretical paper examines how AI-driven communication systems may contribute to processes of linguistic standardization and dialect marginalization within digital environments. Drawing upon Fricker’s (2007) theory of epistemic injustice, the paper introduces the concept of the Digital Language Divide to explain disparities in computational recognition, linguistic legitimacy, and technological visibility among speech communities. Unlike deterministic arguments that portray technology solely as a force of linguistic erosion, this paper adopts a balanced perspective acknowledging that AI technologies may simultaneously expand communication access while also reinforcing structural inequalities in language representation. Recent empirical studies demonstrate that some AI systems exhibit dialect prejudice, particularly toward speakers of African American English and other non-standard language varieties (Hofmann et al., 2024). Additionally, emerging scholarship on techno-linguistic bias argues that language technologies frequently prioritize dominant linguistic norms embedded within training datasets (Helm et al., 2024). Building upon these findings, this paper theorizes how algorithmic standardization may gradually influence linguistic practices, educational expectations, and communicative participation within digitally mediated societies. This paper seeks to establish a conceptual foundation for future interdisciplinary inquiry into the sociolinguistic consequences of AI-mediated communication. The study concludes by proposing directions for future empirical research, policy development, and inclusive AI design frameworks capable of supporting linguistic diversity within digital ecosystems.
During the recent years, the use of linguistic data for language processing increased progressively. Such data are now commonly called language resources. Most of the language resources used for this purpose are collections of texts as the Brown Corpus and the Penn Treebank, but electronic lexicons (WordNet, FrameNet, VerbNet, ComLex, Lexicon-Grammar...) and formal grammars (TAG...) developed recently. Most processes of construction of lexicons and grammars are manual, whereas the construction of corpora has always been highly automated. However, more and more specialists of language processing realize that the information content of lexicons and grammars is richer than that of corpora, and hence the former make more elaborate processing possible. The difference in construction time is likely to be connected with the difference in information content: the handcrafting of lexicons and grammars by linguists would make them more informative than automatically generated data. This situation can evolve into two directions: either specialists of language technology get progressively used to handling manually constructed resources, which are more informative and more complex, or the process of construction of lexicons and grammars is automated and industrialized, which is the mainstream perspective. Both evolutions are already in progress, and a tension exists between them. The relation between linguists and computer scientists depends on the future of these evolutions, since the first implies training and hiring numerous linguists, whereas the other depends essentially on solutions elaborated by computer engineers. The aim of this article is to analyse practical examples of the language resources in question, and to discuss about which of the two trends, handcrafting or generating industrially, or a combination of both, can give the best results or is the most realistic.
While the linguistic shifts between Ottoman and modern Turkish are well-documented qualitatively, quantitative analyses remain scarce.This study addresses this by conducting a comparative computational analysis using two Universal Dependencies treebanks: OTA-DUDU for Ottoman Turkish and TR-BOUN for modern Turkish.By employing descriptive statistics and a log-likelihood ratio test, we demonstrate the change and quantify the magnitude of diachronic variation.The analysis yields three primary statistical findings.First, our data reveals a 77% compliance rate with labial vowel harmony for suffixes, while this value is 98% in modern Turkish.This discrepancy can be explained by the presence of rounding in Ottoman Turkish, which disappears in modern Turkish.On the other hand, the compliance rate of palatal vowel harmony is quite high for both languages, 96% for Ottoman Turkish and 99% for modern Turkish.Second, some suffixes, such as the converb -(y)Ip 1 and the dative infinitive -mAyA, changed by reducing their allomorphs in modern Turkish.Third, we demonstrate that Arabic and Persian pluralization rules, which constituted 28% of plural nouns in Ottoman Turkish, lost their pluralizing function in modern Turkish, although the words remain with singular meaning.
Research has widely documented age differences among adults in the ability to identify facial emotion expressions from pictures. In the current study, we examine how emotion perception ability is related to cognitive abilities, empathy, and relationship satisfaction. We use data from 151 older couples (M age = 72.6 years, age range: 65–86 years) collected in Switzerland and Southern Germany in 2018 and 2019. Participants were assessed on their abilities to perceive emotion from (dynamic) pictures of faces and on the abilities of fluid and crystallized intelligence, processing speed, and memory. Self-report measures assessed affective and cognitive empathy and relationship satisfaction. Additionally, empathic accuracy was assessed through the comparison of individual and partner positive/negative affect ratings obtained through momentary assessments conducted six times a day for 14 days. Associations between emotion perception and cognitive abilities were examined in a structural equation model. A model including separate latent factors for emotion perception, fluid intelligence, crystallized intelligence, processing speed, memory functioning and showed an acceptable model fit. Older participants scored lower than younger participants on all cognitive abilities except for crystallized intelligence. Covariances revealed that emotion perception ability is significantly related to all cognitive abilities within individuals. Emotion perception ability was not correlated between partners. There were no sex differences in emotion perception ability, and it was not significantly related to empathy (self-reported affective/cognitive empathy and empathic accuracy) or relationship satisfaction. Taken together, our findings suggest that cognitive ability is the closest correlate of emotion perception from images of facial expressions in older adults.
Recent advances in multimodal large language models (MLLMs) have greatly improved image understanding and captioning capabilities. However, existing image captioning benchmarks typically suffer from limited diversity in caption length, the absence of recent advanced MLLMs, and insufficient human annotations, which potentially introduces bias and limits the ability to comprehensively assess the performance of modern MLLMs. To address these limitations, we present a new large-scale image captioning benchmark, termed, ICBench, which covers 12 content categories and consists of both short and long captions generated by 10 advanced MLLMs on 2K images, resulting in 40K captions in total. We conduct extensive human subjective studies to obtain mean opinion scores (MOSs) across fine-grained evaluation dimensions, where short captions are assessed in terms of fluency, relevance, and conciseness, while long captions are evaluated based on fluency, relevance, and completeness. Furthermore, we propose an automated evaluation metric, \textbf{ITIScore}, based on an image-to-text-to-image framework, which measures caption quality through reconstruction consistency. Experimental results demonstrate strong alignment between our automatic metric and human judgments, as well as robust zero-shot generalization ability on other public captioning datasets. Both the dataset and model will be released upon publication.
Background and Aims: Most men consume pornography, with a small but significant percentage losing control over their use. Since ICD-11, problematic pornography use can be diagnosed as "compulsive sexual behavior disorder." Debate persists on whether problematic pornography use is an impulse-control disorder or a behavioral addiction. Mechanisms of learning and memory play a central role in addictive disorders but are presumably less relevant for impulse control disorders. Methods: One hundred thirty-nine heterosexual male users of pornography and gaming participated in our study which was part of a multi-center research project on internet use disorders in Germany. We focus on a subsample of fifty-eight non-problematic (n = 35) and problematic pornography users (n = 23, labeled pathological). FMRI data were collected during appetitive conditioning, extinction and recall. Pornographic, game, and money images served as unconditioned stimuli, geometric shapes as conditioned stimuli (CS). Results: During appetitive conditioning pathological pornography users showed a generally stronger response in ventral striatum to all CSs, whereas altered activations in extinction and recall were specific to the porn-associated CS. Greater activations in the dorsal anterior cingulate cortex during extinction and in the medial orbitofrontal cortex during recall suggest persistence of appetitive memory for pornography in pathological users, supported by valence ratings and skin conductance responses (SCR). Sensitization to the monetary cue also emerged in SCR. Discussion and Conclusions: Based on these new neurobiological findings, which are consistent with current addiction theories about stimulus-specific altered reward sensitivity and appetitive memory, we argue that problematic pornography use should be considered a behavioral addiction.