In Standard Arabic (SA), the (dual) quantifier KILAA/KILTAA displays an interesting paradigm of gender and number agreement that is conditioned by semantic interpretation: while collective KILAA/KILTAA agrees in natural duality number and grammatical gender with its predicate, distributive KILAA/KILTAA reflects agreement in grammatical singularity number and grammatical gender with the predicate in question. This paper shows that a simplified minimalist analysis which utilizes the formal tools of relativized probing based on the minimally-constrained cyclic Agree operation best captures the paradigm in question.
Constructions with si in Spanish are meaningful choices which pursue different communicative goals. These are studied as a prototype with contextual manifestations. Prototypical si-constructions are considered to be those that are structured under the protasis-apodosis schema, and that may or may not convey conditional and/or hypothetical meanings. The particle si functions as a space builder of assertiveness in both prototypical and non-prototypical constructions. All interpretations of si-constructions are dependent on the relationship established between the protasis and the apodosis and also on the meaning introduced by si and its combination with verbal tenses. The indicative present tense promotes the basic assertiveness meaning, interacting with si in most constructions. As a common feature, the present tense is almost categorical, which confirms the assertiveness cognitive domain that the conjunction si builds. Prototypical and non prototypical constructions create diverse pragmatic meanings and are unevenly distributed across different kinds of texts.
We disassociate two stages in visual complex pseudoword recognition, namely syntactic licensing and semantic composition (Neophytou et al. 2018), by comparing pseudowords violating category selection rules to those violating semantic rules of affix attachment in two closely related languages: Slovenian and Bosnian-Croatian-Serbian. Contrary to previous studies relying on argument structure violations in Greek and English (e.g., Manouilidou & Stockall 2014), we based our semantic violations on a more unambiguously semantic restriction of state stability by focusing on prefixes raz-, od-, and vz-/uz-that do not attach to stable state verbs ('to dwell'). In two acceptability judgment tasks and two lexical decision tasks, we show that semantic violations (raz-'dwell') are consistently more acceptable, rejected more slowly, and less accurately than category selection violations (raz-'mother'), across prefixes and languages. This adds evidence for the distinction between two post-decomposition stages from a new semantic dimension and supports the universality of this distinction in lexical processing.
This article explores synchronic and diachronic aspects of anticausativization in Italian, focusing on variability in its morpho-syntactic encoding and the parameters determining it, both verb-related-templatic (reflecting the event structure template of verbs) as well as root/lexical (concerning the type of root and change encoded)-and argument-related, involving thematic (e.g., agentivity) and inherent (e.g., animacy) properties of verbal arguments.
While Modern Standard Arabic (MSA) is well-studied, dialectal Arabic texts, such as Moroccan dialect (Darija), pose unique challenges due to their informal structure and lack of a standardized grammar. In this paper, we provide an in-depth study of Darija detailing its morphological and syntactic features, and we introduce DiMorph (Dialectal Morphological Analyzer), a specialized morphological engine, which is designed to address these complexities automatically. In detail, we focus on DiMorph's multi-phase approach, involving both pre-and post-processing phases. Such approach effectively manages dialectal variability and achieves high accuracy in token recognition and analysis, particularly in social media contexts. Finally, we underscore the importance of developing tools that respect the linguistic diversity of Arabic dialects, laying a strong foundation for advanced computational research in Arabic dialectology.
Language impairments are often observed in neurodegenerative disorders, but the role of short-term phonological and working memory in sentence processing remains unclear. This study assessed sentence comprehension in 18 individuals with Mild Cognitive Impairment due to Alzheimer's disease (MCI-AD) and 18 age-matched controls using a sentence-to-picture matching task. The task involved relative clauses with syntactic complexity manipulated through sentence type, branching direction, and linear length. Cognitive and memory skills were also evaluated. Participants with MCI-AD showed greater difficulty comprehending object-relative clauses and center-embedded structures, with no effect of linear length. The asymmetry between subject-and object-relative clauses confirms that filler-gap dependencies across an intervening noun phrase significantly affect sentence comprehension and that centre-embedding structures built on the subject increase the processing load. This is arguably because the matrix subject is separated from the main verb with which it agrees. Our study also highlights the role of the working memory resources necessary to compute filler-gap dependencies in structures of intermediate complexity, such as centre-embedding SRs. These findings indicate that working memory limitations modulate sentence processing difficulties in MCI-AD, and that their role is visible depending on the grammatical complexity of the structures involved.
The main descriptive and analytical goal of this article is to unify under the umbrella of epenthesis +/- the insertion of meaningless phonological material +/- a number of lexically and morphosyntactically conditioned phenomena that have been treated disparately in past research. Many previous analyses of what we call non-canonical epenthesis degrees have resorted to extra devices such as listed allomorphs and abstract segments. We reanalyze the data as instances of epenthesis with morphosyntactic and lexical conditioning, without resorting to such devices. Throughout, we separate the phonological fact of epenthesis from the lexical and morphosyntactic conditions that determine precisely what segment is epenthesized. Finally, we model the interactions between variously conditioned forms of epenthesis using Boolean Monadic Recursive Schemes (Bhaskar et al. 2020; Chandlee & Jardine 2021).
Reading aloud involves the complex interplay of visual, motor and lexical processes. While eye movements have been extensively investigated in the reading literature, less is known about the coordination of voice, eye and finger movements in oral and finger-point reading. Here we propose a multimodal perspective on these dynamics, emphasising the contribution of integrating eye-tracking, finger-tracking, and voice recording to a more comprehensive understanding of reading proficiency. Our results show that finger and eye movements are strongly coupled in early readers. Conversely, skilled readers show a more flexible coordination of sensorimotor signals and a more adaptive sensitivity to prosodic structures, with voice articulation slowing at key structural points, such as chunk heads and sentence-final boundaries. These findings provide novel insights into how multimodal coordination evolves with reading expertise, contributing to a more fine-grained understanding of reading fluency.
Language is assumed to be a multimodal system in which prosody, manual and non-manual gestures, and body positioning interact with syntactic structure in systematic ways. While gestures have often been analyzed from a semantic perspective, recent work has shown that they can be integrated into the syntactic architecture of the clause. We adopt this perspective and propose an analysis of co-speech gestures in expressive contexts, focusing on emotional meanings such as surprise and disapproval. Data from typologically diverse languages are considered. The discussion addresses three main questions: (i) What triggers the use of gesture in expressive utterances? (ii) Are these gestures language-specific or universal? (iii) What is their role within the grammar? We argue that a formal model integrating syntax, prosody, and gestures is required to account for the structural properties of expressive phenomena and to advance a theory of language as a fundamentally multimodal architecture.
The topic of the differences between handwriting and typewriting has been approached from many angles, but almost never with regard to the effects that the different writing modes have on the structure of the texts produced. Generally, the benefits that handwriting has in the learning phase are highlighted (on areas such as students' cognitive skills, reading ability, and spelling accuracy). On the other hand, typewriting has a positive effect on the amount of writing and on motivation. Instead, this contribution aims to investigate possible differences between handwritten and typed texts. The analysis conducted on a small sample of handwritten and typed texts from a sample of 200 students enrolled in the first year of the Bachelor's degree in Humanities at the University of Bologna shows that the texts are very similar for almost all the parameters investigated. However, some differences emerge for the parameters concerning textual cohesion and coherence. In this case, data suggest that typewriting ultimately seems to favour the production of better organised and better structured texts.
When is sound paired with meaning during language production? The exploration of inner speech offers a unique opportunity to approach this fundamental question, allowing a general reflection on the architecture of human language structure and its evolution. Recent experiments based on awake surgery techniques show that during language production the code exploited by neurons contains acoustic information even in non-acoustic areas such as Broca's area and even during inner speech, that is without externalization. This result implies that inner speech cannot per se allow decoupling of acoustic and syntactic information. In the last part of the paper it will be shown how this can be obtained in a different manner by analysing the brain reaction to homophonous phrases with stereoelectroencephalography and evaluate the results with different statistical models.
Motor aspects and bodily experience in general have traditionally been considered peripheral in models of neural and functional architecture of language. This opinion paper draws on two clinical case studies to explore the deep connection between sensorimotor components and symbolic processing systems. To this end, we report on the use of the Digital Linguistic Biomarker (DLBs) technique for analyzing speech and writing samples from older adults with dementia and adolescents with anorexia nervosa. In the former case, our data show that well-known cognitive symptoms of neurodegeneration (e.g., memory loss and planning deficits, word retrieval and sentence comprehension difficulties) are preceded by subtle acoustic-articulatory alterations. In the latter case, physical symptoms associated with extreme weight loss are accompanied by a gradual decline in syntactic complexity. Overall, these findings indicate that even in disorders that appear highly selective in nature, the deficits affect communicative competence across multiple interconnected levels that are deeply grounded in sensorimotor processes.
Language learning is inherently multimodal, combining auditory, visual, and textual inputs that reflect the complexity of natural language and communication. Multimodality can also promote second language (L2) acquisition by enhancing perceptual and productive language skills. This in perceiving and producing Southern Standard British English (L2) vowels /e/, /ae/, /D/, and /Lambda/. In Experiment 1, we compared audiovisual and temporarily improved perception of visually salient sounds but failed to stabilize auditory representations. Experiment 2 examined vowel production after exposure to TV series clips with L2 (English), L1 (Italian), or no subtitles. L2 subtitles improved /ae/ production, demonstrating the value of textual input in linking visual, phonetic, and orthographic information. However, the L2 vowel /Lambda/ showed limited improvement, suggesting multimodality's effectiveness depends on the phonetic salience and relevance of features within the L1 system. These findings highlight multimodality's potential to its impact.
When girls are pearls, does it mean that they are beautiful or that they are pleasant? Not only are metaphors open to different interpretations but also these interpretations might vary across individuals, even with the same cultural context. However, the literature lacks a description of which patterns of interpretation emerge across individuals and which factors might drive them. Here, we investigated the role of multimodality, intended as the contribution of different dimensions of experience-based information, to explain individual variability in metaphor interpretation. We analyzed participants' interpretations in a metaphor verbalization task according to a series of semantic features of words (affective, cognitive, and sensory) that mirror different cognitive mechanisms. With an innovative method that combines i) Natural Language Processing (NLP), ii) a multivariate statistical technique that derives Intersubject Representational Dissimilarity Matrices (IS-RDMs), and iii) a data-driven clustering method, we were able to identify two groups of participants. One cluster, which we named mentalizers, exhibited a greater use of cognitive and affective terms (e.g., the girls-pearls metaphor was explained as indicating that girls are pleasant), while the other cluster, which we named imagers, capitalized more on words expressing sensory-based features (e.g., girls were described as beautiful). Our study showed that a data-driven approach can capture different interpretative profiles from word-level semantic features and that differences are driven by the sensorimotor vs. sociocognitive dimensions. This suggests that there are alternative routes to derive metaphorical meaning, involving different modality systems in the multimodal network for metaphor.
This paper presents BAMBI (BAby language Models Boostrapped for Italian), a series of Baby Language Models (BabyLMs) trained on data that mimic the linguistic input received by a five-year-old Italian-speaking child. The BAMBI models are tested using a benchmark specifically designed to evaluate language models, which takes into account the amount of training input the models received. The BAMBI models are compared against a large language model (LLM) and a multimodal language model (VLM) to study the contribution of extralinguistic information for language acquisition. The results of our evaluation align with the existing literature on English language models, confirming that while reduced training data support the development of relatively robust syntactic competence, they are insufficient for fostering semantic understanding. However, the gap between the training resources (data and computation) of the BAMBI models and the LLMs is not fully reflected in their performance: despite LLMs' massive training, their performance is not much better than that of BAMBI models. This suggests that strategies beyond scaling training resources, such as data curation, inclusion of multimodal input, and other training strategies such as curriculum learning, could play a crucial role in shaping model performance.
Hungarian compound verbs have become more and more productive in the last few decades. They have traditionally been considered as the outcomes of V <- N back-formation. However, appeals to back-formation as a particular morphological process only scratches the surface of a phenomenon whose formal realization is of secondary importance. The paper demonstrates that the same associative and analogical relations may underlie back-formation, forward-formation, and cross-formation. This formal underspecification is neither language-specific nor construction-specific. It may be observed, for instance, in English compound verbs and nominalized Dutch particle verbs. Formal diversity also raises the issue of generalizations in derivation. Source-oriented generalizations are based on the relationship between distinct constructions, they involve information about the scope and nature of mapping between the base and the derivative. Product-oriented generalizations, for their part, provide schematic information about the semantic and formal outcome of derivation. Source-oriented and product-oriented generalizations are always interrelated in derivational constructions. Nevertheless, derivational constructions also vary in the relative prominence of these generalizations. Hungarian compound verbs may be characterized by predominantly source-oriented generalizations that concern the scope and nature of mapping, and their constructional schema generalizes very little information about the output. However, there are derivational constructions, for instance, some Hungarian onomatopoeic verbs, that do not have bases, and represent the other end of the scale.
French assigns grammatical gender (masculine or feminine) to nominals, and is endowed with a diminutive suffix-et/-ette. In most cases, the diminutive noun resulting from-et(te)-affixation will have the same gender as its base (Bally 1932), but there are a significant number of exceptions to this rule, that most of the previous literature (Dauzat 1937; Milner 1989 i.a.) took to be the result of lexicalization. In this study, we assess how frequent gender mismatches induced by et(te)-affixation are, in either direction (masculine to feminine and vice-versa), and what the exact semantic consequences turn out to be. In particular, we show that there exists a significant frequency asymmetry between-et-affixation and-ette-affixation, which affects both gender-matching and gender-mismatching base-derivative pairs, supporting the idea that gender-mismatching diminutives are to a certain extent morphologically transparent, but also that-ette-affixation may receive an analysis distinct from that of-et-affixation. We provide a analysis within the framework of Distributed Morphology (Halle & Marantz 1993) that is in line with the statistical data and with recent cross-linguistic findings on diminutive an augmentative affixes, according to which such elements may vary in place and manner of attachment, across, and also within, languages (Wiltschko & Steriopolo 2007).
The French suffixation in -ier, -i & egrave;re is a system inherited from Latin, which has been largely remodelled. One of its most striking features is the great number of different meanings that the derived nouns in -ier show. Contrary to what may seem, it is argued that the suffix -ier should not be considered as polysemic and that such a hypothesis even prevents to correctly describe the facts. The suffix is meaningless which means that compositionality is not an option to account for the meaning of the derivatives in -ier. It is proposed that this meaning be developed on the basis of inferences drawn from the semantics of their base name and regulated by the derivational series and families in which the derived Ns are included. Suffixes which supposedly manifest a true polysemy are discussed and it is shown that this is not a true polysemy, as is the case with lexemes, but an effect of the multiplicity of meaning obtained by lexeme formation patterns.
This article explores the notion of derivational paradigm. Although several studies have proposed a paradigmatic approach to derivational morphology, we do not know yet what derivational paradigms look like. A key feature of paradigms in inflection is the mutual predictability of the paradigm cells in terms of the content that they express, but we don't know yet how predictability works in the derivational lexicon. In order to explore predictability in derivation, we propose to use scenarios (i.e. prototypical representations of real-world situations). The idea is to build scenarios using short stories produced by Large Language Models (LLMs). We create stories containing pairs of lexemes belonging to the same word family; the regular content of the stories and the participants that frequently co-occur in them determine the prototypical participants of the scenarios. The participants of the same scenario can be considered as semantically interpredictable and may be realized by lexemes belonging to the same derivational paradigm.
