The language of science, in particular scientific English, has been investigated in analytic philosophy and applied linguistics, but a cognitive science perspective is still missing. Scientific languages might differ from everyday language in ways that have consequences for ease of processing and learnability, and conversely cognitive considerations may be useful to explain properties of scientific languages. Here, I argue that cognition constrains the variety of argument types that some expressions in the language of science can accept (type polymorphism), keeping processing difficulty in check. I examine some implications of this hypothesis for complement coercion with aspectual verbs, propositional attitude verbs, generalized coordination, and concealed questions. I present the results of a corpus study on type polymorphism for aspectual verbs in biomedical language, corroborating the main hypothesis. Finally, I consider the feasibility of a research program on the cognitive science of scientific language.

Most people engage with music informally or through the standard school curriculum rather than through extensive professional practice, resulting in graded levels of accumulated musical training. In this study, we used cross-validated machine learning on whole-brain structural connectomes to investigate whether individual differences in brain wiring predict variation in musical training across 225 adults with none-to-moderate training levels. Connectomes were weighted by fiber bundle capacity (FBC), a quantitative diffusion MRI metric of structural connectivity. Musical training was quantified using the Gold-MSI Musical Training subscale (F3), a self-reported measure encompassing both formal and informal musical practice. Subcortical white-matter pathways linking thalamus, putamen, and pallidum with sensorimotor cortex carried the strongest predictive signal (r = 0.261; R2 = 0.068; p_perm = 0.003). Furthermore, this connectivity predicted the association between musical training and auditory perception skill, which was present in individuals with stronger subcortical-sensorimotor connectivity (+1 SD: β = 0.192, p = 0.004), but absent in those with weaker connectivity (-1 SD: β = -0.066, p = 0.488). Subcortical-sensorimotor connectivity is, therefore, the strongest structural correlate of graded musical training in non-specialists within the general population, identifying the structural neural substrate on which the training-skill association depends. This work warrants renewed targeted focus on subcortical circuits in models of musical expertise.

A major current approach in theoretical linguistics proposes that there is only a partial, defeasible correspondence between syntax and semantics: meaning and grammar are autonomous, parallel components in the architecture of language. This fact still lacks a satisfactory account. One type of explanation could be provided in terms of computational capability: architectures that separate out meaning and grammar into two processing streams might perform better in tasks that require the system to attend to syntactic vs semantic aspects of sentences. We compare single-stream vs dual-stream transformer architectures, which we train and test on data sets from two event-related brain potential (ERP) studies, with a variety of syntactically and semantically challenging sentences. The single-stream model showed greater sensitivity (the probability that a given sentence s from set S is labelled as belonging to S) with syntactic anomalies, whereas the dual-stream model showed greater sensitivity with semantic anomalies. However, overall, the two models classified sentences with similar accuracy: there were no general advantages for either architecture. Our results are a counterexample to the notion that the functional separation of meaning and grammar in the architecture of human language can be explained in strictly computational terms.

This article addresses the relationship between philosophy and cognitive neuroscience, focusing on possible sources of experimental ideas when theoretical constructs are absent, weak, or only partly developed. We argue that philosophical thought experiments and cases provide one such source, and we examine their role and nature as conduits to actual laboratory experiments. We demonstrate that the structure of thought experiments, in ethics and in the philosophies of language and mind, may lend itself naturally to translations into cognitive neuroscience research. Such translations into experimental tasks involve structural preservation as well as modification of the original cases, with implications for the mutual relevance of the newly obtained results and the original philosophical analyses. Given philosophy's long-held place within the cognitive sciences, this analysis sheds additional light on the issue of cognitive neuroscience itself as a disciplinary enterprise.

This scientific commentary refers to ‘A right frontal network for analogical and deductive reasoning’ by Mole et al. (https://doi.org/10.1093/brain/awaf062).

Compositionality is the thesis that the meanings of complex expressions are derived from the meanings of their constituent parts and syntactic structure. We used a new experimental paradigm to study how syntactic mode of composition (predication vs modification) and lexical semantic properties of the expressions combined (intersective vs subsective vs privative adjectives vs verbs) affect electroencephalographic (EEG) signals. We compared nouns in sentences composing either with a verb (predication) or an adjective (modification) to conditions in which the same noun and adjective, or verb, were separated by a syntactic boundary, preventing composition locally. Event-related potentials (ERPs) and alpha-beta power were modulated primarily by the lexical semantic properties of the words combined, not by the syntactic mode of composition. Our findings raise questions about the neurobiological import of linguistic theories that distinguish cases of composition according to syntactic variables, while abstracting away from the specific semantic contributions of the words involved.

The suite of capacities constituting language involves diverse mental representations, from modality-specific information to levels of formal structure and meaning. In the cognitive science of language, a long-standing puzzle is how these representations "hang together" in an architecture that explains the widest possible range of facts about language. Parallelism is the general hypothesis that correlations exist between representations in the language system (e.g., between syntactic structure and compositional meaning) as well as within the mind (e.g., between word meaning and world knowledge). These correlations are mediated by systems of interfaces, but are always only partial and exhibit varying degrees of systematicity: each type of representation is functionally autonomous, that is, constructed according to specific principles, in addition to simple combinatorial mechanisms that apply across the system. This Topic explores new directions in developing or engaging with this hypothesis, in relation to open issues in several areas of current research in linguistics and cognitive brain science.

Study preregistrations and theory development have been proposed as independent strategies for improving the quality, transparency, robustness, and reproducibility of research. Yet, a discussion of how theory and preregistration could interact and complement each other towards improved science is currently lacking. Here, we argue that hypothesis preregistration could stimulate theory development by serving various roles, depending on the availability or quality of theories and hypotheses in a field at any given time. We suggest that, in appropriate conditions, preregistration can increase the quality of hypotheses before they are tested, with indirect beneficial effects on theory development. In fields where theories are less developed or agreed upon, or are lacking altogether, hypothesis preregistration can nudge researchers to improve and share their hypotheses, engaging their community and facilitating accumulation and development of hypotheses. As a field’s theories and hypotheses become more advanced and better understood, hypothesis preregistration can become less important: the theory itself can function as a public repository of hypotheses and can constrain methodological aspects of research. We explore possible relations and synergies between hypothesis preregistration and theory development in a range of scenarios. We conclude with a discussion of implications and recommendations for researchers and meta-scientists.

Two distinct computational approaches provide opportunities for bringing together different theories of cognition. Two distinct computational approaches provide opportunities for bringing together different theories of cognition.

In sentences such as John began the book, the complement noun phrase, lexically denoting an entity, is interpreted as denoting an event. This is known in linguistics as complement coercion: the event associated with the verb is not overtly expressed but can be recovered from the meanings of other constituents, context and world knowledge. We investigate whether language models (LMs) can exploit sentence structure and compositional meaning to recover plausible events in complement coercion. For the first time, we tested different LMs in Norwegian, a low-resource language with high syntactic variation in coercion constructions across aspectual verbs. Results reveal that LMs struggle with retrieving plausible events and with ranking them above less plausible ones. Moreover, we found that LMs do not exploit the compositional properties of coercion sentences in their predictions.

In complement coercion sentences, like John began the book, a covert event (e.g., reading) may be recovered based on lexical meanings, world knowledge, and context. We investigate how context influences coercion interpretation performance for 17 language models (LMs) in Norwegian, a low-resource language. Our new dataset contained isolated coercion sentences (context-neutral), plus the same sentences with a subject NP that suggests a particular covert event and sentences that have a similar effect but that precede or follow the coercion sentence. LMs generally benefit from contextual enrichment, but performance varies depending on the model. Models that struggled in context-neutral sentences showed greater improvements from contextual enrichment. Subject NPs and precoercion sentences had the largest effect in facilitating coercion interpretation.

ROSE is a rare example of a neurocomputational model of language that attempts, and partly manages, to align a formal theory of syntax and parsing with an oscillations-based 'neural code' that could implement the required operations. ROSE successfully reconciles hierarchical and predictive syntactic processing, but I argue that models of language in the brain should make room for the possibility that meaning may also be derived in the absence of any syntactic computation, be it hierarchical or predictive.

In neurolinguistics and the neurobiology of language, processing models that parallelize meaning and grammar have acquired theoretical and empirical support over more modular theories. Yet, parallel models too should account for serial, sequential, or blocking effects of one type of representation on others. In this study, we used ERPs to assess whether and how the mental lexicon restricts the applicability of morphosyntactic operations, and how those, in turn, constrain on-line meaning composition. The stimuli were Norwegian sentences with the form ‘N V Adj’. The adjective was either correctly or incorrectly inflected for gender or number relative to the noun, and the noun was either a real word or a pseudoword built around a pseudoroot. ERPs show that agreement only applies between an adjective and a noun that contains a real lexical root, and that lexical meanings are only composed for correctly inflected words. We could not find agreement effects for pseudonouns or differences between gender and number features. Our results suggest that some grammatical processes may draw from and depend on lexical storage, in particular of lexical roots, and that compositional semantic processes may depend on the well-formedness of the outputs of such grammatical processes.

Various notions of plausibility are used in cognitive science to argue for or against the “goodness of theories.” However, plausibility remains poorly understood and difficult to analyze. We review debates in the philosophy of science on uses of plausibility in the assessment of novel scientific theories as well as recent attempts to formalize, reform, or eliminate specific notions of plausibility. Although these discussions highlight important concerns behind plausibility claims, they fail to identify viable notions of plausibility that are sufficiently different from other criteria of “good theory,” such as prior probability or external coherence. We survey uses of plausibility in linguistics and cognitive science, confirming that plausibility is often a proxy for other criteria of good theory. We argue that the need remains for concepts of plausibility that can be employed to assess the quality of proposals at the early stages of theory development when other criteria are not yet applicable. We identify two such notions: one relating to formal constraints on theories and another capturing initial epistemic consensus, if not necessarily convergence on the truth, about the target system in a community of inquiry. We briefly assess the specificity and added value of these notions of plausibility relative to other criteria for good theory.

Why are some individuals more musical than others? Neither cognitive testing nor classical localizationist neuroscience alone can provide a complete answer. Here, we test how the interplay of brain network organization and cognitive function delivers graded perceptual abilities in a distinctively human capacity. We analyze multimodal magnetic resonance imaging, cognitive, and behavioral data from 200+ participants, focusing on a canonical working memory network encompassing prefrontal and posterior parietal regions. Using graph theory, we examine structural and functional frontoparietal network organization in relation to assessments of musical aptitude and experience. Results reveal a positive correlation between perceptual abilities and the integration efficiency of key frontoparietal regions. The linkage between functional networks and musical abilities is mediated by working memory processes, whereas structural networks influence these abilities through sensory integration. Our work lays the foundation for future investigations into the neurobiological roots of individual differences in musicality.

In a recent paper, Mandelkern Linzen (2024) - henceforth M L - address the question of whether language models' (LMs) words refer. Their argument draws from the externalist tradition in philosophical semantics, which views reference as the capacity of words to "achieve 'word-to-world' connections". In the externalist framework, causally uninterrupted chains of usage, tracing every occurrence of a name back to its bearer, guarantee that, for example, 'Peano' refers to the individual Peano (Kripke 1980). This account is externalist both because words pick out referents 'out there' in the world, and because what determines reference are coordinated linguistic actions by members of a community, and not individual mental states. The "central question to ask", for M L, is whether LMs too belong to human linguistic communities, such that words by LMs may also trace back causally to their bearers. Their answer is a cautious "yes": inputs to LMs are linguistic "forms with particular histories of referential use"; "those histories ground the referents of those forms"; any occurrence of 'Peano' in LM outputs is as causally connected to the individual Peano as any other occurrence of the same proper name in human speech or text; therefore, occurrences of 'Peano' in LM outputs refer to Peano. In this commentary, we first qualify M L's claim as applying to a narrow class of natural language expressions. Thus qualified, their claim is valid, and we emphasise an additional motivation for that in Section 2. Next, we discuss the actual scope of their claim, and we suggest that the way they formulate it may lead to unwarranted generalisations about reference in LMs. Our critique is likewise applicable to other externalist accounts of LMs (e.g., Lederman Mahowald 2024; Mollo Milliere 2023). Lastly, we conclude with a comment on the status of LMs as members of human linguistic communities.

Formal analysis of the minimal computational complexity of verification algorithms for natural language quantifiers implies that different classes of quantifiers demand the engagement of different cognitive resources for their verification. In particular, sentences containing proportional quantifiers, e.g. “most”, provably require a memory component, whereas non-proportional quantifiers, e.g. “all”, “three”, do not. In an ERP study, we tested whether previously observed differences between these classes were modulated by memory load. Participants performed a picture-sentence verification task while they had to remember a string of 2 or 4 digits to be compared to a second string at the end of a trial. Relative to non-proportional quantifiers, proportional quantifiers elicited a sentence-internal sustained negativity. Additionally, an interaction between Digit-Load and Quantifier-Class was observed at the sentence-final word. Our results suggest that constraints on cognitive resources deployed during human sentence processing and verification are of the same nature as formal constraints on abstract machines.

The intergenerational stability of auditory symbolic systems, such as music, is thought to rely on brain processes that allow the faithful transmission of complex sounds. Little is known about the functional and structural aspects of the human brain which support this ability, with a few studies pointing to the bilateral organization of auditory networks as a putative neural substrate. Here, we further tested this hypothesis by examining the role of left-right neuroanatomical asymmetries between auditory cortices. We collected neuroanatomical images from a large sample of participants (nonmusicians) and analyzed them with Freesurfer's surface-based morphometry method. Weeks after scanning, the same individuals participated in a laboratory experiment that simulated music transmission: the signaling games. We found that high accuracy in the intergenerational transmission of an artificial tone system was associated with reduced rightward asymmetry of cortical thickness in Heschl's sulcus. Our study suggests that the high-fidelity copying of melodic material may rely on the extent to which computational neuronal resources are distributed across hemispheres. Our data further support the role of interhemispheric brain organization in the cultural transmission and evolution of auditory symbolic systems.