As people increasingly interact with large language models (LLMs), a critical question emerges: do humans process language differently when communicating with an LLM versus another human? While there is good evidence that people adapt comprehension based on their expectations toward their interlocutor in human–human interaction, human–computer interaction research suggests the adaptation to machines is often suspended until expectation violation occurs. We conducted two event-related potential experiments examining Chinese sentence comprehension, measuring neural responses to semantic and syntactic anomalies attributed to an LLM or a human. Experiment 1 revealed reduced N400 but larger P600 responses to semantic anomalies in LLM-attributed text than human-attributed one, suggesting participants anticipated semantic errors yet required increased composition/integration efforts. Experiment 2 showed enhanced P600 responses to LLM-attributed than human-attributed syntactic anomalies, reflecting greater reanalysis or integration difficulty in the former than in the latter. Notably, neural responses to LLM-attributed semantic anomalies (but not syntactic anomalies) were further modulated by participants’ belief about humanlike knowledge in LLMs, with a larger N400 and a smaller P600 in participants with stronger belief of humanlike knowledge in LLMs. These findings provide the first neurocognitive evidence that people develop mental models of LLM capabilities and adapt neural processing accordingly, offering theoretical insights aligned with multidisciplinary frameworks and practical implications for designing effective human–AI communication systems.

In this article, we develop a theory of the form and interpretation of nonrestrictive nominal appositives (NAPs) by combining two recent syntactic and pragmatic approaches. Following Ott (2016), we assume that NAPs are independent elliptical speech acts, which are linearly interpolated into their host sentences in production. Building on insights in Onea 2016, we argue that NAPs make their pragmatic contribution as short answers to discourse-structuring Potential Questions. We show how these two assumptions combine to yield a comprehensive theory of NAPs that captures their central syntactic, semantic, and pragmatic properties and furthermore sheds light on the mechanisms that govern their linear interpolation.

This article proposes a syntax and a semantics for intonation in English and some related languages. The semantics is ‘surface-compositional’, in the sense that syntactic derivation constructs information-structural logical form monotonically, without rules of structural revision, and without autonomous rules of ‘focus projection’. This is made possible by the generalized notion of syntactic constituency afforded by combinatory categorial grammar (CCG)—in particular, the fact that its rules are restricted to string-adjacent type-driven combination. In this way, the grammar unites intonation structure and information structure with surface-syntactic derivational structure and Montague-style compositional semantics, even when they deviate radically from traditional surface structure.

The article revises and extends earlier CCG-based accounts of intonational semantics, grounding hitherto informal notions like ‘theme’ and ‘rheme’ (a.k.a. ‘topic’ and ‘comment’, ‘presupposition’ and ‘focus’, etc.) and ‘background’ and ‘contrast’ (a.k.a. ‘given’ and ‘new’, ‘focus’, etc.) in a logic of speaker/hearer supposition andupdate, usingaversion of Rooth's alternative semantics. A CCG grammar fragment is defined that constrains language-specific intonation and its interpretation more narrowly than previous attempts.

We present evidence for the influence of semantics on the order of subject, object, and verb in Brazilian Sign Language (Libras) sentences. While some have argued for a prevailing pattern of SVO in Libras, we find a strong tendency for this order in sentences that do not presuppose the existence of the verb's object, but not in sentences that do, which instead favor SOV. These findings are coherent with those of a recent study on gesture. We argue that the variable influence of the relevant predicates is particularly salient in sign languages, due to the iconic nature of the visual modality.

This article scrutinizes the diachrony of relativizers and complement clause subordinators in Russian, Polish, and Czech. Historical morphology indicates a development from agreeing relative pronouns via noninflected relativizers to complement clause subordinators. This concurs with recent findings on Germanic (Axel-Tober 2017), but contradicts more traditional proposals that derive subordinators from demonstratives. The respective syntactic reanalyses are demonstrated on diachronic Slavic corpus data. Moreover, a quantitative comparison of sixteenth- to seventeenth- century East Slavic texts with and without West Slavic interference suggests that the use of kotoryj ‘which’ as a relative pronoun spread into Russian as an inner-Slavic contact-induced change.

Macaulay and Brice (1997:798) surveyed example sentences in eleven syntax textbooks published from 1969–1994 and found that virtually all of the authors ‘favor male-gendered NPs as subjects and agents, and regularly stereotype both genders’. In this article, we address the question of whether constructed example sentences in more recent textbooks show similar gender bias. We present an analysis of six syntax textbooks published from 2005–2017, from which we randomly sampled 200 example sentences each. We find that the gender skew and stereotypes reported in 1997 are still present today. Male-gendered arguments are almost twice as frequent as female-gendered ones, and more likely to occur as subjects and agents. In addition, example sentences often perpetuate gender stereotypes. We discuss some broader implications and potential interventions to prevent the implicit perpetuation of gender biases in linguistic materials.

Joe Pater's (2019) target article calls for greater interaction between neural network research and linguistics. I expand on this call and show how such interaction can benefit both fields. Linguists can contribute to research on neural networks for language technologies by clearly delineating the linguistic capabilities that can be expected of such systems, and by constructing controlled experimental paradigms that can determine whether those desiderata have been met. In the other direction, neural networks can benefit the scientific study of language by providing infrastructure for modeling human sentence processing and for evaluating the necessity of particular innate constraints on language acquisition.

While published linguistic judgments sometimes differ from the judgments found in large-scale formal experiments with naive participants, there is not a consensus as to how often these errors occur nor as to how often formal experiments should be used in syntax and semantics research. In this article, we first present the results of a large-scale replication of the Sprouse et al. 2013 study on 100 English contrasts randomly sampled from Linguistic Inquiry 2001-2010 and tested in both a forced-choice experiment and an acceptability rating experiment. Like Sprouse, Schütze, and Almeida, we find that the effect sizes of published linguistic acceptability judgments are not uniformly large or consistent but rather form a continuum from very large effects to small or nonexistent effects. We then use this data as a prior in a Bayesian framework to propose a small n acceptability paradigm for linguistic acceptability judgments (SNAP Judgments). This proposal makes it easier and cheaper to obtain meaningful quantitative data in syntax and semantics research. Specifically, for a contrast of linguistic interest for which a researcher is confident that sentence A is better than sentence B, we recommend that the researcher should obtain judgments from at least five unique participants, using at least five unique sentences of each type. If all participants in the sample agree that sentence A is better than sentence B, then the researcher can be confident that the result of a full forced-choice experiment would likely be 75% or more agreement in favor of sentence A (with a mean of 93%). We test this proposal by sampling from the existing data and find that it gives reliable performance.

This study describes a change in which relative clause extraposition is in the process of being lost in English, Icelandic, French, and Portuguese. This current change in progress has never been observed before, probably because it is so slow that it is undetectable without the aid of multiple diachronic parsed corpora (treebanks) with time depths of over 500 years each. Building on insights from Kiparsky (1995), the study shows that the change may date as far back as the innovation of Proto-Germanic and Proto-Romance relative clauses, as these varieties differentiated from Proto-Indo-European. It also shows that the unusually slow speed of the change is due to partial specialization of the construction along the dimension of prosodic weight, following the argument made at greater length in Fruehwald & Wallenberg 2016. Finally, the change is shown to have important consequences for the syntax of extraposition, supporting the adjunction analysis of Culi- cover and Rochemont (1990). The article also discusses the implications of Sauerland's (2003) analysis of English relative clauses, and while modern English data supports his analysis, the diachronic extraposition data is not yet fine-grained enough to bear on the ‘raising’ analysis of relatives in general. This is identified as an important question for further research on this change.

There is evidence to suggest that finiteness marking on verbs and subject-auxiliary inversion are related phenomena in English. In contrast, in Spanish there is evidence consistent with finiteness marking on verbs and the apparently similar phenomenon of subject-verb inversion being unrelated. In both cases, most of the evidence adduced comes from adult acceptability judgments and other adult psycholinguistic work. In the present article, we present evidence from child English and Spanish that supports the interrelatedness of finiteness and inversion in English, but not Spanish. Specifically, we show that child English speakers who are in the optional infinitive stage have variable judgments of finiteness that are predictive of their independently measured inversion judgments. In contrast, no such relationship appears to hold between the variance in finiteness and the variance in inversion judgments of child Spanish speakers of the same preschool age. We take this to be novel confirmation of the hypothesis that subject-auxiliary inversion in English takes finite tense as a necessary condition. In Spanish, in contrast, it appears that it is not necessary for a finite verb to move to the left periphery for subject-verb inversion.

There has been a recent spate of work on recursion as a central design feature of language. This short report points out that there is little evidence that unlimited recursion, understood as center-embedding, is typical of natural language syntax. Nevertheless, embedded pragmatic construals seem available in every language. Further, much deeper center-embedding can be found in dialogue or conversation structure than can be found in syntax. Existing accounts for the ‘performance’ limitations on center-embedding are thus thrown into doubt. Dialogue materials suggest that center-embedding is perhaps a core part of the human interaction system, and is for some reason much more highly restricted in syntax than in other aspects of cognition.

Classical Sanskrit is well known for making extensive use of compounding. I argue, within a lexicalist framework, that the major rules of compounding in Sanskrit can be most appropriately characterized in syntactic, not morphological, terms. That is, Classical Sanskrit ‘compounds’ are in fact very often syntactic phrases. The syntactic analysis proposed captures the fact that compound formation is closer to a morphological process than other aspects of syntax, and so permits some acknowledgment of the gradient nature of the word-phrase divide, even within a strictly lexicalist theory.

The birthdate of both generative linguistics and neural networks can be taken as 1957, the year of the publication of foundational work by both Noam Chomsky and Frank Rosenblatt. This article traces the development of these two approaches to cognitive science, from their largely autonomous early development in the first thirty years, through their collision in the 1980s around the past-tense debate (Rumelhart & McClelland 1986, Pinker & Prince 1988) and their integration in much subsequent work up to the present. Although this integration has produced a considerable body of results, the continued general gulf between these two lines of research is likely impeding progress in both: on learning in generative linguistics, and on the representation of language in neural modeling. The article concludes with a brief argument that generative linguistics is unlikely to fulfill its promise of accounting for language learning if it continues to maintain its distance from neural and statistical approaches to learning.

In many different domains of language acquisition, there exists an apparent learnability problem to which innate knowledge of some aspect of Universal Grammar (UG) has been proposed as a solution. The present article reviews these proposals in the core domains of (i) identifying syntactic categories, (ii) acquiring basic morphosyntax, (iii) structure dependence, (iv) subjacency, and (v) the binding principles. We conclude that, in each of these domains, the innate UG-specified knowledge posited does not, in fact, simplify the task facing the learner.

The lexicon divides into parts of speech (or lexical categories), and there are cross-cutting regularities (features). These two dimensions of analysis take us a long way, but several phenomena elude us. For these the term ‘split’ is used extensively (‘case split’, ‘split agreement’, and more), but in confusingly different ways. Yet there is a unifying notion here. I show that a split is an ADDITIONAL PARTITION, whether in the part-of-speech inventory or in the feature system. On this base an elegant typology can be constructed, using minimal machinery. The typology starts from four external relations (government, agreement, selection, and anti-government), and it specifies four types of split within each (sixteen possibilities in all). This typology (i) highlights less familiar splits, from diverse languages, and fits them into the larger picture; (ii) introduces a new relation, anti-government, and documents it; (iii) elucidates the complexities of multiple splits; and (iv) clarifies what exactly is split, which leads to a sharpening of our analyses and applies across different traditions.

This article examines the distribution of gender in arguments in example sentences in contemporary linguistics publications. Prior studies have shown that example sentences in syntax textbooks systematically underrepresent women and perpetuate gender stereotypes (Macaulay & Brice 1994, 1997, Pabst et al. 2018). Here we examine example sentences in articles published over the past twenty years in Language, Linguistic Inquiry, and Natural Language & Linguistic Theory and find striking similarities to this prior work. Among our findings, we show a stark imbalance of male (N = 12,117) to female (N = 5,571) arguments, where male-gendered arguments are more likely to be subjects, and female-gendered arguments nonsubjects. We show that female-gendered arguments are more likely to be referred to using a kinship term, to exhibit positive emotions, and to be the object of affection, whereas male-gendered arguments are more likely to have occupations, to exhibit negative emotions, and to perpetrate violence. We show that this pattern has remained stable, with little change, over the course of the twenty years that we examine, leading up to the present day. We conclude with a brief discussion of possible remedies and suggestions for improvement.

This article reports on the successful implementation of the DISCOVERY METHOD in an undergraduate Generative Syntax course. Rather than reading textbooks on syntactic analysis and doing exercises to apply what was read, students embark on a nine-stage learning cycle consisting of interactive, guided, and independent work. They engage with problem sets to build their own analyses of syntactic phenomena step by step. This method—meant to resemble the work of real-life scientists—allows students to hypothesize, experiment, and build knowledge from trial and error (Kolb 1984, Prince 2004). To measure the effectiveness of our pedagogy, we analyzed course evaluations submitted from 2012 to the present. Learner feedback shows that they find the assignments and activities helpful, that the course challenges them to think and learn, and that they consider the weekly write-ups demanding, yet highly beneficial for their learning.

We argue for an extension of the proposal that grammars are in part shaped by processing systems. Hawkins (2014) and others who have advanced this idea focus primarily on parsing. Our extension focuses on production, and we use that to explore explanations for certain subject/object asymmetries in extraction structures. The phenomenon we examine, which we term the mirror asymmetry, runs in opposite directions for within-clause and across-clause (long-distance) extraction, showing a preference for subject extraction in the former and for object extraction in the latter. We review several types of evidence suggesting that the mirror asymmetry and related phenomena are best explained by an account of the formation of grammars that assigns an important role to properties of sentence planning in production.

I examine the behavior of rare and other frequency adjectives in the tough-construction (TC). Due to the effects of a heretofore overlooked semantic selectional restriction, such adjectives have not generally been recognized as grammatical in the TC. I show here that they do occur grammatically in this construction when the relevant selectional restriction is satisfied. Specifically, as it does in non-TC sentences, rare in the TC requires that its subject be kind-denoting, a requirement not imposed on the embedded-clause gap position whose reference the TC matrix subject controls. In this, TCs with rare exemplify a previously unattested selectional and thematic asymmetry in the construction. On their face, the facts appear to argue strongly in favor of treating the rare-TC matrix subject as a thematic argument of the TC matrix predicate, an intriguing and challenging prospect given the fact that such an analysis has been roundly (though not universally) rejected for canonical TCs. Instead, I take the prima facie counterintuitive position that the rare-TC matrix subject is not a thematic argument of the TC matrix predicate; I argue that rare-TCs are thematically and syntactically identical to canonical TCs in this respect. I propose that the kind-denotation requirement for rare-TC matrix subjects is imposed indirectly, through the interaction of a selectional restriction on the infinitival argument of rare and the Agree calculus (Chomsky 2000, 2001) that identifies the TC matrix subject with the embedded gap position whose reference it controls (Rezac 2004, 2006). Beyond its contribution to our theoretical understanding of the perennially thorny TC, the present study constitutes, to the best of my knowledge, the first detailed empirical investigation of the behavior of adjectives like rare in the TC.