Language and Linguistics Compass最新文献

Large Language Models (LLMs) have dramatically transformed the AI landscape. They can produce remarkable fluent text and exhibit a range of natural language understanding and generation capabilities. This article explores how LLMs might be used for sociolinguistic research and, conversely, how sociolinguistics can contribute to the development of LLMs. It argues that both areas of research will benefit from a thoughtful, engaging collaboration. Sociolinguists are not merely end users of LLMs; they have a crucial role to play in the development of LLMs.

This article originally published online on 28 September 2024 with incorrect spellings. The article has been since updated to comply with standard US English spelling.

Sound symbolism refers to associations between language sounds (i.e., phonemes) and particular properties (e.g., certain shapes). For example, phonemes like /m/ are associated with roundness, while phonemes like /k/ are associated with spikiness. In this paper I review the accumulating evidence that different instances of sound symbolism can be observed as patterns in real words in existing lexicons (e.g., /m/ occurring more frequently in words for round things). The properties examined include shape, size, texture, valence and arousal. Such effects are an instance of both iconicity (words whose forms resemble their meanings) and systematicity (largescale patterns in the forms of related words in a language). I also discuss open questions on the topic, including how such patterns emerge, and their effects on language processing.

Truthmaker semantics is a non-classical logical framework that has recently garnered significant interest in philosophy, logic, and natural language semantics. It redefines the propositional connectives and gives rise to more fine-grained entailment relations than classical logic. In its model theory, truth is not determined with respect to possible worlds, but with respect to truthmakers, such as states or events. Unlike possible worlds, these truthmakers may be partial; they may be either coherent or incoherent; and they are understood to be exactly or wholly relevant to the truth of the sentences they verify. Truthmaker semantics generalises collective, fusion-based theories of conjunction; alternative-based theories of disjunction; and nonstandard negation semantics. This article provides a gentle introduction to truthmaker semantics aimed at linguists; describes applications to various natural language phenomena such as imperatives, ignorance implicatures, and negative events; and discusses its similarities and differences to related frameworks such as event semantics, situation semantics, alternative semantics, and inquisitive semantics.

How can modern neural networks like language models be useful to the field of language acquisition, and more broadly cognitive science, if they are not a priori designed to be cognitive models? As developments towards natural language understanding and generation have improved leaps and bounds, with models like GPT-4, the question of how they can inform our understanding of human language acquisition has re-emerged. As such, it is critical to examine how in practice linking hypotheses between models and human learners can be safely established. To address these questions, we propose a model taxonomy, including four modelling approaches, each having differing goals, from exploratory hypothesis generation to hypothesis differentiation and testing. We show how the goals of these approaches align with the overarching goals of science and linguistics by connecting our taxonomy to the realist versus instrumentalist approaches in philosophy of science. We survey recent work having adopted each of our modelling approaches and address the importance of computational modelling in language acquisition studies.

Children acquiring language must learn a variety of words mapping on to different kinds of concepts. Typically, word learning accounts focus on how children strategically acquire words for entities (nouns) and events (verbs). Often underrepresented are word for properties (adjectives). While adjectives, like verbs, may trail behind many nouns in early production, and may not always be required to refer or express truth-conditional meaning, they are essential for distinguishing among entities and events in the world, and communicating clearly in disciplines such as language arts and science. Moreover, nearly every language has the grammatical category of adjectives. It is therefore critical to know what challenges children face in acquiring adjectives, and what supports their acquisition. Here, I outline three main aspects that, together, render the process of learning adjectives both vexing and linguistically intriguing. First, there is cross-linguistic variation in how languages around the world treat words for properties. Second, the syntactic position of adjectives varies across and within languages, and is often correlated with subtle differences in meaning. Third, within the category of adjectives, a wide range of meanings may be encoded. I review findings across these areas that document children's successes and challenges with these aspects of adjective distribution and meaning. To counter these challenges, I end by covering three effective and empirically documented strategies for adjective learning, which rely on attending to key information in both the linguistic and extralinguistic context. I close with a brief discussion of prospects for future research on this topic.

Psycholinguistic studies using the event-related brain potential (ERP) technique have found both N400 and P600 effects for gender stereotype violations. The finding of a P600 effect for this type of world knowledge violation is surprizing given that this component is traditionally associated with syntactic violations. In this paper, we set out to systematically analyse design- and task patterns of ERP studies investigating gender stereotype violations. Based on our review, we propose a scheme that predicts a P600 effect for gender stereotype violations for stimuli comprising entire sentences, and specifically when the gender stereotype serves to establish coherence in inferences (i.e., expecting a specific referent gender or a stereotype-appropriate behaviour of the sentence subject based on the gender stereotype information). We predict an N400 effect for gender stereotype information that does not serve to establish coherence in inferences, as well as in priming paradigms. By extension, our predictive scheme suggests that the N400 may reflect a hybrid process of semantic retrieval and integration, while the P600 may reflect a cognitive process of error monitoring, and mental revision. Our study can aid in the interpretation of previous findings and inform future studies investigating gender stereotyping.