Words have meaning. Sentences also have meaning, but their meaning is different in kind from any collection of the meanings of the words they contain. I discuss two puzzles related to this difference. The first is how the meanings of the parts of a sentence combine to give rise to a unified sentential meaning, as opposed to a mere collection of disparate meanings (UP1). The second is why the formal ontology of linguistic meaning changes when grammatical structure is built up (...) (UP2). For example, the meaning of a sentence is a proposition evaluable for truth and falsity. In contrast, a collection of the meanings of its parts does not constitute a proposition and is not evaluable for truth. These two puzzles are closely related, since change in formal ontology is the clearest sign of the unity of meaning. The most popular strategy for answering them is taking the meanings of the parts as abstractions from primitive sentence meanings. However, I argue that, given plausible psychological constraints, sentence meanings cannot be taken as explanatory primitives. Drawing on recent work in Generative Grammar and its philosophy, I suggest that the key to both unity questions is to distinguish strictly between lexical and grammatical meaning. The latter is irreducible and determines how lexical content is used in referential acts. I argue that these referential properties determine a formal ontology, which explains why and how formal ontology changes when grammatical structure is built up (UP2). As for UP1, I suggest that, strictly speaking, lexical meanings never combine. Instead, whenever grammar specifies a formal ontology for the lexical meanings entering a grammatical derivation, further lexical (or phrasal) meanings can only specify aspects of this recursive grammatical process. In this way, contemporary grammatical theory can be used to address old philosophical problems. (shrink)

Arguments from the Logical Problem of Language Acquisition suggest that since linguistic experience provides few negative data that would falsify overgeneral grammatical hypotheses, innate knowledge of the principles of Universal Grammar must constrain learners hypothesis formulation. Although this argument indicates a need for domain-specific constraints, it does not support their innateness. Learning from mostly positive data proceeds unproblematically in virtually all domains. Since not every domain can plausibly be accorded its own special faculty, the probative value of the argument in (...) the linguistic case is dubious. In ignoring the holistic and probablistic nature of theory construction, the argument underestimates the extent to which positive data can supply negative evidence and hence overestimates the intractability of language learning in the absence of a dedicated faculty. While nativism about language remains compelling, the alleged Logical Problem contributes nothing to its plausibility and the emphasis on the Problem in the recent acquisition literature has been a mistake. (shrink)

Cognitive Science, Volume 45, Issue 11, November 2021.

As modern deep networks become more complex, and get closer to human‐like capabilities in certain domains, the question arises as to how the representations and decision rules they learn compare to the ones in humans. In this work, we study representations of sentences in one such artificial system for natural language processing. We first present a diagnostic test dataset to examine the degree of abstract composable structure represented. Analyzing performance on these diagnostic tests indicates a lack of systematicity in representations (...) and decision rules, and reveals a set of heuristic strategies. We then investigate the effect of training distribution on learning these heuristic strategies, and we study changes in these representations with various augmentations to the training set. Our results reveal parallels to the analogous representations in people. We find that these systems can learn abstract rules and generalize them to new contexts under certain circumstances—similar to human zero‐shot reasoning. However, we also note some shortcomings in this generalization behavior—similar to human judgment errors like belief bias. Studying these parallels suggests new ways to understand psychological phenomena in humans as well as informs best strategies for building artificial intelligence with human‐like language understanding. (shrink)

Rumelhart and McClelland's chapter about learning the past tense created a degree of controversy extraordinary even in the adversarial culture of modern science. It also stimulated a vast amount of research that advanced the understanding of the past tense, inflectional morphology in English and other languages, the nature of linguistic representations, relations between language and other phenomena such as reading and object recognition, the properties of artificial neural networks, and other topics. We examine the impact of the Rumelhart and McClelland (...) model with the benefit of 25 years of hindsight. It is not clear who “won” the debate. It is clear, however, that the core ideas that the model instantiated have been assimilated into many areas in the study of language, changing the focus of research from abstract characterizations of linguistic competence to an emphasis on the role of the statistical structure of language in acquisition and processing. (shrink)

According to a “selective” (as opposed to “instructive”) model of human language capacity, people come to know more than they experience. The discrepancy between experience and eventual capacity (the “poverty of the stimulus”) is bridged by genetically provided information. Hence any hypothesis about the linguistic genotype (or “Universal Grammar,” UG) has consequences for what experience is needed and what form people's mature capacities (or “grammars”) will take. This BBS target article discusses the “trigger experience,” that is, the experience that actually (...) affects a child's linguistic development. It is argued that this must be a subset of a child's total linguistic experience and hence that much of what a child hears has no consequence for the form of the eventual grammar. UG filters experience and provides an upper bound on what constitutes the triggering experience. This filtering effect can often be seen in the way linguistic capacity can change between generations. Children only need access to robust structures of minimal (“degree-0”) complexity. Everything can be learned from simple, unembedded “domains” (a grammatical concept involved in defining an expression's logical form). Children do not need access to more complex structures. (shrink)