According to much of theoretical linguistics, a fair amount of our linguistic knowledge is innate. One of the best-known (and most contested) kinds of evidence for a large innate endowment is the argument from the poverty of the stimulus (APS). An APS obtains when human learners systematically make inductive leaps that are not warranted by the linguistic evidence. A weakness of the APS has been that it is very hard to assess what is warranted by the linguistic evidence. Current artificial neural networks appear to offer a handle on this challenge, and a growing literature has started to explore the potential implications of such models to questions of innateness. We focus on Wilcox, Futrell, and Levy’s (2024) use of several different networks to examine the available evidence as it pertains to wh -movement, including island constraints. WFL conclude that the (presumably linguistically neutral) networks acquire an adequate knowledge of wh -movement, thus undermining an APS in this domain. We examine the evidence further, looking in particular at parasitic gaps and across-the-board movement, and argue that current networks do not succeed in acquiring or even adequately approximating wh -movement from training corpora roughly the size of the linguistic input that children receive. We also show that the performance of one of the models improves considerably when the training data are artificially enriched with instances of parasitic gaps and across-the-board movement. This finding suggests, albeit tentatively, that the networks’ failure when trained on natural, unenriched corpora is due to the insufficient richness of the linguistic input, thus supporting the APS.

The typology of the logical vocabulary in natural language is highly skewed. In the domain of logical connectives, AND and OR are often lexicalized, lexicalizations of NOR are less common and tend to be structurally complex, and no other logical connective is ever lexicalized. Existing accounts fail to fully derive this major crosslinguistic pattern, and moreover resort to otherwise unwarranted assumptions. The goal of this article is to provide an account that is less stipulative and has wider empirical coverage than previous accounts, based on a novel notion of communicative stability. Using a model of a rational speaker, we observe that attested languages are stable languages—that is, languages in which the optimal message for a speaker to choose when they want to convey a particular state they are in is not affected by which states they take to be more likely and which ones less. We argue that Stability can explain both why AND and OR are the only simple connectives lexicalized and why NOR is the only complex connective lexicalized.

We develop an evaluation metric for Optimality Theory that allows a learner to induce a lexicon and a phonological grammar from unanalyzed surface forms. We wish to model aspects of knowledge such as the English-speaking child’s knowledge that the aspiration of the first segment of k h æt is predictable and the French-speaking child’s knowledge that the final l of table ‘table’ is optional and can be deleted while that of parle ‘speak’ cannot. We show that the learner we present succeeds in obtaining this kind of knowledge and is better equipped to do so than other existing learners in the literature.

Agreement and concord involve mismatches between the position where a syntactic or semantic contribution is made and where it is expressed morphologically. Such mismatches are often taken to involve either movement or the spreading and realization of features. I examine a number of superficially different instances of this kind of mismatch that seem to resist an account based on realization and movement alone. I argue that the patterns involved should receive a unified account in terms of licensors, elements subject to a condition of structural economy that associate indirectly with semantically contentful heads through features that those heads spread and that require a c-commanding licensor.