Abstract:Neural network language models (LMs) have been shown to successfully capture complex linguistic knowledge. However, their utility for understanding language acquisition is still debated. We contribute to this debate by presenting a case study where we use LMs as simulated learners to derive novel experimental hypotheses to be tested with humans. We apply this paradigm to study cross-dative generalization (CDG): productive generalization of novel verbs across dative constructions (she pilked me the ball/she pilked the ball to me) -- acquisition of which is known to involve a large space of contextual features -- using LMs trained on child-directed speech. We specifically ask: "what properties of the training exposure facilitate a novel verb's generalization to the (unmodeled) alternate construction?" To answer this, we systematically vary the exposure context in which a novel dative verb occurs in terms of the properties of the theme and recipient, and then analyze the LMs' usage of the novel verb in the unmodeled dative construction. We find LMs to replicate known patterns of children's CDG, as a precondition to exploring novel hypotheses. Subsequent simulations reveal a nuanced role of the features of the novel verbs' exposure context on the LMs' CDG. We find CDG to be facilitated when the first postverbal argument of the exposure context is pronominal, definite, short, and conforms to the prototypical animacy expectations of the exposure dative. These patterns are characteristic of harmonic alignment in datives, where the argument with features ranking higher on the discourse prominence scale tends to precede the other. This gives rise to a novel hypothesis that CDG is facilitated insofar as the features of the exposure context -- in particular, its first postverbal argument -- are harmonically aligned. We conclude by proposing future experiments that can test this hypothesis in children.

The emergence of case matching in discontinuous DPs

The Interpretation of Disjunction in VP Ellipsis: the Case of Mandarin Chinese

Revisiting English Written VP-ellipsis and VP-substitution: a Dependency-Based Analysis

The parameters of case marking and spell out driven movement

The Emergence of Disjunction: A History of Constructionalization in Chinese

Diagnostics for Constituents: Dependency, Constituency, and the Status of Function Words.

NPs, Not DPs: the NP Vs. DP Debate in the Context of Dependency Grammar

Case alternation in lexicalized grammar

Bracketing paradoxes: a dependency grammar analysis in terms of morph catenae

Weak determinism and the computational consequences of interaction

Case and Agreement Variation in Contact: A Multifactorial Investigation of It-Clefts Across World Englishes

Extending the Quantitative Pattern-Matching Paradigm

Generating novel experimental hypotheses from language models: A case study on cross-dative generalization

Existence and Consequences of Dormant Features: Evidence from L2 Acquisition of Chinese Object Ellipsis

Characterizing the Dynamics of Learning in Repeated Reference Games

Episodic versus generic eventualities and nominals

Same Sentences, Different Grammars, Different Brain Responses: An MEG study on Case and Agreement Encoding in Hindi and Nepali Split-Ergative Structures

The correlation between nativelike selection and prototypicality: a multilingual onomasiological case study using semantic embedding

Diachronic Extension Of Linguistic Inventory Mightiness: Evolution Of Directional Resultative Verb-Compounds In Chinese

The natural stability of autonomous morphology