From Novel Combinations to Informative Absence: Structural permission, communicative opportunity, and form choice in language learning

Language learners must infer both which novel expressions are grammatically licensed and when an expression's continued absence becomes evidence against it. We develop a generative theory uniting these capacities through counterfactual expression opportunities, decomposing utterance generation into structural permission, communicative opportunity, and form choice. First, we show that shared predictive continuations enable learners to infer structural permission for unseen word–construction pairings across thirty artificial grammars (mean AUC 0.945 after 120 sentences). Second, to explain how learners interpret subsequent omissions under hierarchical uncertainty, we derive a joint evidence law governed by context diversity and within-context repetition, alongside an exact finite-sample context-refinement inequality. This generative coupling yields two key predictions: positive observations of related expressions calibrate expected usage to sharpen subsequent negative evidence, and candidates with stronger initial permission can undergo steeper posterior declines when accompanied by higher expected use (reversing their initial ordering after six omissions in a fixed-rate example). Third, we prove that whereas raw occurrence counts confound semantic availability with form selection, scene-level opportunity observations identify both factors. Empirically, reanalyzing artificial-language experiments with adults and children shows meaning-conditioned restriction and finds that judgments of restricted combinations improve predictions of held-out production across 399 participants (+0.036 nats per production). Furthermore, naturalistic child-directed and child speech across eleven families and six corpora exhibits pervasive source-level usage clustering beyond lexical, number, and dependency-role controls, empirically grounding hierarchical rate uncertainty. Finally, cross-linguistic modeling of causative judgments across five languages demonstrates that structural compatibility and calibrated usage probabilities draw on distinct distributional, semantic, and discriminative cues. Together, these findings establish how the organization and calibration of communicative opportunities govern the evidential value of linguistic absence.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24360441
Primary Topic
Language Development and Disorders
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

From Novel Combinations to Informative Absence: Structural permission, communicative opportunity, and form choice in language learning

Yanzhe Zhao
Zenodo (CERN European Organization for Nuclear Research)
Language Development and Disorders
preprint

From Novel Combinations to Informative Absence: Structural permission, communicative opportunity, and form choice in language learning

Yanzhe Zhao
preprint en

Abstract

Language learners must infer both which novel expressions are grammatically licensed and when an expression's continued absence becomes evidence against it. We develop a generative theory uniting these capacities through counterfactual expression opportunities, decomposing utterance generation into structural permission, communicative opportunity, and form choice. First, we show that shared predictive continuations enable learners to infer structural permission for unseen word–construction pairings across thirty artificial grammars (mean AUC 0.945 after 120 sentences). Second, to explain how learners interpret subsequent omissions under hierarchical uncertainty, we derive a joint evidence law governed by context diversity and within-context repetition, alongside an exact finite-sample context-refinement inequality. This generative coupling yields two key predictions: positive observations of related expressions calibrate expected usage to sharpen subsequent negative evidence, and candidates with stronger initial permission can undergo steeper posterior declines when accompanied by higher expected use (reversing their initial ordering after six omissions in a fixed-rate example). Third, we prove that whereas raw occurrence counts confound semantic availability with form selection, scene-level opportunity observations identify both factors. Empirically, reanalyzing artificial-language experiments with adults and children shows meaning-conditioned restriction and finds that judgments of restricted combinations improve predictions of held-out production across 399 participants (+0.036 nats per production). Furthermore, naturalistic child-directed and child speech across eleven families and six corpora exhibits pervasive source-level usage clustering beyond lexical, number, and dependency-role controls, empirically grounding hierarchical rate uncertainty. Finally, cross-linguistic modeling of causative judgments across five languages demonstrates that structural compatibility and calibrated usage probabilities draw on distinct distributional, semantic, and discriminative cues. Together, these findings establish how the organization and calibration of communicative opportunities govern the evidential value of linguistic absence.

Zenodo (CERN European Organization for Nuclear Research)
Tianjin University (CN)
Language Development and Disorders
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon