Plausibility as Commonsense Reasoning: Humans Succeed, Large Language Models Do not

TL;DR

This study investigates whether large language models process syntactic ambiguity in a human-like way by testing their attachment preferences in Turkish relative clauses, revealing they do not utilize plausibility cues as effectively as humans.

Contribution

The paper introduces a novel cross-linguistic diagnostic for evaluating plausibility-driven syntactic attachment in language models using Turkish relative clauses.

Findings

Humans show a strong, plausibility-driven attachment preference.

Language models exhibit weak, unstable, or reversed plausibility effects.

Turkish RC attachment serves as a useful diagnostic beyond standard benchmarks.

Abstract

Large language models achieve strong performance on many language tasks, yet it remains unclear whether they integrate world knowledge with syntactic structure in a human-like, structure-sensitive way during ambiguity resolution. We test this question in Turkish prenominal relative-clause attachment ambiguities, where the same surface string permits high attachment (HA) or low attachment (LA). We construct ambiguous items that keep the syntactic configuration fixed and ensure both parses remain pragmatically possible, while graded event plausibility selectively favors High Attachment vs.\ Low Attachment. The contrasts are validated with independent norming ratings. In a speeded forced-choice comprehension experiment, humans show a large, correctly directed plausibility effect. We then evaluate Turkish and multilingual LLMs in a parallel preference-based setup that compares matched HA/LA continuations via mean per-token log-probability. Across models, plausibility-driven shifts are weak, unstable, or reversed. The results suggest that, in the tested models, plausibility information does not guide attachment preferences as reliably as it does in human judgments, and they highlight Turkish RC attachment as a useful cross-linguistic diagnostic beyond broad benchmarks.