Trivial Vocabulary Bans Improve LLM Reasoning More Than Deep Linguistic Constraints

TL;DR

Trivial vocabulary bans, especially those that are shallow and impose minimal conceptual disruption, significantly enhance reasoning in language models more than complex linguistic constraints.

Contribution

The study demonstrates that simple, shallow vocabulary restrictions outperform deep linguistic constraints in improving language model reasoning, challenging previous assumptions.

Findings

Neutral filler-word ban improved reasoning accuracy by 6.7 percentage points.

E-Prime had the smallest positive effect on reasoning performance.

Shallow constraints outperform deep linguistic modifications in enhancing reasoning.

Abstract

A previous study reported that E-Prime (English without the verb "to be") selectively altered reasoning in language models, with cross-model correlations suggesting a structural signature tied to which vocabulary was removed. I designed a replication with active controls to test the proposed mechanism: cognitive restructuring through specific vocabulary-cognition mappings. The experiment tested five conditions (unconstrained control, E-Prime, No-Have, elaborated metacognitive prompt, neutral filler-word ban) across six models and seven reasoning tasks (N=15,600 trials, 11,919 after compliance filtering). Every prediction from the cognitive restructuring hypothesis was disconfirmed. All four treatments outperformed the control (83.0%), including both active controls predicted to show null effects. The neutral filler-word ban, banning words like "very" and "just" with no role in logical inference, produced the largest improvement (+6.7 pp), while E-Prime produced the smallest (+3.7 pp). The four conditions ranked in perfect inverse order of theoretical depth. The cross-model correlation signature did not replicate (mean r=0.005). These results are consistent with a simpler mechanism: any constraint that forces a model off its default generation path acts as an output regularizer, improving reasoning by disrupting fluent but shallow response patterns. The shallowest constraints work best because they impose monitoring load with minimal conceptual disruption. I present these findings as a case study in discovery through disconfirmation.