[Paper Review] Language Hierarchization Provides the Optimal Solution to Human Working Memory Limits

LuYao Chen, W. Gao+3 more|arXiv (Cornell University)|Jan 6, 2026

Neurobiology of Language and Bilingualism0 citations

TL;DR

The paper argues that hierarchical language processing optimally fits human working memory constraints, using a likelihood framework where tehta_MLE equals the mean of units; simulations and natural language validation show hierarchical structure outperforms linear processing and aligns with development patterns in children.

ABSTRACT

Language is a uniquely human trait, conveying information efficiently by organizing word sequences in sentences into hierarchical structures. A central question persists: Why is human language hierarchical? In this study, we show that hierarchization optimally solves the challenge of our limited working memory capacity. We established a likelihood function that quantifies how well the average number of units according to the language processing mechanisms aligns with human working memory capacity (WMC) in a direct fashion. The maximum likelihood estimate (MLE) of this function, tehta_MLE, turns out to be the mean of units. Through computational simulations of symbol sequences and validation analyses of natural language sentences, we uncover that compared to linear processing, hierarchical processing far surpasses it in constraining the tehta_MLE values under the human WMC limit, along with the increase of sequence/sentence length successfully. It also shows a converging pattern related to children's WMC development. These results suggest that constructing hierarchical structures optimizes the processing efficiency of sequential language input while staying within memory constraints, genuinely explaining the universal hierarchical nature of human language.

Motivation & Objective

• Motivate the question of why human language is hierarchical.
• Develop a likelihood-based framework to quantify alignment between language processing units and working memory capacity (WMC).
• Show that hierarchical processing constrains tehta_MLE under human WMC better than linear processing.
• Validate with simulations and natural language data.
• Link findings to developmental patterns in children's WMC.

Proposed method

• Define a likelihood function that measures alignment between expected units under language processing and human WMC.
• Derive that the maximum likelihood estimate tehta_MLE equals the mean of units.
• Perform computational simulations with symbol sequences to compare hierarchical vs. linear processing.
• Validate the approach using analyses of natural language sentences.
• Assess how tehta_MLE behaves as sequence length grows and across developmental stages.

Experimental results

Research questions

• RQ1 Can hierarchical language processing maximize alignment with human working memory capacity?
• RQ2 What is the mathematical form and interpretation of tehta_MLE in this framework?
• RQ3 Does hierarchical processing constrain tehta_MLE more effectively than linear processing as sequence length increases?
• RQ4 Do the results generalize to developmental patterns in children’s working memory capacity?
• RQ5 Do analyses of natural language data corroborate findings from simulations?

Key findings

• Hierarchical processing far surpasses linear processing in constraining tehta_MLE under human WMC as sequence length increases.
• The maximum likelihood estimate tehta_MLE is the mean of units under the proposed framework.
• Computational simulations and validation analyses of natural language sentences support the superiority of hierarchical processing for memory-constrained language understanding.
• There is a converging pattern in tehta_MLE related to children’s WMC development.
• The results suggest hierarchical structure optimizes processing efficiency within memory limits, explaining universal hierarchical language.