Component-Level Lesioning of Language Models Reveals Clinically Aligned Aphasia Phenotypes

2026-01-27

AI Summary

This study proposes a clinically informed, component-level perturbation approach to systematically simulate the linguistic impairments characteristic of Broca’s and Wernicke’s aphasia within both modular Mixture-of-Experts (MoE) and dense Transformer-based large language models. By selectively lesioning specific functional components and quantitatively evaluating outcomes using the Western Aphasia Battery (WAB) and Aphasia Quotient (AQ), the work establishes, for the first time, a systematic mapping between aphasia subtypes and internal model components. Results demonstrate that targeted perturbations significantly outperform random lesions in inducing aphasia-like language degradation. Notably, in MoE architectures, the relationship between lesioned components and resulting behavioral phenotypes is more localized and interpretable, underscoring the potential of modular large language models as platforms for investigating the cognitive mechanisms of aphasia and testing rehabilitation hypotheses.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Search and Retrieval-Augmented AI: Large language models for search
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Large language models (LLMs) increasingly exhibit human-like linguistic behaviors and internal representations that they could serve as computational simulators of language cognition. We ask whether LLMs can be systematically manipulated to reproduce language-production impairments characteristic of aphasia following focal brain lesions. Such models could provide scalable proxies for testing rehabilitation hypotheses, and offer a controlled framework for probing the functional organization of language. We introduce a clinically grounded, component-level framework that simulates aphasia by selectively perturbing functional components in LLMs, and apply it to both modular Mixture-of-Experts models and dense Transformers using a unified intervention interface. Our pipeline (i) identifies subtype-linked components for Broca's and Wernicke's aphasia, (ii) interprets these components with linguistic probing tasks, and (iii) induces graded impairments by progressively perturbing the top-k subtype-linked components, evaluating outcomes with Western Aphasia Battery (WAB) subtests summarized by Aphasia Quotient (AQ). Across architectures and lesioning strategies, subtype-targeted perturbations yield more systematic, aphasia-like regressions than size-matched random perturbations, and MoE modularity supports more localized and interpretable phenotype-to-component mappings. These findings suggest that modular LLMs, combined with clinically informed component perturbations, provide a promising platform for simulating aphasic language production and studying how distinct language functions degrade under targeted disruptions.

Problem

• aphasia
• language models
• lesioning
• language production
• computational simulation

Innovation

• component-level lesioning
• aphasia simulation
• large language models
• Mixture-of-Experts
• clinically aligned phenotypes