Predictive Coding or Just Feature Discovery? An Alternative Account of Why Language Models Fit Brain Data

2022/11/04 by Richard Antonello, Alexander G. Huth · 1 voice · 47 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · Mathematics · #Artificial intelligence #Biomedical Text Mining and Ontologies #Coding (social sciences) #Computer science #Explainable Artificial Intelligence (XAI) #Feature (linguistics) #Linguistics #Machine learning #Mathematics #Natural language processing #Predictive coding #Statistics #Topic Modeling

paper · pdf · doi:10.1162/nol_a_00087

published in Neurobiology of Language 5(1), 1-16

openalex publication_date 2022/11/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/01

Abstract

Many recent studies have shown that representations drawn from neural network language models are extremely effective at predicting brain responses to natural language. But why do these models work so well? One proposed explanation is that language models and brains are similar because they have the same objective: to predict upcoming words before they are perceived. This explanation is attractive because it lends support to the popular theory of predictive coding. We provide several analyses that cast doubt on this claim. First, we show that the ability to predict future words does not uniquely (or even best) explain why some representations are a better match to the brain than others. Second, we show that within a language model, representations that are best at predicting future words are strictly worse brain models than other representations. Finally, we argue in favor of an alternative explanation for the success of language models in neuroscience: These models are effective at predicting brain responses because they generally capture a wide variety of linguistic phenomena.

Citations

• Smoothing and Differentiation of Data by Simplified Least Squares Procedures.

Cited by

• Stimulus dependencies—rather than next-word prediction—can explain pre-onset brain encoding in naturalistic listening designs
• Abstraction Induces the Brain Alignment of Language and Speech Models
• Contextual Semantic Relevance Tracks fMRI BOLD Responses During Naturalistic Speech Comprehension
• Coherence in the brain unfolds across separable temporal regimes
• Distributional Semantics: Meaning Through Culture and Interaction
• Exploring Similarity between Neural and LLM Trajectories in Language Processing
• A systematic evaluation of Dutch large language models’ surprisal estimates in sentence, paragraph and book reading
• A shared model-based linguistic space for transmitting our thoughts from brain to brain in natural conversations
• Divergences between Language Models and Human Brains
• Do Large Language Models Think Like the Brain? Sentence-Level Evidences from Layer-Wise Embeddings and fMRI
• Learning Interpretable Representations Leads to Semantically Faithful EEG-to-Text Generation
• Improve Language Model and Brain Alignment via Associative Memory
• Prediction during language comprehension: what is next? [europepmc]
• Lexical-Semantic Content, Not Syntactic Structure, Is the Main Contributor to ANN-Brain Similarity of fMRI Responses in the Language Network. [europepmc]
• Artificial Neural Network Language Models Predict Human Brain Responses to Language Even After a Developmentally Realistic Amount of Training. [europepmc]
• Computational Language Modeling and the Promise of In Silico Experimentation. [europepmc]
• Scaling laws for language encoding models in fMRI. [europepmc]
• A shared model-based linguistic space for transmitting our thoughts from brain to brain in natural conversations. [europepmc]
• Divergences between Language Models and Human Brains. [europepmc]
• Beyond Letters: Optimal Transport as a Model for Sub-Letter Orthographic Processing. [europepmc]
• SIGNAL: Dataset for Semantic and Inferred Grammar Neurological Analysis of Language. [europepmc]
• Stimulus dependencies-rather than next-word prediction-can explain pre-onset brain encoding in naturalistic listening designs. [europepmc]
• Does the brain really know what word is coming next? [europepmc]
• Evaluating scientific theories as predictive models in language neuroscience [europepmc]

Discussions

• 🏆 Best Paper — Predictive Coding or Just Feature Discovery? An Alternative Account of Why Language Models Fit Brain Data: doi.org/10.1162/nol_... Richard Antonello and @alexanderhuth.bsky.social [bsky, 25 points, 1 comments]

Related

• The neural architecture of language: Integrative modeling converges on predictive processing
• The Predictive Brain: Neural Correlates of Word Expectancy Align with Large Language Model Prediction Probabilities
• Long-range and hierarchical language predictions in brains and algorithms
• Does the brain represent words? An evaluation of brain decoding studies of language understanding
• Stimulus dependencies—rather than next-word prediction—can explain pre-onset brain encoding in naturalistic listening designs