Evidence Hub

Predictive coding unifies neural and behavioral signatures of contextual facilitation

Nour Eddine S, Yacovone A, Kuperberg GR

2026

Abstract

During comprehension, words that are primed or predictable are processed more quickly and elicit smaller N400 responses than words that are less consistent with their preceding context. Although these behavioral and neural effects are both commonly interpreted as evidence of facilitated processing, they have typically been explained within different computational frameworks that impose seemingly conflicting constraints: behavioral facilitation requires more activity to accumulate over expected representations, allowing a decision threshold to be reached earlier, whereas N400 facilitation requires less neural activity to be generated overall. Here, we show that an implemented model of predictive coding provides a unified mechanistic account of both phenomena and, in doing so, resolves three longstanding puzzles. First, it explains how facilitated processing can simultaneously produce faster behavioral responses and reduced neural activity: expected representations accumulate activity more rapidly in state units while generating less activity in error units. Second, an asymmetry between top-down priors and bottom-up evidence allows context to pre-activate multiple lexical candidates without inducing premature competition or misprediction costs. Third, it reconciles apparently conflicting claims about the timing of contextual facilitation by showing that behavioral and neural measures can reflect different phases along a single continuous trajectory of lexico-semantic…

Read the paper

Related studies

Go deeper in Evidence Hub

Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.

Open this study Sign up free