Limits to Language Prediction: Findings From Diverse Populations

Abstract

For a model in cognitive science to adequately explain cognitive processes across different populations, empirical findings from diverse participant groups are essential. This paper selectively reviews studies that investigated prediction in different populations and discusses what they reveal about the mechanisms and role of language prediction. To build a generalizable model of prediction and to understand the role of prediction in language comprehension and learning, it is crucial to understand both the brain's potential and limits to prediction. Findings from WEIRD (Western, Educated, Industrialized, Rich, and Democratic) populations have demonstrated the brain's remarkable ability to predict upcoming language. However, findings from diverse populations (e.g., second‐language speakers, heritage speakers, low‐literacy adults, children and older adults, adults from different socioeconomic backgrounds) highlight important limits to language prediction and suggest that prediction is modulated by individual differences beyond what has been found in WEIRD populations. I suggest that these individual differences can be explained by the currently proposed prediction models without assuming qualitatively different mechanisms for different populations.