A Look around at What Lies Ahead: Prediction and Predictability in Language Processing

Abstract

Traditionally, prediction has been considered an ineffi cient and cognitively expensive processing mechanism in the domain of language comprehension, where there are many possible ways for relaying a single thought, meaning or desire and the chances of mispredicting are accordingly high. Predictive linguistic processing, however, does not seem untenable given its similarity to other neural processing domains that are contextually grounded and appear to implement knowledgeand experience-based mental representations anticipatorily. Here, we examine linguistic prediction from multiple perspectives, ranging from theoretical models that analyze predictability at the level of ambiguity resolution, to experimental evidence primarily from event-related brain potentials (ERPs) that supports a “strong” model of prediction in which items are not just incrementally integrated, but are wholly or featurally pre-activated via accruing mental sentential representations. We also explore possible consequences of a neural language parser (aka, brain) that may be prone to mispredicting, and what electrophysiological evidence for such processing may look like. We conclude by arguing for the importance of investigating such linguistic eff ects as yet another example of a neural system in which probability estimation is inherent, with a proposal to move beyond the debate of whether there is linguistic prediction, toward focusing research eff orts on how pre-activation may occur and what is pre-activated. POTENTIAL BENEFITS OF LINGUISTIC