Probabilistic word pre-activation during language comprehension inferred from electrical brain activity

Area of Science:

• Cognitive Neuroscience
• Psycholinguistics
• Computational Linguistics

Background:

• Anticipatory cognitive processes are observed across species.
• The role of specific word anticipation in human language processing remains debated.

Purpose of the Study:

• To investigate whether the human brain pre-activates upcoming words during real-time language comprehension.
• To determine if this pre-activation is graded and predictable.

Main Methods:

• Utilized a phonological regularity of English indefinite articles ('a' vs. 'an').
• Employed event-related brain potential (ERP) recordings.
• Correlated brain activity with offline word probability estimates.

Main Results:

• Evidence suggests that readers' brains pre-activate individual words in a graded manner.
• The degree of pre-activation correlates with the estimated probability of a word completing a sentence fragment.
• Demonstrates anticipatory word processing in real-time language comprehension.

Conclusions:

• The human brain actively predicts upcoming words during reading.
• World knowledge and contextual cues are used to estimate the likelihood of future words.
• Supports a predictive coding framework for language processing.