Rational Adaptation in Lexical Prediction: The Influence of Prediction Strength

Area of Science:

• Cognitive Psychology
• Psycholinguistics
• Computational Neuroscience

Background:

• Processing unexpected words incurs costs, particularly when initial predictions are strong.
• Comprehenders rationally adapt their predictive strategies based on context.

Purpose of the Study:

• To investigate if the disconfirmation of strong predictions triggers adaptation mechanisms.
• To examine how the frequency of disconfirmed predictions influences subsequent prediction costs.

Main Methods:

• Two experiments conducted in Hebrew and English using speeded congruency judgments.
• Participants evaluated two-word phrases with either highly constraining or non-constraining first words.
• Manipulation of the proportion of disconfirmed predictions in highly constraining contexts.

Main Results:

• Confirmed the processing costs associated with disconfirming strong predictions.
• Demonstrated a reduction in these costs when participants encountered a high proportion of disconfirmed strong predictions.
• Showed that participants adjust prediction strength when strong predictions are frequently discouraged.

Conclusions:

• The study provides evidence for adaptive prediction mechanisms in language processing.
• Findings suggest that prediction costs are modulated by the perceived likelihood of prediction success.
• A Bayesian adaptation model successfully accounts for trial-by-trial behavioral data.