Low-level predictive inference in reading: the influence of transitional probabilities on eye movements

Area of Science:

• Cognitive Science
• Psycholinguistics
• Computational Linguistics

Background:

• Reading behavior is complex, influenced by various linguistic factors.
• Transitional probability quantifies word-to-word predictability in text.
• Understanding these statistical patterns can illuminate reading mechanisms.

Purpose of the Study:

• To investigate the predictive power of transitional probability on reading behavior.
• To analyze how forward and backward transitional probabilities affect eye movement metrics.
• To determine if transitional probability's influence varies across word classes.

Main Methods:

• Analysis of an eye-movement corpus from newspaper text reading.
• Calculation of forward (P(n/n-1)) and backward (P(n/n+1)) transitional probabilities.
• Statistical modeling to correlate transitional probability with fixation durations and positions.

Main Results:

• Both forward and backward transitional probabilities significantly predicted first fixation and gaze durations.
• Higher transitional probability correlated with shorter fixation times.
• Forward transitional probability showed a slight rightward shift in initial fixation position for predictable words.
• The predictive power of transitional probability was consistent for both function and content words.

Conclusions:

• Transitional probability is a key factor in explaining reading behavior, specifically fixation times.
• The predictability of a word influences how quickly and where readers fixate.
• These findings highlight the role of statistical learning in fluent reading.