Specifying Precision in Visual-orthographic Prediction Error Representations for a Better Understanding of Efficient...

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psycholinguistics

Background:

• Efficient visual word recognition is crucial for reading.
• Orthographic prediction error (oPE) representations are hypothesized to underlie this process.
• Predictive coding frameworks suggest optimizing perception by focusing on informative sensory input.

Purpose of the Study:

• To explore alternative implementations of oPE representations.
• To test if increased precision (binary signaling, realistic lexicons) improves models of efficient word recognition.
• To evaluate model performance using behavioral and electrophysiological (EEG) data.

Main Methods:

• Developed a neurocognitive computational model based on predictive coding.
• Implemented and compared two oPE representations: original (less-precise) and alternative (more-precise: binary signaling, frequency-sorted lexicon).
• Evaluated models against behavioral response times and EEG data.

Main Results:

• More precise oPE representations (binary signaling, common word lexicon) best explained behavioral and EEG data at 300 ms post-stimulus onset.
• The original, less-precise oPE representation best explained early brain activation.
• A dynamic adaptation pattern was observed, with initial graded errors converting to binary representations.

Conclusions:

• Efficient word recognition involves a dynamic shift in orthographic prediction error representations.
• Initial graded prediction errors are refined into binary representations for accurate word meaning retrieval.
• This provides a neuro-cognitively plausible account of visual word recognition, highlighting the role of oPE.