Hemispheric asymmetries in the split-fovea model of semantic processing

Area of Science:

• Computational neuroscience
• Cognitive psychology
• Psycholinguistics

Background:

• The brain's semantic processing of written words involves complex neural computations.
• Hemispheric asymmetries in language processing are well-documented but their computational basis remains debated.

Purpose of the Study:

• To model semantic processing of single English words using neural networks.
• To investigate the influence of visual field splitting on orthography-to-semantics mapping.
• To explore hemispheric asymmetries in word meaning and priming.

Main Methods:

• Developed neural network models simulating left and right hemisphere word processing.
• Implemented foveal splitting to mimic visual field input.
• Trained models on four-letter English words, considering frequency and position.
• Analyzed differences in processing ambiguous words and priming effects.

Main Results:

• Models exhibited processing asymmetries comparable to human behavioral studies.
• Cueing effects for ambiguous words differed based on presentation side (left vs. right).
• Priming effects for related words were stronger in the left input compared to the right.
• These asymmetries stemmed from coarser semantic coding in the right model half.

Conclusions:

• Neural network models can replicate observed hemispheric asymmetries in word reading.
• Coarser coding in the right hemisphere model may explain weaker priming effects.
• Informational asymmetries in English word structure contribute to processing differences.
• These findings offer a potential computational origin for behavioral data in single-word reading.