How are visual words represented? Insights from EEG-based visual word decoding, feature derivation and image...

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• The neural basis of reading, particularly visual-orthographic processing, is an active area of research.
• Understanding the fine-grained structure of neural representations for reading is crucial but remains challenging.
• Previous studies have explored the cortical regions involved in reading but lacked detailed insights into representational structure.

Purpose of the Study:

• To investigate if electroencephalography (EEG) patterns can decode and reconstruct internal representations of visually presented words.
• To examine the temporal dynamics and neurocomputational principles underlying word recognition.
• To identify individual differences in visual-orthographic representations during reading.

Main Methods:

• Utilized electroencephalography (EEG) data from healthy adults viewing visually presented words.
• Applied machine learning techniques to decode word identity and reconstruct visual representations from EEG patterns.
• Analyzed the spatiotemporal structure of EEG signals to understand the timing of neural processes.

Main Results:

• Accurate above-chance word classification and image reconstruction were achieved using EEG data.
• The temporal profile of these decoding and reconstruction abilities showed an early onset around 100 ms, peaking at 170 ms.
• Reconstruction accuracy was significantly explained by visual-orthographic properties of the words, and individual differences were observed.

Conclusions:

• EEG-based decoding and reconstruction of word representations are feasible.
• The findings elucidate the early dynamics and neurocomputational basis of visual word recognition.
• This research highlights the potential of EEG for understanding the neural mechanisms of reading and individual variability.