Decoding the temporal representation of facial expression in face-selective regions.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Facial expression recognition is crucial for social interaction.
• Rapid neural computations in distributed face-selective regions underlie this ability.

Purpose of the Study:

• To investigate the temporal dynamics of neural responses in face-selective regions during facial expression discrimination.
• To determine the nature of information processed in these regions.

Main Methods:

• Magnetoencephalography (MEG) was used to measure neural responses to facial expressions.
• Time-resolved decoding analyzed neural activity in specific face-selective regions (LO-faces, IP-faces, FG-faces, pSTS-faces).
• Convolutional Neural Networks (CNNs) performed similarity analyses between neural responses and stimulus representations.

Main Results:

• Facial expressions were classified from neural data starting at approximately 100-150 ms post-stimulus onset.
• LO-faces and IP-faces demonstrated higher classification accuracy compared to FG-faces and pSTS-faces.
• Neural responses in LO-faces and IP-faces correlated more with deep emotional category representations than low-level image features.

Conclusions:

• LO-faces and IP-faces are critical for early and accurate facial expression discrimination.
• These regions process abstract emotional information rather than just visual features.
• A link exists between behavioral performance and neural activity in LO-faces and IP-faces.