Abrupt Scene Onsets and Gradually Emerging Scene Information Produce Distinct EEG Decoding Dynamics.

Area of Science:

• Cognitive Neuroscience
• Neuroimaging
• Visual Perception

Background:

• Multivariate analysis of M/EEG data often relies on discrete stimulus onsets.
• This approach typically shows high decoding performance initially, followed by a sustained lower level.
• Naturalistic visual input lacks clear, discrete onsets, posing challenges for traditional analysis.

Purpose of the Study:

• To investigate time-resolved EEG decoding of natural scene processing during gradual visual transitions.
• To compare EEG decoding for stimuli with artificial onsets versus those emerging naturalistically.
• To understand how stimulus presentation context influences neural decoding patterns.

Main Methods:

• Created video sequences with smooth scene transitions using blended panoramas and moving apertures.
• Compared EEG decoding for artificially presented scenes (first scene) and naturalistically emerging scenes (second scene).
• Utilized deep neural networks to realign category-diagnostic content for the second scene.

Main Results:

• Robust EEG category decoding observed for artificially presented scenes (from 60 ms), with a distinct peak structure.
• Markedly weaker decoding and no discernible peak structure for naturalistically emerging scenes.
• Classifiers trained on initial scenes showed diffuse generalization to later scenes, not a hierarchical pattern.

Conclusions:

• Time-resolved EEG decoding is sensitive to the stimulus presentation context.
• Traditional decoding patterns from discrete onsets may not generalize to continuous visual input and gradual transitions.
• The way visual information is presented significantly impacts the analysis of neural responses.