Visual Prediction Error Spreads Across Object Features in Human Visual Cortex.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Visual Cognition

Background:

• Visual cognition relies on contextual expectations.
• Distinct neural signatures exist for expected versus unexpected stimuli.
• How the brain combines multiple concurrent object feature expectations is unknown.

Purpose of the Study:

• Investigate how unexpected object features affect concurrent processing of other expected features.
• Test models of independent processing, surprise spread, or feature unbinding.
• Understand neurocomputational principles of multifeature expectations.

Main Methods:

• Human functional magnetic resonance imaging (fMRI).
• Behavioral task independently manipulating color and motion expectations.
• Computational modeling of multifeature expectations.

Main Results:

• Behavioral and fMRI data supported a model of prediction error mixing.
• Surprise in one feature spreads to other features, rendering the entire object unexpected.
• Objects appear to be the unit of selection for predictive vision.

Conclusions:

• The brain combines multiple concurrent expectations at the object level.
• Prediction errors for one feature impact the processing of other features.
• This reveals object-level predictive processing in vision.