Prediction suppression and surprise enhancement in monkey inferotemporal cortex.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Processing

Background:

• Neuronal responses in the visual system are influenced by learned associations and predictions.
• Predictive coding models propose neurons signal prediction errors, crucial for updating internal models of the world.

Purpose of the Study:

• To investigate how neuronal visual responsiveness in monkey inferotemporal cortex (area TE) is affected by sequential image learning.
• To determine whether observed prediction-modulated firing arises from prediction suppression or surprise enhancement.

Main Methods:

• Monkeys were trained with sequential image pairs, establishing predictive relationships between leading and trailing images.
• Neuronal firing rates were recorded and compared under prediction-confirming, prediction-violating, and prediction-neutral conditions.

Main Results:

• Neurons exhibited significantly weaker responses to trailing images when they confirmed predictions (prediction suppression).
• Neurons showed stronger responses to the same images when they violated predictions (potential surprise enhancement).
• Evidence strongly supported prediction suppression, with limited support for surprise enhancement.

Conclusions:

• Monkey inferotemporal neurons display prediction-modulated firing, but the signal appears unsigned, unlike the signed prediction errors in standard predictive coding models.
• The findings suggest a visual system mechanism that emphasizes unpredicted events, offering an alternative to traditional predictive coding frameworks.