Visual Predictions Operate on Different Timescales.

Area of Science:

• Cognitive neuroscience
• Human perception and action
• Learning and memory

Background:

• Humans operate in dynamic environments requiring adaptive prediction.
• Perceptual and motor skills depend on adjusting internal models to environmental changes.
• Understanding how the brain manages short-term versus long-term learning is crucial.

Purpose of the Study:

• To investigate the interplay between short-term and long-term associative learning in human predictive processing.
• To examine how different timescales of learning influence adaptation to environmental volatility.
• To determine if long-term associations can spontaneously replace short-term ones under changing probabilities.

Main Methods:

• Two electroencephalography (EEG) experiments were conducted.
• Participants viewed sequences of Gabor patches and responded to a target stimulus.
• Stimuli were presented in phases with varying probabilities, including short-term and long-term learning phases, followed by a neutral test phase.

Main Results:

• In the neutral phase, participants initially relied on short-term learned probabilities.
• Participants without prior long-term learning adapted to neutral probabilities over time.
• Participants with prior long-term learning showed spontaneous recovery of those associations, overriding short-term learning.

Conclusions:

• Long-term and short-term associations appear to be stored and controlled independently.
• Short-term associations are transiently used and abandoned when environmental context shifts.
• Spontaneous recovery of long-term associations suggests robust, persistent memory traces.