Saccade selection when reward probability is dynamically manipulated using Markov chains.

Area of Science:

• Cognitive Neuroscience
• Behavioral Economics
• Computational Neuroscience

Background:

• Markov chains model transitions between states, commonly applied to behaviors like foraging.
• Understanding how primates learn and utilize Markov chain probabilities in behavior is limited.
• Saccadic eye movements are suitable for studying probability-based learning due to their link with information processing.

Purpose of the Study:

• To investigate whether humans can learn and incorporate Markov chain probabilities into saccade target selection.
• To explore if Markov chains offer a confoud-free method for studying probability's influence on saccades.
• To assess how humans adapt behavior when reward probabilities and magnitudes vary dynamically.

Main Methods:

• Human participants performed a saccade target selection task where reward probabilities were assigned via a Markov chain.
• Stimuli were selected by saccade, followed by feedback on reward.
• Sessions involved 200-600 trials, with variations in uniform rewards and reward magnitudes.

Main Results:

• Subjects learned to match stimulus selection frequency to reward probability in uniform reward conditions.
• When informed about the Markov chain, participants increased selection of high-probability rewards, approaching reward maximization.
• In variable reward magnitude conditions, subjects balanced probability and magnitude, preferring stimuli with higher expected value.

Conclusions:

• Markov chains can dynamically assign probabilities that are effectively learned and exploited by humans during saccade selection.
• This research demonstrates rapid human adaptation to probabilistic reward structures.
• The findings highlight the role of expected value in guiding behavior under complex probabilistic environments.