Animal Learning in a Multidimensional Discrimination Task as Explained by Dimension-Specific Allocation of Attention

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Animal Behavior

Background:

• Reinforcement learning explains how rewards reinforce actions, but complex multi-dimensional stimuli pose challenges for understanding cognitive processes.
• Existing models effectively explain simple learning but struggle with stimuli varying across multiple features.

Purpose of the Study:

• To investigate the cognitive mechanisms underlying learning and decision-making in multi-sensory discrimination tasks with complex stimuli.
• To develop and validate a computational model that accounts for learning and decision-making under shifting relevance of stimulus dimensions.

Main Methods:

• Adapted an intra-dimensional/extra-dimensional set-shifting paradigm for rats using multi-sensory (spatial, olfactory, visual) cues.
• Proposed a weighted attention model (WAM) where learning rules are applied to features within each dimension, and decisions are based on weighted averages.
• Estimated WAM parameters from rat behavior and compared its performance against an alternative feature-combination learning model.

Main Results:

• The weighted attention model (WAM) significantly outperformed the alternative model in explaining rat behavior.
• Estimated decision weights revealed an experience-based bias, where previously relevant dimensions retained high weights even after their relevance shifted.
• This bias in decision weights provides an explanation for observed performance deficits during extra-dimensional shifts.

Conclusions:

• The weighted attention model (WAM) offers a robust framework for understanding learning and decision-making with complex, multi-sensory stimuli.
• Decision weights quantified by the WAM can illuminate experience-based biases that influence learning and performance in shifting environments.
• This research provides insights into the flexibility and limitations of cognitive processes in adapting to changing environmental contingencies.