Learning about environmental geometry: an associative model

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Animal Behavior

Background:

• K. Cheng's (1986) hypothesis proposed a specialized geometric module for spatial learning, independent of other spatial information.
• Empirical observations of blocking and overshadowing failures in geometry learning align with this modular view.
• Existing models often treat spatial feature learning and geometric learning separately.

Purpose of the Study:

• To propose and evaluate an operant model for spatial learning that integrates the competition between learning local features and geometric properties.
• To explain how the interplay between feature and geometry learning can account for diverse experimental findings in spatial navigation.
• To investigate the influence of enclosure shape and feature type on the dynamics of spatial learning.

Main Methods:

• Development of an operant conditioning model based on the Rescorla-Wagner framework.
• Simulation of learning processes where the associative strength of cues at a location dictates behavioral choices and reward contingencies.
• Testing the model's ability to reproduce established findings from spatial learning experiments in both dry arenas and water mazes.

Main Results:

• The model demonstrates that competitive learning between local features and geometry can produce apparent potentiation, blocking, or independence effects.
• These effects are shown to be contingent upon the specific enclosure shape and the nature of the local features present.
• The model successfully replicates numerous empirical findings from both dry arena and water maze experiments, validating its predictive power.

Conclusions:

• Spatial learning involves a dynamic competition between the acquisition of local environmental features and the understanding of geometric layouts.
• The proposed model provides a unified framework for understanding seemingly disparate outcomes in spatial learning paradigms.
• This competitive learning mechanism offers a parsimonious explanation for complex navigation behaviors observed in various experimental settings.