Minimal cross-trial generalization in learning the representation of an odor-guided choice task

Area of Science:

• Cognitive science
• Animal behavior
• Computational neuroscience

Background:

• Task representation can vary between individuals, even when experiencing identical events.
• Experimenters often assume subjects share their intended task representation, particularly in animal studies lacking explicit instructions.
• Understanding how animals represent tasks is crucial for interpreting their behavior.

Purpose of the Study:

• To investigate how rats represent an odor-guided choice task.
• To determine if rats generalize learning from forced-choice trials to free-choice trials based on odor cues.
• To assess whether rats adopt a parsimonious task representation.

Main Methods:

• Rats performed an odor-guided choice task with forced and free-choice trials.
• Reinforcement-learning models with different task representations were fitted to individual rat behavior.
• The degree to which rats generalized across trial types (used a parsimonious representation) was quantified.

Main Results:

• Most rats did not acquire the hypothesized parsimonious task representation.
• Behavioral data indicated a lack of generalization across odors in the choice task.
• Model comparison showed that simpler representations did not fit most individual rat data.

Conclusions:

• Rats may not automatically adopt the most parsimonious task representation, even with extensive experience.
• It is essential to empirically test and validate assumed task representations in animal research.
• Observed animal behavior should not be presumed to align with the experimenter's designed task structure without formal assessment.