Dissociating task acquisition from expression during learning reveals latent knowledge

Area of Science:

• Neuroscience
• Cognitive Science
• Animal Behavior

Background:

• Cognitive task performance during learning is a common metric for knowledge acquisition.
• However, this performance is known to be influenced by contextual factors, leading to controversy regarding its accuracy as a knowledge measure.

Purpose of the Study:

• To investigate the extent to which testing context influences performance during learning, versus reflecting underlying knowledge.
• To differentiate between knowledge acquisition and knowledge expression in animal models.

Main Methods:

• Mice, rats, and ferrets were trained on various tasks, interleaving reinforced trials with probe trials where reinforcement was omitted.
• A network model was employed to capture the observed results, with learning occurring during reinforced trials and context modulating readout parameters.

Main Results:

• Across all species and tasks, animals performed significantly better during probe trials (without reinforcement) compared to reinforced trials.
• Inter-animal variability in performance was substantially reduced during probe trials.
• Reinforcement feedback was found to be critical for behavioral improvement but paradoxically masked the expression of latent knowledge.

Conclusions:

• The study demonstrates that reinforcement feedback is crucial for learning-related behavioral improvements but can obscure the true expression of underlying knowledge.
• Omitting reinforcement during testing (probing) effectively uncovers latent knowledge.
• Contextual factors, rather than inherent "smartness," are identified as the primary source of individual variability in cognitive task performance during learning.