Being right matters: Model-compliant events in predictive processing.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psychology

Background:

• Prediction errors (PE) are known to drive learning via internal model adaptation.
• The role of model-compliant events, like checkpoints (CP), in predictive processing remains less understood.
• CPs may serve as on-line reference points for evaluating internal model validity, especially in uncertain situations.

Purpose of the Study:

• To investigate the functional relationship between checkpoints (CP) and prediction errors (PE) using electroencephalography (EEG).
• To determine if CPs and PEs elicit similar neural responses, specifically the P3b component, indicating internal model recourse.
• To examine if PEs elicit a distinct mismatch signal (N400) compared to CPs.

Main Methods:

• Utilized electroencephalography (EEG) to record brain activity during a serial pattern detection task.
• Analyzed event-related potentials (ERPs), focusing on the P3b and N400 components.
• Compared neural responses to checkpoints (CP), prediction errors (PE), and standard stimuli (STD).
• Conducted behavioral subgroup analyses to explore the influence of initial cue learning on processing these events.

Main Results:

• Event-related findings supported the hypothesis that both CP and PE elicit a joint P3b component, suggesting internal model recourse.
• The N400 component was observed for PE, consistent with a mismatch signal, when compared to CP.
• Behavioral data indicated that reliance on initial cue learning might hinder the processing of informative reference points like CPs, impacting flexible model adaptation.

Conclusions:

• Checkpoints (CP) function as model-compliant reference points for internal model evaluation, similar to prediction errors (PE).
• While PEs instigate model adaptation, CPs are crucial for ongoing model assessment.
• The effective utilization of reference points like CPs for flexible cognitive adaptation may be modulated by prior learning experiences and uncertainty levels.