Neurophysiological Evidence for Cognitive Map Formation during Sequence Learning.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Human sequence learning involves parsing statistical regularities.
• Cognitive map theories suggest learning relies on representations of latent spaces.
• Temporal discounting influences sequence learning and resembles spatial navigation models.

Purpose of the Study:

• Investigate neural underpinnings of cognitive map formation during sequence learning.
• Link behavioral models of sequence learning with neural activity.
• Explore the role of temporal discounting in creating space-like representations.

Main Methods:

• Behavioral modeling of human reaction times in a sequential task.
• Intracranial electroencephalography (iEEG) recordings.
• Fitting a model with temporal discounting as a free parameter to estimate latent space representations.

Main Results:

• Neural activity, particularly in the temporal lobe, reflects individual estimates of the latent space.
• Low-dimensional neural activity representations facilitate feature separation, similar to spatial navigation.
• Sequence learning occurs rapidly (within ~500 trials) and is influenced by latent space structure and individual temporal discounting.

Conclusions:

• Neural activity supports the formation of cognitive maps through temporal discounting in sequence learning.
• Findings bridge behavioral sequence learning models and neural correlates.
• Highlights domain-general cognitive map principles applicable to diverse learning scenarios.