Internally generated sequences in learning and executing goal-directed behavior.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Goal-directed behavior and decision-making rely on a network of brain structures, including the hippocampus, prefrontal cortex, and striatum.
• The precise neural mechanisms governing these complex cognitive functions remain largely unknown.
• Internally generated sequences, defined as structured multi-neuron firing patterns driven by internal brain dynamics, are hypothesized to be involved.

Purpose of the Study:

• To review the role of internally generated sequences in neural networks.
• To propose a computational model where these sequences function as a component of goal-directed decision systems.
• To elucidate the neural basis of decision-making and goal-directed behavior.

Main Methods:

• Review of neurophysiological studies investigating neural firing patterns.
• Computational modeling to explore the function of internally generated sequences.
• Analysis of sequence generation based on internal brain dynamics, not solely external stimuli.

Main Results:

• Neurophysiological evidence suggests internally generated sequences support memory consolidation, representation augmentation, internal simulation, and information recombination.
• Computational models propose these sequences act as a sampling-based inference engine.
• This engine is suggested to optimize goal acquisition at multiple timescales, including on-line choice, action control, and learning.

Conclusions:

• Internally generated sequences are a critical, yet previously underappreciated, component of the neural systems underlying goal-directed behavior and decision-making.
• Viewing these sequences as a sampling-based inference engine provides a novel framework for understanding how the brain makes optimal choices and learns.
• Further research integrating neurophysiology and computational modeling is warranted to fully understand these internal dynamics.