Metastable attractors explain the variable timing of stable behavioral action sequences.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Self-initiated actions display significant timing variability despite well-learned sequences.
• The neural basis for this mix of reliability and stochasticity in motor control remains unclear.

Purpose of the Study:

• To investigate how neural circuits generate both reliable sequencing and variable timing in self-initiated actions.
• To identify a potential neural network motif underlying this behavior.

Main Methods:

• Trained rats to perform stereotyped action sequences.
• Recorded neural ensemble activity in the secondary motor cortex (M2).
• Utilized hidden Markov models to link neural activity patterns to actions and analyzed network dynamics.

Main Results:

• Identified metastable attractors in neural activity, characterized by reliable sequential structure and variable transition timing.
• Demonstrated that a network of coupled high-dimensional recurrent and low-dimensional feedforward networks can generate these metastable attractors.
• Found that correlated variability within this mesoscale feedback loop drives transitions between attractors, matching empirical M2 recordings.

Conclusions:

• A novel mesoscale network motif involving reciprocally coupled recurrent and feedforward networks, leveraging correlated variability, can explain the observed reliability and stochasticity in action timing.
• This motif provides a mechanism for naturalistic animal behavior and motor control.