A cost minimisation and Bayesian inference model predicts startle reflex modulation across species

Area of Science:

• Behavioral neuroscience
• Computational neuroscience
• Evolutionary biology

Background:

• Rapid defensive reflexes are crucial for survival across species.
• Startle reflex modulation by environmental states, such as fear-potentiated startle, is well-documented.
• Existing models fail to explain the full range of startle modulation or its evolutionary basis.

Purpose of the Study:

• To develop a normative cost-minimization model of startle reflex modulation.
• To provide a unified explanation for startle modifications across various environmental states.
• To offer testable predictions for future experimental research.

Main Methods:

• Development of a normative cost-minimization model based on Bayesian optimality principles.
• Mathematical modeling of startle reflex modulation incorporating rewards, punishments, and attention.
• Comparison of model predictions with existing empirical data.

Main Results:

• The Bayesian model successfully predicts observed startle reflex modulation across diverse states, including fear, reward, and instructed attention.
• The model's mathematical framework generates novel, experimentally testable predictions.
• Analysis suggests a specific neural implementation approximating optimal Bayesian computations.

Conclusions:

• Startle reflex modulation can be understood within the framework of Bayesian decision theory and predictive coding.
• An adaptive perspective, grounded in computational principles, is essential for interpreting defensive behaviors.
• This work unifies diverse findings on startle modulation and provides a foundation for future research into defensive behaviors.