Conceptually plausible Bayesian inference in interval timing.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychophysics

Background:

• Perception relies on statistical properties of past experiences for optimization in uncertain environments.
• The central tendency effect demonstrates how prior experiences influence current perception, notably in interval timing tasks.
• Current Bayesian observer models often use unimodal distributions to represent priors, which may limit their explanatory power.

Purpose of the Study:

• To critically evaluate the assumptions of traditional unimodal prior models in Bayesian perception.
• To propose a more flexible and plausible model for representing empirical distributions of past experiences.
• To investigate interval timing behavior in healthy adults and individuals with mild cognitive impairment.

Main Methods:

• Developed a novel Bayesian observer model using a mixture of lognormal distributions to represent priors.
• This mixture lognormal model can flexibly mimic various unimodal distributions.
• Fitted the proposed model to published interval timing data from healthy young adults and a clinical population (aged mild cognitive impairment patients and controls).

Main Results:

• The mixture lognormal model demonstrated a superior fit to the behavioral data compared to traditional models.
• The model provided new insights into the mechanisms underlying interval timing in a memory-affected clinical population.
• The enhanced flexibility of the mixture lognormal model allows for better characterization of empirical prior distributions.

Conclusions:

• The mixture lognormal model offers a more flexible and conceptually plausible approach to modeling priors in Bayesian perception.
• This model better explains behavioral data in interval timing tasks across different populations.
• Findings suggest the model can reveal underlying mechanisms, particularly in clinical populations with memory impairments.