Hidden Markov models reveal behavioral state dynamics in depth-related locomotion in mice

Abstract

Understanding how mice process and respond to visual depth cues is crucial for studying visual perception, yet traditional behavioral analyses often miss key aspects of this process, such as the dynamic transitions between behavioral states and the integration of multiple spatial cues that shape depth-related behaviors. Here we demonstrate that mouse responses to visual depth cues are more sophisticated than previously recognized, involving both direct avoidance behaviors and complex modulations of exploratory patterns. By combining a modified circular apparatus with Hidden Markov Model analysis, we reveal that mice transition between three distinct behavioral states—resting, exploring, and navigating—in response to visual depth cues. Using this framework, we uncover several fundamental aspects of mouse visual processing: depth perception has an optimal range of spatial frequencies, with strongest responses to patterns between 6–8 cm; visual processing integrates multiple spatial cues rather than triggering simple avoidance; and initial strong cliff-avoidance responses evolve into more nuanced behavioral adaptations over time. Comparisons between wild-type C57BL/6J mice ( Mus musculus ), retinal degeneration models (rd1-2J, C57BL/6J background, Mus musculus ), and control conditions confirm that these behavioral patterns specifically reflect visual processing rather than general exploratory behavior. These findings reveal that mouse depth perception involves sophisticated neural processing that modulates overall exploratory behavior rather than simply triggering avoidance responses. Our approach establishes a new framework for analyzing complex behavioral sequences in neuroscience research, demonstrating how refined behavioral analysis can reveal previously undetectable aspects of sensory processing.