Research overview: hidden Markov models for behavioral states

Hidden Markov models infer an animal's unobserved behavioral state, such as resting, foraging, or directed travel, from the statistical signatures of its movement, typically step lengths and turning angles. This overview explains how the models assume the animal switches among a small number of latent states, each with characteristic movement, and estimate both the state sequence and the probabilities of switching. Applied to felid tracks, they can segment a trajectory into encamped, hunting, and relocating phases without direct observation, and can link state occupancy to habitat or time of day. The overview covers model selection, the interpretation of states, and the option to incorporate covariates and accelerometer data. Hidden Markov models are a leading tool for extracting behavioral structure from raw movement geometry.