behavior*

Behavior exhibits multiple timescales: from fast control by neural activity to slower variation due to neuromodulation or even aging. Can we extract longer-time scale dynamics, driven by such hidden processes, from short-time behavior? We use a recently developed transfer operator approach, akin to a Fokker-Planck description, to perform a top-down coarse-graining of C. elegans posture dynamics: from long timescale ``roaming’' and ``dwelling’' states (minutes), to a ``run’' and ``pirouette’' description (seconds), down to fine scale movements (milliseconds). We use the transfer operator spectrum to illustrate how genetic and environmental perturbations can impair the generation of long-lived behavioral patterns. We study the effects of starvation and of mutation in a neuropeptide receptor gene, and show that although both perturbations lead to the absence of a ``dwelling’' state, there are important fine scale differences that hint at distinct neuromodulatory mechanisms.