In previous papers, we have established, through a functional analysis of the behavioral sequences recorded on laboratory mice over a twelve-hour period, the existence of two independent strategies of action selection. On the one hand, the choice of ultradian alternations of rest and activity bouts makes it possible for every mouse to maximize its net energy gain over a one-day or a one-night interval. On the other hand, the succession of acts performed within an activity bout might serve to precisely fit the metabolic needs of each animal, because the corresponding results present great inter-individual variability, with some mice increasing and others decreasing their net energy gain. To determine what kind of dynamic system could generate these successions of acts within an activity bout, we performed a nonlinear time series analysis of the energy costs related to the acts displayed by the mice during such a period. The results suggest that chaotic dynamics might be involved in action selection in mice. Thus, the variability mentioned above could be a consequence of the sensitive dependence on initial conditions associated with such dynamics, and would allow mice to rapidly adapt their metabolic needs to the ongoing situation.