Figure 2
Characterization of swimming bouts based on kinematic features. (A) Schematic representation of the underlying assumptions of the Ornstein–Uhlenbeck model. Transitions in behavior occur when the state of the network controlling the behavior passes a certain threshold (upper line). The state of the network is at any point described by a random-walk process which will eventually lead to spontaneous transitions events. In addition, a command signal that actively pushes the state towards the threshold may be present (indicated by the dashed lines). (B–D) Distributions of kinematic features extracted for each bout (total n = ∼120,000), (B) bout duration, (C) bout distance and (D) turning angle. Yellow lines denote fitting of the Ornstein–Uhlenbeck model to experimental data and asterisks show binning density. (E,F) Classification of swimming trajectories reveals differences between swimming behavior observed in individual bouts. (E) Trajectories representing the mean trajectories of 15 bout types identified by the algorithm are shown in the horizontal plane (x-y). (F) The trajectories in (E) shown with time from bout on-set on the y-axis (x-t). The apparent earlier onset of type 15 is caused by data subsampling (see Methods). (G) The relative frequency of the number of observed bouts of each type (class code as in E-F; showing that shorter and straight bouts tend to be more common). Whiskers denote mean and SD of the distribution over all individuals – note the relatively small SD and that the pooled probability (bars) closely matches the average over individuals (asterisks) indicating a small inter-individual variance.
