Fig. 1. High throughput measurement of locomotor kinematics in zebrafish larvae. (A) Simultaneous tracking of multiple larvae. In this example, 24 larvae are tracked over 1000 ms (red), with position and curvature information measured every ms. Scale bar, 2.0 mm. (B) Primary measurements are position, orientation (Bi) and curvature (Bii). Quantitative kinematic descriptions of locomotion are derived from these measurements, yielding measures including C-bend angle (Bii), distance traveled (Biii) and trajectory (Biv). Note that this is a high-resolution image – movement analysis is performed on the lower quality images in A. (C) Four examples showing curvature across time (400 ms). The three lower examples demonstrate the smooth changes in curvature observed for active larvae, compared to the flat curvature function for a stationary fish. (D) Scatter analysis of stationary (red squares, N=175) and active larvae (blue squares, N=156) shows that active larvae can be distinguished from stationary larvae by the maximal signal power of the Fourier transform of the curvature function, together with the maximal three-point derivative of the function. (E) Comparison of automated and manual analysis of a new group of 800 events using the criteria established in D demonstrates that automatic analysis reliably distinguishes stationary and active larvae with 98% accuracy. Larvae moving at the beginning of the video recording are detected with 90% accuracy, being mistaken for larvae initiating movement 7% of the time and stationary larvae 3% of the time. Altogether, 96.8% (775/800) events are correctly recognized.