In a planar free-hand drawing of an ellipse, the speed of movement is proportional to the −1/3 power of the local curvature, which is widely thought to hold for general curved shapes. We investigated this phenomenon for general curved hand movements by analyzing an optimal control model that maximizes a smoothness cost and exhibits the −1/3 power for ellipses. For the analysis, we introduced a new representation for curved movements based on a moving reference frame and a dimensionless angle coordinate that revealed scale-invariant features of curved movements. The analysis confirmed the power law for drawing ellipses but also predicted a spectrum of power laws with exponents ranging between 0 and −2/3 for simple movements that can be characterized by a single angular frequency. Moreover, it predicted mixtures of power laws for more complex, multifrequency movements that were confirmed with human drawing experiments. The speed profiles of arbitrary doodling movements that exhibit broadband curvature profiles were accurately predicted as well. These findings have implications for motor planning and predict that movements only depend on one radian of angle coordinate in the past and only need to be planned one radian ahead.optimal control | motor system | free-hand drawing | power law N atural body movements have surprising regularities despite the complexity of the movements and the many degrees of freedom that are involved. These regularities provide insights into principles underlying the mechanisms that generate the movements. We investigate here conditions when a known regularity fails to adequately describe motor behavior, which adds new insights into how movements are generated by the nervous system.One of the best-studied movement regularities is the inverse relationship between the speed and curvature of 2D hand movements observed when subjects are instructed to freely draw curved shapes such as ellipses, which follow a remarkably simple power law (1):This empirical relationship is called the one-third power law between speed and curvature, or equivalently, the two-thirds power law between angular speed and curvature. On a log curvature vs. log speed plot of hand movement, the power law appears as a single straight line with a slope of −1=3 (Fig. 1A). However, deviations from Eq. 1 have been observed for some movement trajectories. One class of deviation occurs when the curvature changes sign: The power law predicts the speed should diverge to infinity as the curvature approaches zero, which is physically implausible, whereas actual movements exhibit smooth, finite speed profiles at such inflection points. Deviations have also been observed for movements without inflection points (1-4). These noninflectional deviations occur in complex movements and are characterized by fragmentation of the log speed vs. log curvature plot into multiple line segments (Fig. 1B). These observations led to the segmented control hypothesis that complex movements could be generated by concatenating smaller and simpler movement...