Reaching tasks are considered well-executed if they appear "smooth," a quality that is typically quantified by its opposite, jerk, the rate of change of acceleration. While jerk is a theoretically sound measure, its application to spastic individuals sometimes yields counter-intuitive results, and does not reveal motor impairment across the workspace. To more generally quantify spontaneous accelerative transients (SATs) within a movement, a pseudo-wavelet transform was devised that iteratively compared angular trajectories to a series of straight-line approximants. Cumulative linear fit errors were expressed in terms of flexion angle, yielding an SAT map of the entire motion. To compare SAT maps with traditional smoothness measures, two scalar indices were extracted from them: residual excursion deviation (RED), representing the integral over Deltatheta and the ratio of peak error to mean error (PEME) on the map. Fifteen subjects, including five subjects with chronic stroke performed elbow flexions throughout their entire ranges of motion, Deltatheta, at a comfortable pace with their arms supported in the transverse plane. Maps revealed that stroke subjects were significantly less coordinated than controls, as measured both by RED: 8.0+/-2.9 x 10(-3) versus 3.1+/-0.8 x 10(-3) and PEME: 6.6+/-0.9 versus 12.1+/-1.9, both P<0.001. Comparable jerk metrics, including integrated average jerk, did not report a significant performance deficit at the P<0.05 level. Map metrics for all subjects were independent of average velocity (correlation with theta : rho0.31), but jerk-based metrics for stroke subjects were spuriously co-variant with velocity rho=0.85, which may relate to the significantly higher mean arrest period ratio in stroke subjects (0.26+/-0.19 versus 0.09+/-0.08, P<0.001). We conclude that SAT maps provide reliable information on regional movement impairments at a wide range of proficiency levels.