Although speed-accuracy trade-offs and planning and execution of rapid goaldirected movements have garnered significant research interest, far fewer studies have reported results on the lower end of the movement speed spectrum. Not only do very interesting observations exist that are unique to slow movements, but an explanation of these observations is highly relevant to motor function recovery and motor skill learning, where movements are typically slow at the initiation of therapy or learning, and movement speed increases through practice, exercise or therapy. In the first part of this thesis, based on data from nine stroke patients who underwent a month-long hybrid traditional and robotic therapy protocol, a correlation analysis shows that measures of movement quality based on minimum jerk theory for movement planning correlates significantly and strongly with clinical measures of motor impairment. In contrast, measures of movement speed lack statistical significance and show only weak to moderate correlations with clinical measures. These results constitute an important step towards establishing a much-needed bridge between clinical and robotic rehabilitation research communities. In the second part, the origins of movement intermittency or variability in slow movements are explored. A study with five healthy subjects who completed a manual circular tracking task shows that movement intermittency increases in distal direction along the arm during multi-joint movements. This result suggests that a neuromuscular noise option is favored against a submovement-based central planning alternative, as the source of variability in slow movements. An additional experimental study with eight healthy subjects who completed slow elbow flexion movements at a constant slow speed target under varying resistive torque levels demonstrates that resistive torques can significantly decrease movement speed variability. The relationship between resistive torque levels and speed variability, however, is not monotonic. This finding may constitute a basis for proper design of novel human skill augmentation methods for delicate tasks and improve motor rehabilitation and learning protocols. Finally, a neuro-musculoskeletal model of the elbow suggests that movement speed variability in slow movements cannot be solely attributed to variability in the mechanics of muscle force generation. Together, these analyses, simulations, and experiments shed light on variability in slow movements, and will inform the development of novel paradigms for robotic rehabilitation, motor skill learning and augmentation.