The paper by Thoroughman and Shadmehr explores how the brain learns and controls movements through the flexible combination of motor primitives. They use a time-series analysis to investigate how humans learn the dynamics of reaching movements, finding that the brain learns through the combination of Gaussian-like tuning functions that encode hand velocity. The study reveals that the wide tuning of these primitives limits the brain's ability to represent viscous dynamics, which aligns with subjects' adaptation to novel force fields. The mathematical properties of these primitives resemble the tuning curves of Purkinje cells in the cerebellum, suggesting that these cells may encode the primitives involved in learning dynamics. The authors also demonstrate that errors experienced in one movement can affect subsequent movements, and they quantify the sensitivity of the internal model (IM) to these errors. They find that errors in one direction can influence the IM for other directions, with the influence decaying as the angular distance between directions increases. The study further predicts that wide Gaussians, which have a broader tuning, can lead to S-shaped movement trajectories and overcompensation in force fields, while narrow Gaussians do not. Finally, they show that learning with wide Gaussians imposes limits on adaptation to high spatial frequency force fields, consistent with human performance.The paper by Thoroughman and Shadmehr explores how the brain learns and controls movements through the flexible combination of motor primitives. They use a time-series analysis to investigate how humans learn the dynamics of reaching movements, finding that the brain learns through the combination of Gaussian-like tuning functions that encode hand velocity. The study reveals that the wide tuning of these primitives limits the brain's ability to represent viscous dynamics, which aligns with subjects' adaptation to novel force fields. The mathematical properties of these primitives resemble the tuning curves of Purkinje cells in the cerebellum, suggesting that these cells may encode the primitives involved in learning dynamics. The authors also demonstrate that errors experienced in one movement can affect subsequent movements, and they quantify the sensitivity of the internal model (IM) to these errors. They find that errors in one direction can influence the IM for other directions, with the influence decaying as the angular distance between directions increases. The study further predicts that wide Gaussians, which have a broader tuning, can lead to S-shaped movement trajectories and overcompensation in force fields, while narrow Gaussians do not. Finally, they show that learning with wide Gaussians imposes limits on adaptation to high spatial frequency force fields, consistent with human performance.