The paper introduces the concept of *dynamical movement primitives* (DMPs), a method for modeling attractor behaviors in autonomous nonlinear dynamical systems using statistical learning techniques. The authors aim to address the challenge of creating goal-directed behavior in nonlinear systems, which are often parameter-sensitive and difficult to analyze and predict. DMPs transform simple dynamical systems into weakly nonlinear systems with prescribed attractor dynamics through a learnable forcing term. This approach allows for the generation of both point attractors and limit cycle attractors of arbitrary complexity, with multiple degrees of freedom coordinated by arbitrary phase relationships. The stability of the model equations is guaranteed, and the method provides a metric for comparing different dynamical systems. The authors evaluate their approach in motor control and robotics, demonstrating its effectiveness in imitation learning, online modulation, synchronization, and movement recognition. The paper also discusses variations and design principles of the DMPs, emphasizing the importance of structural equivalence, autonomy, and analyzable stability properties.The paper introduces the concept of *dynamical movement primitives* (DMPs), a method for modeling attractor behaviors in autonomous nonlinear dynamical systems using statistical learning techniques. The authors aim to address the challenge of creating goal-directed behavior in nonlinear systems, which are often parameter-sensitive and difficult to analyze and predict. DMPs transform simple dynamical systems into weakly nonlinear systems with prescribed attractor dynamics through a learnable forcing term. This approach allows for the generation of both point attractors and limit cycle attractors of arbitrary complexity, with multiple degrees of freedom coordinated by arbitrary phase relationships. The stability of the model equations is guaranteed, and the method provides a metric for comparing different dynamical systems. The authors evaluate their approach in motor control and robotics, demonstrating its effectiveness in imitation learning, online modulation, synchronization, and movement recognition. The paper also discusses variations and design principles of the DMPs, emphasizing the importance of structural equivalence, autonomy, and analyzable stability properties.