This paper presents a model for representing context-dependent variation of articulator movements. Our model explains the contextual effect based on a multidimensional phonemic task and dynamic constraints of movements. The task determines the articulatory target so that invariant features of phoneme articulation are achieved. The dynamic constraints represent smoothly moving behavior of the articulators. Because the dimension of the task is smaller than that of the articulator variables, there are unconstrained degrees-of-freedom of the articulator variables. These redundant components are used to represent the contextual effect by smoothly interpolating the adjacent tasks. The phonemic invariant feature is defined as a linear transformation that minimizes a normalized articulatory variation. Simulation of articulatory movements is performed and the results are compared with actual movements.