We tested the hypothesis that force variability and error during maintenance of submaximal isometric knee extension are greater in subacute stroke patients than in controls and are related to motor impairments. Contralesional (more-affected) and ipsilesional (less-affected) legs of 33 stroke patients with sufficiently high motor abilities (62 ± 13 yr, 16 ± 2 days postinjury) and the dominant leg of 20 controls (62 ± 10 yr) were tested in sitting position. After peak knee extension torque [maximum voluntary contraction (MVC)] was established, subjects maintained 10, 20, 30, and 50% of MVC as steady and accurate as possible for 10 s by matching voluntary force to the target level displayed on a monitor. Coefficient of variation (CV) and root-mean-square error (RMSE) were used to quantify force variability and error, respectively. The MVC was significantly smaller in the more-affected than less-affected leg, and both were significantly lower than in controls. The CV was significantly larger in the more-affected than less-affected leg at 20 and 50% MVC, whereas both were significantly larger compared with controls across all force levels. Both more-affected and less-affected legs of patients showed significantly greater RMSE than controls at 30 and 50% MVC. The CV and RMSE were not related to the Fugl-Meyer motor score or to the Rivermead Mobility Index. The CV negatively correlated with MVC in controls but only in the less-affected leg of patients. It is concluded that isometric knee extension strength and force control are bilaterally impaired soon after stroke but more so in the more-affected leg. Future studies should examine possible mechanisms and the evolution of these changes.