The aims of the present study were to quantify the impairment in ankle coactivation on the paretic and non-paretic sides of subjects with hemiparesis and to examine the relationship of ankle coactivation with postural instability, motor deficit of the paretic lower extremity and locomotor performance. Electromyography of the medial gastrocnemius (MG) and tibialis anterior (TA) muscles were recorded bilaterally during gait in 30 subjects (62.1+/-9.9 years) who had suffered a recent stroke (<6 months) as well as on one side of 17 healthy controls (59.3+/-9.1 years) walking at very slow speed. Ankle muscle coactivation was calculated by dividing the time of overlap between MG and TA signals (threshold of 20 microV) by the duration of the gait phases of interest: stance, swing, first and second double support sub-phases and single support sub-phase. The time spent in single support and the peak plantarflexor moment of force on the paretic side were used to measure, respectively, postural stability and dynamic strength of the paretic plantarflexors. The subjects with hemiparesis demonstrated less coactivation on the paretic side during the single support sub-phase (p<0.01) and more coactivation during first and second double support sub-phases on the non-paretic side (p<0.001) compared to control values. The patients with coactivation patterns that differed the most from controls were the patients with the more severe impairments and disabilities. While the reduced coactivation on the paretic side may contribute to poor postural stability and poor locomotor performance, the presence of excessive coactivation on the non-paretic side when both limbs were in ground contact may be an adaptation to help maintain postural stability during gait.