Antagonist cocontraction was hypothesized to limit net moment production in children with spastic diplegic cerebral palsy (CP). A second hypothesis was that concontraction would vary with joint angle. To test these hypotheses, surface EMG activity and moment data from the quadriceps and hamstrings muscle groups were obtained from children with CP and compared with normally developing children during isometric flexion and extension exertions. A biomechanical model was developed to predict individual moments produced by the agonist and antagonist muscle groups. Concontraction was defined as the percentage of the net moment that was negated by the antagonist moment. The model performed well in predicting the measured moment as illustrated by high R2 correlation coefficients and low prediction errors. The mean maximum moment produced was greater in normally developing children than children with CP in both flexion and extension. Antagonist cocontraction during extension was greater in children with CP (12.2 +/- 14.4%) than in normally developing children (4.9 +/- 3.8%), implying that antagonist cocontraction is one explanation for the observed extension weakness in children with CP. However, during flexion, cocontraction was not significantly different between the two groups. Cocontraction differed significantly with joint angle in both groups during flexion and in the normally developing children during extension. Although quantifying coactivation based on EMG activity alone produced similar results, it underestimated the effect of the antagonist. The quantification of cocontraction has potential applications for characterizing spastic muscle dysfunction and thereby improving clinical outcomes in children with CP.