The purpose of this study was to determine whether cross-education, defined as the increase in strength of an untrained limb after training of the contralateral homologous limb, is specific to low and high velocity eccentric training. Twenty-six subjects were randomized into two groups ( n=13 each) that performed unilateral eccentric training of the elbow flexors on an isokinetic dynamometer at velocities of either 30 degrees s(-1) (0.52 rad s(-1)) or 180 degrees s(-1) (3.14 rad s(-1 )). Subjects trained three times per week for 8 weeks. Ten subjects served as controls and did not train. Subjects were tested before and after training for peak torque of the elbow flexors during eccentric and concentric contractions at 30 degrees s(-1) and 180 degrees s(-1). Eccentric peak torque at the velocity of 180 degrees s(-1) in the untrained arm increased only for the group that trained at that velocity (P<0.05). There were no other changes in untrained arms for any of the groups at velocities of 30 degrees s(-1) or 180 degrees s(-1). For the trained arm, the increase in eccentric torque (pooled over velocities) was greatest for the group training at 180 degrees s(-1), whereas the increase in concentric torque was similar for the groups training at 30 degrees s(-1) and 180 degrees s(-1). For the trained arm, there was no specificity for velocity or contraction type. We conclude that cross-education was specific to contraction type and velocity when fast (but not slow) eccentric contractions were used during training; whereas there was no specificity of training in the trained arm.