Nine male karate athletes and 13 untrained men did maximal voluntary isometric (MVC) and ballistic elbow extension actions, the latter unloaded (L0) and against a load equal to 10% MVC (L10). The karate group achieved greater (P < 0.05) isometric (32%) and ballistic action peak torque with L0 (30%) and L10 (40%). With L10 the ratio of ballistic action to isometric action, peak torque was 13% greater in the karate group, indicating a load specific training adaptation. With L0 the corresponding ratio did not differ significantly between groups. Ballistic action peak rate of torque development (51%, 51%) and peak acceleration (15%, 9%) with L0 and L10, respectively, were greater in the karate group. In contrast, peak velocity and movement time did not differ significantly between groups. Electromyographic recordings of agonist triceps and antagonist biceps were made during the isometric and ballistic actions. Since ballistic actions (L10) were initiated from a preloaded condition, the occurrence and duration of premovement agonist depression were monitored. In ballistic actions there were no group differences in agonist activation, the ratio of ballistic to isometric action agonist activation, or antagonist coactivation. Premovement agonist depression occurred infrequently in both groups, with no group differences. It is concluded that karate athletes have enhanced elbow extension ballistic performance, but it could not be related to amplified agonist activation, altered antagonist activation, or more frequent occurrence of agonist premovement depression.