Movement related EMGs become more variable during learning of fast accurate movements

Human subjects performed simple flexion and extension movements about the elbow in a visual step-tracking paradigm. Movements were self-terminated. Subjects were instructed to increase movement velocity while maintaining end-point accuracy during practice. The effects of practice on the pattern and variability of EMG activity of the biceps and triceps muscles were studied. Initial movements were performed using reciprocal phasic activation of agonist and antagonist muscles as indicated by surface EMGs. With practice, increases in movement speed were associated with larger agonist and antagonist bursts and an earlier onset of the antagonist burst. Decreased duration of the premovement antagonist silence was also observed during practice. Decreases in variability of movements during practice were not accompanied by equivalent decreases in variability of the associated EMGs. Surprisingly, both agonist and antagonist EMGs were more variable in faster, practiced movements. The combined agonist-antagonist EMG variability depended on both movement speed and trajectory variability. Lower variability in movements in the presence of greater variability in the related EMGs occurred because of linked variations in agonist and antagonist muscle activities. Variations in the first agonist burst were often compensated for by associated variations in the antagonist and late agonist bursts. These linked variations maintained the limb trajectory relatively constant in spite of large variations in the first agonist burst. Modifications to impulse-variability models are therefore needed to explain compensations for variability in accelerative impulses (produced by the first agonist burst) by linked variations in impulses for deceleration (produced by the antagonist and late agonist bursts).