The aim of the present study was to investigate the effect of experimental and chronic neck-shoulder pain on the magnitude of cycle-to-cycle variability of task timing, kinematics and muscle activation during repetitive arm movement performed for 3 or 5 min. In an experimental part, acute muscle pain was induced in healthy subjects by intramuscular injection of hypertonic saline in trapezius (n = 10) and infraspinatus (n = 10) muscles. In a clinical part, workers with (n = 12) and without (n = 6) chronic neck-shoulder pain were compared. Cycle-to-cycle standard deviations of task duration, arm and trunk movement in 3D and surface electromyographic (EMG) root mean square activity were computed to assess the degree of variability. The variability in task timing increased in presence of both experimental and chronic pain (P < 0.05) compared with non-painful conditions. Experimental pain increased the variability of the starting position of the arm (P < 0.05), the arm range of motion (P < 0.01), the arm and trunk movement area (P < 0.01) and the acceleration of the arm (P < 0.01). In the chronic pain condition, the variability of arm and trunk acceleration (P < 0.01) and EMG activity (P < 0.05) was decreased compared with healthy controls. These results indicate that pain alters the magnitude of motor variability, and that the transition from acute to chronic pain is accompanied by changes in motor patterns. Experimental pain likely resulted in a quest for a motor solution reducing nociceptive influx, while chronic pain was characterised by a diminished motor flexibility.