Surface neuromuscular electrical stimulation (sNMES) after stroke aims to improve upper limb function and reduce shoulder pain, but current evidence of effectiveness is inconclusive. We have undertaken a randomized controlled trial to evaluate sNMES to the shoulder after acute stroke. One hundred seventy-six patients, within 10 days of stroke onset, were randomized to receive sNMES or placebo in addition to stroke unit care. The primary outcome measure was upper limb function measured by the Action Research Arm Test (ARAT) 3 months after stroke. Secondary outcome measures included other measures of upper limb function, upper limb impairment, pain, disability, and global health status. Outcome assessments were blinded. There was no difference in arm function between groups in terms of the primary outcome measure. The median ARAT at 3 months was 50 in the intervention group and 55.5 in the control group (P=0.068). Significant differences were seen at 3 months in favor of the control group for other measures of arm function and impairment: grasp and gross movement subsections of the ARAT, Frenchay Arm Test, and the arm subsection of the Motricity Index. Secondary analysis suggested that these differences were most marked in subjects with severe initial upper limb weakness. A 4-week program of sNMES to the shoulder after acute stroke does not improve functional outcome and may worsen arm function in severely impaired stroke patients. "Routine" use of sNMES to the proximal affected upper limb after acute stroke cannot be recommended.