Objective: To assess the predictive value of early shoulder and hand movement after stroke for subsequent hand movement and function. Design: An inception cohort design following 71 patients for 3 months after first stroke. Setting: Rehabilitation medicine units in three Sydney teaching hospitals. Patients: A consecutive sample of 71 first-stroke patients, mean age 67 years, was assessed for upper limb function. Those with preexisting impairment in the hemiplegic limb, early good recovery in hand function, or impaired comprehension were excluded. Data were available for 65 patients at 1 month, 50 at 2 months, and 46 at 3 months. Main Outcome Measures: Hand movement and hand function were assessed at 1, 2, and 3 months. A standardized hand movement scale and four specific hand function tasks were used. The outcome variables were (1) 'good' hand movements, ie, independent index finger extension or opposition of finger/s to thumb and (2) ability to perform one of the specified hand function tasks. Results: Initial shoulder shrug predicted good hand movement (odds ratios 7.3, 7.0, 6.0) and hand function (odds ratios 13.8, 5.3, 11.3) at 1, 2, and 3 months, respectively. Initial presence of synergistic hand movement predicted good hand movement (odds ratios 10, 13.8, 12.2) at 1, 2, and 3 months and hand function (odds ratios 27.9, 7.5) at 1 and 2 months. Initial active shoulder abduction predicted good hand movement at 1 month and hand function at 1 and 2 months only. Conclusions: Early shoulder shrug and synergistic hand movements are useful bedside predictors of hand outcome after stroke.