We prospectively verified whether an ad-hoc questionnaire administered by phone supports pre-hospital suspicion of stroke in order to reduce the time before care is given. From June 1996 to May 1997, physicians of the Emergency Medical Service in the area of Bergamo, Italy asked all people calling for a patient with symptoms and signs suggesting a cerebral vascular injury to immediately answer some questions on common symptoms and signs of stroke. The medical records of the patients hospitalized at Ospedali Riuniti of Bergamo were reviewed at the end of the study by a single neurologist, skilled in stroke management and blinded to the questionnaires. Sensitivity and specificity, in addition to positive and negative predictive values, of single questions versus final diagnosis were assessed. Logistic regression analysis was also performed to identify those questions useful to suspect strokes. We collected 143 valid questionnaires, related to 63 men and 80 women, aged 34-99 years (mean, 71.8 years). The question concerning headache had the lowest sensitivity and specificity, respectively 57.1% and 36.5%, and the question concerning leg palsy had the highest sensitivity and specificity, respectively 82.0% and 52.4%. Multivariate analysis identified questions on facial and leg palsy as independent predictors of a final diagnosis of stroke. A few questions on motor deficits proposed by emergency medical service operators may be useful in the pre-hospital identification of stroke patients. Concordance of any questions versus final diagnosis of stroke was, however, far to be satisfying. Thus, our experience supports the need for an educational program to improve the efficiency of a pre-hospital diagnosis of stroke.