Activating emergency medical services (EMS) is the most important factor in reducing delay times to hospital arrival for stroke patients. Determining who calls 911 for stroke would allow more efficient targeting of public health initiatives. The T.L.L. Temple Foundation Stroke Project is an acute stroke surveillance and intervention project in nonurban East Texas. Prospective case ascertainment allowed chart abstraction and structured interviews for all hospitalized stroke patients to determine if EMS was activated, and if so, by whom. Of 429 validated strokes, 38.0% activated EMS by calling 911. Logistic regression analysis comparing those who called 911 with those who did not activate EMS found that individuals who were employed were 81% less likely to have EMS activated (OR 0.19, 95% CI 0.04 to 0.63). Of the 163 cases in which 911 was called, the person activating EMS was: self (patient), 4.3%; family member of significant other, 60. 1%; paid caregiver, 18.4%; and coworker or other, 12.9%. Significant associations between the variables age group (P=0.02), insurance status (P=0.007), and living alone (P=0.05) with who called 911 was found on chi(2) analysis. Educational efforts directed at patients themselves at risk for stroke may be of low yield. To increase the use of time dependent acute stroke therapy, interventions may wish to concentrate on family, caregivers, and coworkers of high-risk patients. Large employers may be good targets to increase utilization of EMS services for acute stroke.