We measured blood pressure differences by cuff size in 181 adults aged 25 to 74 years, allocated to a random sequence that involved the measurement of blood pressure using a small cuff, a large cuff, and an appropriate cuff as determined by standardized arm circumference measurement. Systolic and diastolic blood pressure were underestimated by 3-5 mm Hg in men and 1-3 mm Hg in women when the cuff was one size larger than appropriate. Systolic and diastolic blood pressure were overestimated by 2-6 mm Hg in men and 3-4 mm Hg in women when the cuff was one size smaller than appropriate. In addition, 30-40% of subjects were "misclassified" when blood pressure cutpoints were used to define hypertension.