Blood viscosity is elevated in hypertensive subjects, but the association of viscosity with arterial blood pressure in the general population, and the influence of social, lifestyle and disease characteristics on this association, are not established. In the Edinburgh Artery Study, 1592 men and women aged 55-74 years selected randomly from the general population attended a university clinic. A fasting blood sample was taken for the measurement of blood viscosity and its major determinants (haematocrit, plasma viscosity and fibrinogen). Systolic pressure was related univariately to blood viscosity (P < 0.001), plasma viscosity (P < 0.001) and plasma fibrinogen (P < 0.01), but the association with fibrinogen did not persist after adjusting for body mass index. Diastolic pressure was related univariately to blood viscosity (P < 0.001) and plasma viscosity (P < 0.001) and haematocrit (P < 0.001) but not to fibrinogen. The only difference between the sexes was that the association between blood viscosity and systolic pressure was confined to males. Blood viscosity was associated equally with systolic and diastolic pressures in males, and remained independently related on multivariate analysis adjusting for age, sex, body mass index, social class, smoking, alcohol intake, exercise, angina, HDL and non-HDL cholesterol, diabetes mellitus, plasma viscosity, fibrinogen, and haematocrit.