Trained nurse-specialists obtained 84 000 blood-pressure measurements in 1240 obese subjects using cuffs of the three standard adult sizes in a randomised order. The differences in readings between the three cuffs were smallest in non-obese subjects and became progressively greater with increasing arm circumference (AC) in the obese population. The regular cuff (12 X 23 cm) showed the greatest bias in relation to AC. Formulae and a table have been derived to correct the measurement error caused by cuffs of inappropriate size at various ACs. The reported high prevalence of hypertension in obese subjects may be greatly overestimated.