Four faults are reported in the Hawksley Random Zero Sphygmomanometer (RZS). Our study is of their mechanism.(i) Compared with a mercury sphygmomanometer the RZS underestimates blood pressure (BP). We confirm this: for 240 measurements by three experienced operators in 12 patients, systolic BP was 3.4 mm Hg lower in the RZS; diastolic pressure was not underestimated. A cause of under-estimation in 89% of measurements was that mercury stuck in the manometer giving a false high reading of random zero (RZ). Tilting the RZS before reading RZ reduced under-reading by 1.6 mm Hg. A rare cause is failure of the operator to completely close the reservoir tap.(ii) Values of RZ are not randomly distributed; non-randomness is most marked in measurements made by experienced operators whose speed of measurement provides insufficient time during cuff inflation for filling of the diaphragm chamber. Smaller contributions are made by the sticking of mercury in the manometer and by a leak of air through the air bleed screw.(iii) Consecutive RZ estimates often have similar value. This has two causes: short cuff inflation time and short interval between opening the reservoir tap and spinning the thumb-wheel.(iv) An inverse relation of RZ and BP suggested by earlier work and by our own data is probably an artifact: when BP is low, measurement is quick and RZ is falsely high; when BP is high, measurement takes longer and RZ is lower. These four faults could be partly or wholly avoided by a change in the operators’ technique.