The degree of systolic hypotension causing failure and recovery were tested simultaneously with three oximeters (CSI 504US, Nellcor N-200, and Ohmeda 3740) in nine normal male volunteers. Perfusion of the right hand was slowly reduced and restored by 1) elevation of the hand plus systemic hypotension with nitroprusside if needed (EL); 2) clamp compression of the brachial artery (CL); 3) brachial cuff inflation (CU); and 4) intraarterial norepinephrine (NE). With EL, pulse pressure was normal whereas right radial arterial systolic pressure (SP) was 25.3 ± 12.4 mmHg at failure and 34.1 ± 13.3 at recovery (mean of three oximeters, n = 189). With CL, pulse pressure fell more than did mean pressure, and failure occurred at 37.3 ± 9.8 and recovery at 46.8 ± 17.6 mmHg, n = 84. With CL, threshold of function, defined as the average of failure SP and recovery SP, was 47.1 ± 13.5, n = 41 for Nellcor, higher than for either CSI (38.7 ± 14.5, n = 17) or Ohmeda (36.0 ± 3.4, n = 26) (P< 0.05). With EL, no difference among instruments was found (mean 29.7 ± 12.8, n = 189). Threshold was 58.2 ± 8.4, n = 17 with CU if cuff inflation was slow (filling veins), but recovery was similar to EL after rapid cuff occlusion. With NE, SP threshold was increased to 58.3 ± 21.0 with CL but only to 41.0 ± 13.8 with EL. Hypoxia to Sao270% reduced SP thresholds with CSI to 25.1 ± 12.0 and with Ohmeda to 21.0 ± 7.6 with EL, but had no effect on the Nellcor, nor on any oximeter with CL. With systolic pressure just above failure, flow measured by finger plethysmograph was usually zero and detection of an induced hypoxic transient was delayed in one test as much as 6 min. In normoxic subjects, Spo2decreased at threshold to 88.8 ± 11.5%, n = 210, suggesting that arterial O2was lost to tissue by pulsatile movement in and out of capillaries, or by transarterial diffusion. This suggests that failure of pulse oximeters on fingers occurs at pressures too low for adequate tissue perfusion except when vasoconstriction is present. There are minor threshold differences between instruments, but efforts to increase sensitivity may not be justified in view of the local arterial desaturation evident at current thresholds during normoxia.