Estimation of regional tissue oxygenation (rStO2) by near infrared spectroscopy enables non-invasive end-organ oxygen balance monitoring and could be a valuable tool in intensive care. However, the diverse absolute values and dynamics of different devices, and overall poor repeatability of measurements are a problem. The aim of the present study is to test the hypothesis that INVOS 5100C, FORE-SIGHT and NONIN EQUANOX 7600 have similar properties concerning absolute values, repeatability, and sensitivity to changes in rStO2. To test repeatability the sensors were repositioned 20 times during hemodynamic steady state on the adult forearm. Afterwards six vascular occlusions by inflation of an upper arm cuff were done to achieve low oxygenation in the forearm. Absolute values were compared by repeated-measures ANOVA, repeatability was estimated by the within-subject standard deviation, Sw, and response to changing oxygenation by the down slope of rStO2 during vascular occlusion in the respective arm. 10 healthy adults, 21-29 years old, with double skinfolds on the forearm less than 10 mm participated. The median rStO2 was 70.7% (interquartile range (IQR) 7.7%), 68.4% (IQR 8.4%), and 64.6% (IQR 4.8) with INVOS, NONIN, and FORE-SIGHT, respectively, the median rate of decline was 13.2%/min (IQR 9.6), 22.8 %/min (IQR 18.0), and 10.8%/min (IQR 6.0), and the same-site repeatability was 2.9% (95% CI 2.4-3.3), 4.6% (CI 3.9-5.3), and 2.0% (CI 1.7-2.3). INVOS gave significantly higher steady state values than FORE-SIGHT, and NONIN had the steepest decline in rStO2, but the poorest repeatability. Two measures of signal-to-noise were similar among devices. This suggests that good repeatability comes at the expense of low sensitivity to changes in oxygenation. Values of rStO2 on the forearm from INVOS, NONIN and FORE-SIGTH cannot be used interchangeably.