Currently available near infrared spectroscopy (NIRS) devices are unable to discriminate between arterial and venous blood, a potential source of artifact. The purpose of this study was to test the hypothesis that oscillations in NIR signals at the respiratory and cardiac frequency could be attributed to venous and arterial blood, respectively, and thereby isolated. After written informed consent was obtained, a two-wavelength NIRS device was placed over the left frontal cortex in 20 volunteers. After 5 min of unimpeded spontaneous ventilation, an impedance threshold device (ITD, average resistance-7 cm H2O) was applied and an additional two minutes of data recorded. Tissue saturation (StO2) calculated at the ventilatory and cardiac frequencies was compared to non-pulsatile StO2, before and after application of the ITD using spectral peak and power algorithms. The ITD increased non-pulsatile cerebral saturation by 3.6 %. The ITD had no discernable effect on pulsatile estimates of StO2 at either the ventilatory or cardiac frequencies. StO2 estimated at the NIRS spectral peak from 0.75 to 1.75 Hz was 24 % higher than non-pulsatile StO2 (p = 0.0013). There were no other significant differences between pulsatile and non-pulsatile algorithms in the estimation of StO2. In 64 % of cases, both the low (ventilator) and high (cardiac) frequency estimates of StO2 were either both larger or both smaller than non-pulsatile StO2, suggesting that they were interrogating the same vascular bed. Frequency domain analysis cannot reliably separate NIRS waveforms into arterial and venous components.