1. Near-infrared (IR) spectroscopy is based on the relative transparency of skin, skull and brain to the light in the near-IR region (700-1100 nm) and on the oxygen-dependent tissue absorption changes of haemoglobin. 2. We evaluated the most relevant factors (reproducibility, venous return, age and sex) that might affect reliability of near-IR spectroscopy to test CO2 cerebrovascular reactivity. 3. Thirty-four healthy volunteers were enrolled in the study. The protocol consisted of a 3-min baseline, a 3-min hypercapnia (5% CO2 in air) and a 2-min recovery. Transcranial Doppler sonography measurements were simultaneously performed. The CO2 reactivity test was repeated on 27 subjects after 1 h to assess reproducibility. CO2 reactivity was also evaluated at different body positions (supine, 35 degrees Trendelenburg and 35 degrees reverse Trendelenburg), and over a gradual increase of the inspired CO2. 4. Changes in near-IR spectroscopy and transcranial Doppler sonography parameters were significantly correlated with variations of end-tidal CO2 (P < 0.005). A significant correlation between the reactivity indexes of near-IR spectroscopy parameters and flow velocity was also found (P < 0.01). A high reproducibility was also found for deoxyhaemoglobin (r(i) = 0.76), oxyhaemoglobin (r(i) = 0.68) and flow velocity (r(i) = 0.60) reactivity indexes. No significant differences between the reactivity indexes of different body positions were found (P > 0.05). The reactivity index of oxyhaemoglobin and deoxyhaemoglobin decreased (P < 0.05) and increased (P < 0.01) with age respectively. 5. We found that near-IR spectroscopy is a reliable and reproducible method for the evaluation of cerebrovascular reactivity and might be considered, after appropriate validation, for the assessment of patients with cerebrovascular disease.