Brain near-infrared spectroscopy (NIRS) is an emerging neurophysiological tool that combines straightforward activity localization with cost-economy, portability and patient compatibility. NIRS is proving its empirical utility across specific cognitive and emotional paradigms. However, a potential limitation is that it is not only sensitive to haemodynamic changes taking place in the cortex, and task-related cardiovascular responses expressed in the perfusion of extracranial layers may be confounding. Existing literature reports correlations between brain NIRS and systemic blood pressure, yet it falls short of establishing whether in normal participants the blood pressure changes encountered in experimental settings can have confounding effects. Here, we tested this hypothesis by performing two experimental manipulations while recording from superficial occipital cortex, encompassing striate and extrastriate regions. Visual stimulation with reversing chequerboards evoked cortical haemodynamic responses. Simultaneously and independently, transient systemic blood pressure changes were generated through rapid arm-raising. Shallow-penetration NIRS recordings, probing only extra-cerebral tissues, highlighted close haemodynamic coupling with blood pressure. A different coupling pattern was observed in deep-penetration recordings directed at haemodynamic signals from visual cortex. In absence of blood-pressure changes, NIRS signals tracked differences in visual stimulus duration. However when blood pressure was actively manipulated, this effect was absent and replaced by a very large pressure-related response. Our observations demonstrate that blood pressure fluctuations can exert confounding effects on brain NIRS, through expression in extracranial tissues and within the brain itself. We highlight the necessity for continuous blood pressure monitoring alongside brain NIRS, and for further research on methods to correct for physiological confounds. © 2011 Elsevier B.V.