Background: The reliability of near-infrared spectroscopy (NIRS) for measuring cerebral oxygenation (ScO₂) is controversial due to the possible contamination from extracranial tissues. We compared ScO₂ measured with the NIRS optode on the forehead, the skull and the dura mater in anaesthetised patients undergoing craniotomy. We hypothesised that ScO₂ measured directly on the skull and the dura mater would differ from ScO₂ measured on the skin. Methods: This prospective observational study included 17 adult patients scheduled for elective craniotomy. After induction of general anaesthesia, ScO₂ was measured on the forehead skin, as well as on the skull and on the dura mater in the surgical field. The primary comparison was the difference in ScO₂ measured on the dura mater and on ScO₂ measured on the skin; secondary comparisons were the differences in ScO₂ on the skull and ScO₂ on the skin and the dura mater, respectively. Data were described with median (5%–95% range) and analysed with the Wilcoxon signed-rank test. Results: ScO₂ values on the dura mater were obtained in 11 patients, and median ScO₂ (48%, 29%–95%) did not differ significantly from ScO₂ on the skin (73%, 49%–92%; p =.052), median difference −25% (−35.6% to −1.2%). ScO₂ on the skull (N = 16) was lower than that on the skin (63% [43%–79%] vs. 75% [61%–94%]; p =.0002), median difference −10% (−20.8 to −3.0). Conclusion: In adults undergoing craniotomy, NIRS-based ScO₂ measured on the dura mater did not reach statistically significantly lower values than ScO₂ measured on the skin, whereas values on the skull were lower than on the skin, indicating a contribution from scalp tissue to the signal.