The concept of using diffuse reflectance spectroscopy to distinguish intraoperatively between pediatric brain tumors and normal brain parenchyma at the edge of resection cavities is evaluated using an in vivo human study. Diffuse reflectance spectra are acquired from normal and tumorous brain areas of 12 pediatric patients during their tumor resection procedures, using a spectroscopic system with a handheld optical probe. A total of 400 spectra are acquired at the rate of 33 Hz from a single investigated site, from which the mean spectrum and the standard deviation are calculated. The mean diffuse reflectance spectra collected are divided into the normal and the tumorous categories in accordance with their corresponding results of histological analysis. Statistical methods are used to identify those spectral features that effectively separated the two tissue categories, and to quantify the spectral variations induced by the motion of the handheld probe during a single spectral acquisition procedure. The results show that diffuse reflectance spectral intensities between 600 and 800 nm are effective in terms of differentiating normal cortex from brain tumors. Furthermore, probe movements induce large variations in spectral intensities (i.e., larger standard deviation) between 400 and 600 nm.