The intrinsic autofluorescence properties of biological tissues can be affected by the occurrence of histological and biochemical alterations induced by pathological processes. In this study the potential of autofluorescence to distinguish tumor from normal tissues was investigated with the view of a real-time diagnostic application in neurosurgery to delineate glioblastoma resection margins. The autofluorescence properties of nonneoplastic and neoplastic tissues were analyzed on tissue sections and homogenates by means of a microspectrofluorometer, and directly on patients affected by glioblastoma multiforme, during surgery, with a fiber-optic probe. Scan-microspectrofluorometric analysis on tissue sections evidenced a reduction of emission intensity and a broadening of the main emission band, along with a redshift of the peak position, from peritumoral nonneoplastic to neoplastic tissues. Differences in both spectral shape and signal amplitude were found in patients when the glioblastoma lesion autofluorescence was compared with those of cortex and white matter taken as healthy tissues. Both biochemical composition and histological organization contribute to modify the autofluorescence emission of neoplastic, with respect to nonneoplastic, brain tissues. The differences found in the in vivo analysis confirm the prospects for improving the efficacy of tumor resection margin delineation in neurosurgery.