This investigation describes the discovery of novel possible cerebrospinal fluid (CSF) biomarkers for frontotemporal dementia (FTD) using surface-enhanced laser desorption/ionization time-of-flight (SELDI-TOF) mass spectrometry (MS). Sixteen clinically diagnosed FTD patients and 12 non-demented controls were included in the study. CSF was collected and analyzed for protein expression by SELDI-TOF MS. The samples were analyzed on four different array surfaces using two different energy-absorbing molecules as matrices. In total each sample was subjected to eight different surface/matrix conditions. About 2000 protein peaks (mass/charge ratios) were detected. Forty-two peaks were differentially expressed in FTD (P < 0.01). After exclusion of peaks with low signal-to-noise ratio and/or poor resolution and peaks representing differentially charged proteins, 10 peaks remained, five of which were increased and five decreased in FTD cases compared to controls. Using partial least square discriminant analysis (PLS-DA), the combination of these biomarkers discriminated FTD from non-demented controls with a sensitivity of 94%, a specificity of 83% and an accuracy of 89%. Five of the peaks were purified further and identified by tandem MS as a fragment of neurosecretory protein VGF, transthyretin, S-cysteinylated transthyretin, truncated cystatin C and a fragment of chromogranin B. With use of these potential biomarkers, FTD can be distinguished from control subjects with high accuracy in this pilot study.