The discrimination of cell types is a crucial task in cell biology. Available techniques, based on an irreversible treatment of the cells, do not allow a sensitive label-free characterization under in situ conditions. Infrared spectroscopic imaging is a new and useful tool for studying individual cells. It has established itself as a powerful method to probe the molecular composition and to indicate the biochemistry of cells. Monolayers of cultivated U343, T1115 and T508 human glioma cells were characterized using infrared spectroscopic imaging. A classification algorithm based on linear discriminant analysis was developed to distinguish different cells without labeling. The classification is based upon spectral features which mainly arise from proteins, nucleic acids, and cholesterol. An accuracy of 91% and 84% was obtained for cells of U343 and T1115, respectively. Cells of the T508 cell line exhibit some misclassifications resulting in a lower accuracy rate of 73%. As the results demonstrate, the potential of infrared spectroscopic imaging method to assess the overall molecular composition of cells in a non-destructive manner opens the possibility to characterize cells on a molecular level without labels or an irreversible treatment.