Optical spectroscopy methods are fast emerging as potential alternatives for early diagnosis of cancer. A Raman spectroscopy method for discrimination of normal and malignant oral tissues has been developed by us earlier. It is necessary to evaluate and establish the validity of the approach before it can be routinely used. In the present study, our Raman spectroscopy investigations are extended further to evaluate the efficacy of the technique to discriminate between normal, inflammatory, premalignant, and malignant conditions in oral tissue. Spectral profiles of normal, malignant, premalignant, and inflammatory conditions show pronounced differences between one another. Spectra of normal tissues can be attributed mainly to lipids whereas pathological tissue spectra are dominated by proteins. Principal components analysis (PCA) of the spectral data sets belonging to the four different categories showed that scores of factors differentiated between normal and all pathological conditions but gave only poor discrimination among the three pathological states. PCA combined with multiparameter limit tests allow match/mismatch criteria to be applied to test samples when pathologically certified calibration sets are available in each class. It is shown that by this method all the four tissue types could be discriminated and diagnosed correctly. The biochemical differences between normal and pathological conditions of oral tissue are also discussed from spectral differences of the different classes of spectra.