Raman spectroscopy has been shown to have the potential for providing differential diagnosis in the cervix with high sensitivity and specificity in previous studies. The research presented here further evaluates the potential of near-infrared Raman spectroscopy to detect cervical dysplasia in a clinical setting. Using a portable system, Raman spectra were collected from the cervix of 79 patients using clinically feasible integration times (5 seconds on most patients). Multiple Raman measurements were taken from colposcopically normal and abnormal areas prior to the excision of tissue. Data were processed to extract Raman spectra from measured signal, which includes fluorescence and noise. The resulting spectra were correlated with the corresponding histopathologic diagnosis to determine empirical differences between different diagnostic categories. Using histology as the gold standard, logistic regression discrimination algorithms were developed to distinguish between normal ectocervix, squamous metaplasia, and high-grade dysplasia using independent training and validation sets of data. An unbiased estimate of the accuracy of the model indicates that Raman spectroscopy can distinguish between high-grade dysplasia and benign tissue with sensitivity of 89% and specificity of 81%, while colposcopy in expert hands was able to discriminate with a sensitivity and specificity of 87% and 72%.