A non-invasive method of characterizing burn injuries using near infrared spectroscopy is described. In the method, a beam of light is emitted into the burnt tissue portion at two or more different tissue depths. The spectra are then compared using multivariate analysis to determine diagnostic regions of the spectra. This information is used to categorize the burn. In some cases, the diagnostic regions correspond to wavelengths related to the hemodynamics of the tissue portion. The spectra can also be repeated over time, thereby allowing trends and changes in the spectra to be measured. This data is in turn used to categorize the burn as either a superficial burn, partial thickness burn, deep partial burn or a full thickness burn. Once the burn has been categorized, the clinician can intervene as needed to treat the burn.