Raman spectroscopy is a technique that utilises inelastic scattering processes to provide a biochemical fingerprint that has been shown to successfully discriminate oesophageal pathologies. The aim of this study was to develop Raman spectroscopy as a clinical tool; both in vivo for ‘targeted biopsy’, and in ex vivo for ‘automated histopathology’. Two different Raman probes were evaluated and compared and tissue classification models generated ex vivo. A preliminary classification model of a novel single collection fibre probe demonstrated potential for the probe design. Both probes were shown to discriminate three different oesophageal pathology groups. A cross-validated tissue classification model (88 samples) discriminated normal, Barrett’s and neoplasia with an overall accuracy of 86.5% with a sensitivity of 83.3-89.5% and specificity of 89.2-97.1%. A novel rapid Raman mapping technique was evaluated. It was shown that sufficient biochemical information for pathology diagnosis could be extracted from low signal to noise ratio data using multivariate analysis providing the dataset was sufficiently large, thus demonstrating the feasibility of automated histopathology in a clinically realistic time frame. Furthermore, it was demonstrated that high spatial resolution imaging was not necessarily required for automated histopathology using novel interpretation of multivariate techniques. A tissue classification model generated from two rapid Raman maps containing separated substrate, normal, HGD, luminescence and fibrous connective tissue with an overall training performance of 97.5% Problems limiting clinical implementation of Raman techniques were investigated and methods of overcoming devised.