Histopathology remains the gold standard technique for the diagnosis of intraepithelial neoplasia (dysplasia) in Barrett's oesophagus, but it is highly subjective and relies on blind biopsy targeting. The aim of this study was to evaluate Raman spectroscopy, a rapid, non-invasive, molecular, specific analytical technique, for the objective identification and classification of Barrett's neoplasia in vitro. A secondary objective was to demonstrate the need for a rigorous gold standard in the development of new diagnostic techniques. Forty-four patients with a mean age of 69 years (range 34–89 years) undergoing surveillance for Barrett's oesophagus were included in the study. Three consultant pathologists independently assessed snap-frozen oesophageal biopsy specimens. Raman spectra were measured on 87 histopathologically homogeneous samples. Spectral classification models were developed using multivariate analysis for the prediction of pathology. Histopathology and Raman classification results were compared. Raman spectral prediction with a consensus pathology classification model gave sensitivities between 73% and 100% and specificities of 90–100%. A high level of agreement (κ = 0.89) was demonstrated between the three-subset biopsy targeting model and consensus pathology opinion. This compares favourably with the agreement measured between an independent pathologist and the consensus pathology opinion for the same spectra (κ = 0.76). Raman spectroscopy appears to provide a highly sensitive and specific technique for the identification and classification of neoplasia in Barrett's oesophagus. Copyright © 2003 John Wiley & Sons, Ltd.