Localized proton NMR spectroscopy in vivo allows focal studies of cerebral metabolites in both man and laboratory animals from image-defined regions as small as 1 mL or 64 microL, respectively. Although brain tumours lead to remarkable spectral alterations relative to normal brain, a number of problems may compromise the interpretation of the results. Potential complications arise from the chosen experimental conditions (method, TE, size and location of volume of interest), from regional metabolic heterogeneity in and around tumours, from differences between human tumours and animal models, and from discrepancies between in vivo and in vitro findings. Strategies and pitfalls are illustrated with use of selected examples from primary brain tumours, a rat tumour model and perchloric acid extracts of resected specimens.