More than 50% of cancers fail to respond to any individual treatment and tumour follow-up after treatment plays a major role in routine therapy planning and pharmacological research. Today, MRS is the only technological approach providing non-invasive access to tumour biochemistry. Ten years ago, expectations were raised concerning 31P MRS as an exciting and promising technical approach to the study of tumours. However the expectations have not always come to fruition. How close are we now to seeing routine 31P NMR in clinical oncology? This review of the 127 published papers shows spectroscopy results in more than 150 experimental animal tumour models. These tumour/host/treatment systems provide us with a useful basis to evaluate the current state of the art, summarize the basic knowledge presently available, determine the key points underlying the present disappointment of some clinical oncologists and stimulate new basic research. The information collected concerns the discussion of the reliability of experimental models in oncology, the technical improvement of magnetic resonance technology and the monitoring of bioenergetic status, pH regulation and phospholipid metabolism in treated and untreated tumours. Recent advances (two-thirds of the papers have been published in the last 5 years) seem to provide more optimistic perspectives than those generally accepted a few years ago, in the depressing period following early pioneering work.