Energy and lipid metabolism as well as tumor pH in two murine tumor lines, the KHT and RIF-1 sarcomas, were studied using 31P NMR spectroscopy. Possible relationships between spectral parameters on the one hand and volume fraction of necrosis and fraction of radiobiologically hypoxic cells on the other were investigated. For both tumor lines the PCr and NTP beta resonances decreased and the Pi resonance increased significantly with increasing tumor volume in the volume range 100-4000 mm3. This decrease in bioenergetic status was accompanied by a decrease in tumor pH from about 7.2 to about 6.8. The NTP beta resonance and the tumor pH tended to be somewhat higher and the Pi resonance somewhat lower for the KHT than for the RIF-1 tumors. Linear relationships were found between tumor pH and Pi or (PCr + NTP beta)/Pi for both tumor lines (P much less than 0.05). The PME resonance increased slightly and the PDE resonance decreased slightly during tumor growth and were not significantly different for the KHT and the RIF-1 tumors. The volume fraction of necrosis was about 5 per cent in both lines at a tumor volume of 100 mm3 and increased to about 30 per cent (KHT) and 50 per cent (RIF-1) at a tumor volume of 4000 mm3. The fraction of radiobiologically hypoxic cells was found to increase from 12 to 23 per cent for the KHT line and from 0.9 to 1.7 per cent for the RIF-1 line when tumor volume was increased from about 200 to about 2000 mm3. The volume-dependence of the 31P NMR spectral parameters indicated increased nutritional deprivation and development of hypoxia and necrosis during tumor growth, and was thus qualitatively in good agreement with the changes observed in necrotic and hypoxic fraction. However, quantitative relationships between any spectral parameter and necrotic or hypoxic fraction across tumor lines were not found, implying that other physiological parameters and/or cellular characteristics may contribute significantly to a 31P NMR tumor spectrum. Consequently, 31P NMR spectra of untreated tumors have to be supplemented with other tumor data, e.g. rate of oxygen consumption, cell survival time under hypoxic stress and/or fraction of metabolically active, non-clonogenic hypoxic cells, to be useful in quantitative determination of tumor hypoxia and hence prediction of tumor radioresistance caused by hypoxia.