Proton magnetic resonance spectroscopy was investigated as a non-invasive technique to observe biochemical changes in the brains of children who had sustained perinatal hypoxic, ischaemic and haemorrhagic brain injury. Methods: Proton spectra were acquired from the centre of the brain in premature infants, and from the parieto-occipital region in older children. Phosphorus spectra were also collected and compared with the proton spectra. Unlocalized phosphorus spectra were acquired at different repetition times. Some children had localized phosphorus spectroscopy examinations with two and four dimensional chemical shift imaging. The children were between 33 weeks post-conceptional age and four years and three months postnatal age at the time of their initial spectroscopy examination. The ability of early proton spectroscopy to predict outcome was considered in relation to the clinical neurological state at eighteen months or more. Because certain assumptions were made about the proton spectra (e.g. no T2 measurements were made), proton spectra were acquired from adults with central nervous system tumours. At surgery, biopsies were taken from the tumours and from normal brain and were analysed with in vitro spectroscopy, histology and established biochemical techniques. The metabolite ratios were compared with those from the in vivo spectra. Modifications were made to commercially available monitoring and ventilation equipment to provide the same standards of care within the magnetic field for sick patients, as on the neonatal or intensive care units. Results: All the proton spectra had peaks attributable to N-Acetyl aspartate (NAA), choline containing compounds (Cho) and creatine plus phosphocreatine (Cr). The NAA/Cho and NAA/Cr peak height ratios increased with age, while the Cho/Cr ratio decreased. The NAA/Cr ratios were significantly decreased in all children with an abnormal neurological outcome when compared with the NAA/Cr ratios from children with a normal outcome. The NAA/Cho ratios were significantly decreased in those children with a moderate outcome but not in those with a severe neurological outcome. There were no significant changes in the Cho/Cr ratios. The phosphorus spectra showed changes; phosphocreatine (PCr) to inorganic phosphate (Pi) decreased after injury and there was a marked increase in pH in the children with the poorest outcome. The apparent Tl of Pi was increased in the first month after birth in the children with a severe outcome. Few changes were seen with localized phosphorus spectroscopy in children who had focal lesions. Phosphorus spectra returned to normal within weeks of birth, while the proton spectra remained abnormal. The adult tumour proton spectra compared well with the in vitro spectra and histology of the biopsies. The concentrations changes of metabolites in vivo, were consistent with the measurements made with established biochemical techniques. Discussion: Hypoxic-ischaemic injury produced changes in the proton spectrum from neonatal brain. These changes persisted with time. Some of these changes correlated with outcome. Phosphorus spectra showed acute changes in response to injury, but the changes resolved within weeks. NAA is located in neurons; the decrease in NAA could be due to failure of neurons to develop normally, or to areas of neuronal loss and gliosis resulting from hypoxic-ischaemic damage. Phosphorus spectra may return to normal because neurons and glia have similar phosphorus metabolite ratios. Proton spectroscopy combined with magnetic resonance imaging, may become a useful technique for studying the anatomy and biochemistry of the brain in children who have suffered brain injury.