The goals of this study were to evaluate 31P MR spectroscopic imaging (MRSI) for clinical studies and to survey potentially significant spatial variations of 31P metabolite signals in normal and pathological human brains. In normal brains, chemical shifts and metabolite ratios corrected for saturation were similar to previous studies using single-volume localization techniques (n = 10; pH = 7.01 +/- 0.02; PCr/Pi = 2.0 +/- 0.4; PCr/ATP = 1.4 +/- 0.2; ATP/Pi = 1.6 +/- 0.2; PCr/PDE = 0.52 +/- 0.06; PCr/PME = 1.3 +/- 0.2; [Mg2+]free = 0.26 +/- 0.02 mM.) In 17 pathological case studies, ratios of 31P metabolite signals between the pathological regions and normal-appearing (usually homologous contralateral) regions were obtained. First, in subacute and chronic infarctions (n = 9) decreased Pi (65 +/- 12%), PCr (38 +/- 6%), ATP (55 +/- 6%), PDE (47 +/- 9%), and total 31P metabolite signals (50 +/- 8%) were observed. Second, regions of decreased total 31P metabolite signals were observed in normal pressure hydrocephalus (NPH, n = 2), glioblastoma (n = 2), temporal lobe epilepsy (n = 2), and transient ischemic attacks (TIAs, n = 2). Third, alkalosis was detected in the NPH periventricular tissue, glioblastoma, epilepsy ipsilateral ictal foci, and chronic infarction regions; acidosis was detected in subacute infarction regions. Fourth, in TIAs with no MRI-detected infarction, regions consistent with transient neurological deficits were detected with decreased Pi, ATP, and total 31P metabolite signals. These results demonstrate an advantage of 31P MRSI over single-volume 31P MRS techniques in that metabolite information is derived simultaneously from multiple regions of brain, including those outside the primary pathological region of interest. These preliminary findings also suggest that abnormal metabolite distributions may be detected in regions that appear normal on MR images.