Elucidation of the role of cerebral hyperthermia as a secondary factor that worsens outcome after brain injury, and the therapeutic application of modest brain hypothermia would benefit from noninvasive measurements of absolute brain temperature. The present study was performed to evaluate the feasibility of using 1H magnetic resonance (MR) spectroscopy to measure absolute brain temperature in human subjects on a clinical imaging spectroscopy system operating at a field strength of 1.5 T. In vivo calibration results were obtained from swine brain during whole-body heating and cooling, with concurrent measurements of brain temperature via implanted probes. Plots of the frequency differences between the in vivo MR peaks of water and N-acetyl-aspartate and related compounds (NAX), or water and choline and other trimethylamines versus brain temperature were linear over the temperature range studied (28-40 degrees C). These relationships were used to estimate brain temperature from 1H MR spectra obtained from 10 adult human volunteers from 4 cm3-volumes selected from the frontal lobe and thalamus. Oral and forehead temperatures were monitored concurrently with MR data collection to verify normothermia in all the subjects studied. Temperatures determined using N-acetyl-aspartate or choline as the chemical shift reference did not differ significantly, and therefore results from these estimates were averaged. The brain temperature (mean +/- SD) measured from the frontal lobe (37.2 +/- 0.6 degrees C) and thalamus (37.7 +/- 0.6 degrees C) were significantly different from each other (paired t-test, p = 0.035). We conclude that 1H MR spectroscopy provides a viable noninvasive means of measuring regional brain temperatures in normal subjects and is a promising approach for measuring temperatures in brain-injured subjects.