We studied 26 infants (1-18 months old) and 27 children (18 months or older) with acute nonaccidental (n = 21) or other forms (n = 32) of traumatic brain injury using clinical rating scales, a 15-point MRI scoring system, and occipital gray matter short-echo proton MRS. We compared the differences between the acutely determined variables (metabolite ratios and the presence of lactate) and 6- to 12-month outcomes. The metabolite ratios were abnormal (lower NAA/Cre or NAA/Cho; higher Cho/Cre) in patients with a poor outcome. Lactate was evident in 91% of infants and 80% of children with poor outcomes; none of the patients with a good outcome had lactate. At best, the clinical variables alone predicted the outcome in 77% of infants and 86% of children, and lactate alone predicted the outcome in 96% of infants and 96% of children. No further improvement in outcome prediction was observed when the lactate variable was combined with MRI ratios or clinical variables. The findings of spectral sampling in areas of brain not directly injured reflected the effects of global metabolic changes. Proton MRS provides objective data early after traumatic brain injury that can improve the ability to predict long-term neurologic outcome.