We measured regional cerebral metabolic rates for glucose and selected cerebral metabolites in rats fed one of the following diets for 6 to 7 weeks: (1) regular laboratory chow; (2) high-fat, carbohydrate-free ketogenic diet deriving 10% of its caloric value from proteins and 90% from fat; and (3) high-carbohydrate diet deriving 10% of its caloric value from proteins, 78% from carbohydrates, and 12% from fat. In preliminary experiments, we found that moderate ketosis could not be achieved by diets deriving less than about 90% of their caloric value from fat. Rats maintained on the ketogenic diet had moderately elevated blood beta-hydroxybutyrate (O.4 mM) and acetoacetate (0.2 mM), and a five- to 10-fold increase in their cerebral beta-hydroxybutyrate level. Cerebral levels of glucose, glycogen, lactate, and citrate were similar in all groups. 2-Deoxyglucose studies showed that the ketogenic diet did not significantly alter regional brain glucose utilization. However, rats maintained on the high-carbohydrate diet had a marked decrease in their brain glucose utilization and increased cerebral concentrations of glucose 6-phosphate. These findings indicate that long-term moderate ketonemia does not significantly alter brain glucose phosphorylation. However, even marginal protein dietary deficiency, when coupled with a carbohydrate-rich diet, depresses cerebral glucose utilization to a degree often seen in metabolic encephalopathies. Our results support the clinical contention that protein dietary deficiency coupled with increased carbohydrate intake can lead to CNS dysfunction.