The 2-deoxyglucose (2-DG) method was used to study the effect of working memory processing on local cerebral glucose utilization (LCGU) in the diencephalon of the rhesus monkey. Monkeys were given [ 14 C]2-DG while performing either one of three tasks that engaged working memory (WORK group) or one of two control tasks (CONT group) that used associative or non associative processes. The tasks of the WORK group—spatial delayed response, spatial delayed alternation, and delayed object alternation—are alike in that the information guiding a correct response changes from trial to trial and only the accurate record of the preceding response (or cue) is relevant for each successive trial. The CONT group, in contrast, performed on either a visual pattern discrimination test, in which the correct stimulus–response association was invariant across all trials and all test sessions, or on a sensorimotor task in which there was no explicit memory requirement. LCGU was examined in five diencephalic regions: the mammillary bodies, the anteroventral and anteromedial thalamus, and the parvocellular and magnocellular components of the mediodorsal thalamic nucleus. Comparisons across the two groups showed that mean LCGU in the anterior and mediodorsal thalamic nuclei was significantly elevated (by 12–16%) in the WORK group relative to the CONT group. Mean LCGU in the mammillary bodies also was higher in the WORK group than in the CONT group, but this difference was not significant. The present findings suggest that the anterior and mediodorsal thalamic nuclei represent diecephalic components of a neural network processing working memory. Together with our previous report on the enhancement of metabolic activity in the hippocampus and dentate gyrus, these results show that working memory has a wide-ranging influence on cerebral metabolism and emphasize the cooperative, rather than dissociable, roles of the hippocampus and medial thalamus in this function.