Test–retest reliability of resting regional cerebral metabolic rate of glucose (rCMR) was examined in selected subcortical structures: the amygdala, hippocampus, thalamus, and anterior caudate nucleus. Findings from previous studies examining reliability of rCMR suggest that rCMR in small subcortical structures may be more variable than in larger cortical regions. We chose to study these subcortical regions because of their particular interest to our laboratory in its investigations of the neurocircuitry of emotion and depression. Twelve normal subjects (seven female, mean age 5 32.42 years, range 21–48 years) underwent two FDG-PET scans separated by approximately 6 months (mean 5 25 weeks, range 17–35 weeks). A region-of-interest approach with PET-MRI coregistration was used for analysis of rCMR reliability. Good test–retest reliability was found in the left amygdala, right and left hippocampus, right and left thalamus, and right and left anterior caudate nucleus. However, rCMR in the right amygdala did not show good test–retest reliability. The implications of these data and their import for studies that include a repeat-test design are considered.