With magnetic resonance spectroscopic imaging (MRSI), it is possible to simultaneously map distributions of several brain metabolites with relatively good spatial resolution in a short time. Although other functional imaging modalities have taken advantage of population-based inferences using spatially extended statistics, this approach remains little utilized for MRSI. In this study, statistical nonparametric mapping (SnPM) was applied to two-dimensional MRSI data from the medial walls of the human brain to assess the effect of normal aging on metabolite concentrations. The effects of different preprocessing steps on these results were then explored. Short echo time MRSI of left and right medial walls was acquired in conjunction with absolute quantification of total choline, total creatine (tCr), glutamate and glutamine, myo-inositol, and N-acetyl-aspartate. Individual images were spatially warped to a common anatomical frame of reference. Age effects were assessed within SnPM as were the effects of voxel subsampling, variance smoothing, and spatial smoothing. The main findings were: (1) regions in the bilateral dorsal anterior cingulate and in the left posterior cingulate exhibited higher tCr concentrations with age; (2) voxel subsampling but not spatial smoothing enhanced the cluster-level statistical sensitivity; and (3) variance smoothing was of little benefit in this study. Our study shows that spatially extended statistics can yield information about regional-specific changes in metabolite concentrations obtained by short echo time MRSI. This opens up the possibility for systematic comparisons of metabolites in the medial wall of the brain.