We develop a model of continuous spherical shapes and use it to analyze the anatomy of the hippocampus. To account for the geometry of bends and folds, the model relies on a geodesic metric that is sensitive to first-order deformations. We construct an atlas of the hippocampus as a mean shape and develop statistical models to characterize quantitative and qualitative normal shape variation. We also develop a localization tool to identify local contrasts in the anatomy of different populations. The tool is applied to the detection, characterization and visualization of anatomical differences such as local enlargement and gains in volume on the right hippocampus of blind subjects.