Somatosensory discrimination of unseen objects relies on processing of proprioceptive and tactile information to detect spatial features, such as shape or length, as acquired by exploratory finger movements. This ability can be impaired after stroke, because of somatosensory-motor deficits. Passive somatosensory discrimination tasks are therefore used in therapy to improve motor function. Whereas the neural correlates of active discrimination have been addressed repeatedly, little is known about the neural networks activated during passive discrimination of somatosensory information. In the present study, we applied functional magnetic resonance imaging (fMRI) while the right index finger of ten healthy subjects was passively moved along various shapes and lengths by an fMRI compatible robot. Comparing discriminating versus non-discriminating passive movements, we identified a bilateral parieto-frontal network, including the precuneus, superior parietal gyrus, rostral intraparietal sulcus, and supramarginal gyrus as well as the supplementary motor area (SMA), dorsal premotor (PMd), and ventral premotor (PMv) areas. Additionally, we compared the discrimination of different spatial features, i.e., discrimination of length versus familiar (rectangles or triangles) and unfamiliar geometric shapes (arbitrary quadrilaterals). Length discrimination activated mainly medially located superior parietal and PMd circuits whereas discrimination of familiar geometric shapes activated more laterally located inferior parietal and PMv regions. These differential parieto-frontal circuits provide new insights into the neural basis of extracting spatial features from somatosensory input and suggest that different passive discrimination tasks could be used for lesion-specific training following stroke.