Using echolocation, bats can not only locate objects in space but also discriminate objects of different shape. The acoustic image of an object is its impulse response (IR). The current experiments investigate whether bats just perceive changes in echo composition or whether bats perceive the IR itself through a detailed comparison of the emitted sound with the echo. The bat Megaderma lyra was trained to classify unknown virtual objects according to learned reference objects of different temporal and spectral composition. The bats' spontaneous classification was compared to predictions based on variousphysical and simulated peripheral auditory representations of the objects. The results show that the bats developed an accurate internal representation of the objects' IRs. In the auditory periphery, the IRs of small objects (< 4-6 cm) are coded along the tonotopic frequency axis.