Various lines of evidence have been used to infer the origin of human bipedalism, but the paucity of hominoid postcranial fossils and the diversity of inferred locomotor modes have tended to confound the reconstruction of ancestral morphotypes. Examination of the bony labyrinth morphology of the inner ear of extinct and living hominoids provides independent evidence for inferring the evolution of hominoid locomotor patterns. New computed tomography data and morphometric analyses of the Late Miocene ape Lufengpithecus indicate that it and other stem great apes possess labyrinths similar to one another and show that hominoids initially evolved from a positional repertoire that included orthogrady, below-branch forelimb suspension and progression, above-branch bipedalism, climbing, clambering, and leaping (hylobatid-like) to one that comprised above-branch quadrupedalism, below-branch forelimb suspension, vertical climbing, limited leaping, terrestrial quadrupedal running and walking, possibly with knuckle walking, and short bouts of bipedalism (chimpanzee-like). The bony labyrinth morphology of Lufengpithecus indicates that it probably conforms more closely to the last common ancestors of crown hominoids and hominids in its locomotor behavior than do other Miocene hominoids. Human bipedalism evolved from this common archetypal Lufengpithecus-like locomotor repertoire. The low evolutionary rate of semicircular canal morphology suggests that Lufengpithecus experienced a relative stasis in locomotor behavior, probably due to the uplift of the Tibetan Plateau, which created a stable environment in the Miocene of southwestern China.