The origin of the rotations detected paleomagnetically in the Central Andes is controversial. Tectonic models proposed to explain it involve Late Cenozoic oroclinal bending (the Bolivian Orocline), and/or small-block rotations driven by oblique subduction. In this paper, we report paleomagnetic data from upper Miocene ignimbrites from the Puna, in the presumed southern limb of the orocline. These rocks showed high unblocking temperature and high coercivity magnetization. The resulting pole position (latitude 85.7¿S, longitude 80.5¿E, A95=7.9¿, K=22, N=17) indicates that the study area has not been significantly affected by vertical-axis rotation since late Miocene onward. This implies that the whole Puna has not undergone significant regional rigid-body rotation during the same time span. A farther analysis based on Neogene paleomagnetic data from 31 Andean localities distributed from 11¿S to 31¿S shows the presence of rotated and unrotated areas. Systematic counterclockwise rotations in Per¿ and northern Bolivia are observed, whereas clockwise rotations are present in southern Bolivia, northern Chile, and perhaps northwestern Argentina. No systematic rotations are found farther south. The overall time-spatial distributions of these Neogene paleomagnetic data suggest that if orogenic bending occurred, it must have taken place before the middle Miocene. Thus the oroclinal hypothesis can not explain the rotations observed in middle Miocene rocks or younger ones, which also show orocline-like declination anomalies. We suggest that small-block rotations driven by distributed shear may be a single process that can account for the whole rotations in the Central Andes. It is hypothesized that horizontal shear in the Andean crust could be controlled by ancient structures. ¿ American Geophysical Union 1996