Paleomagnetic results have been obtained from a suite of Paleozoic samples from Sierra de Almeida with the high eastern portions of the Atacama desert. Characteristic directions are discussed for two sequences of pre-Silurian lavas of probable Cambro-Ordovician ages, the Late Cambrian Choscas pluton, three Late Ordovician-Early Silurian plutons, the Devonian-Carboniferous Lila Formation and late Paleozoic volcanic units of the Pular and Cas Formations. The Choschas pluton and one lava series yield similar northerly and shallow directions which for the presence of reversals and their concordance are suggested to represent early Paleozoic Arequipa plate directions. Directions in the three Late Ordovician-Early Silurian plutons pass a tilt test using an overlying erosional unconformity, and these also include reversals. Together with directions from a second lava series (a roof pendant in an Early Silurian pluton) these define poles compatible with Silurian Gondwana results from Africa and Australia. In situ directions from the basal red beds of the Devonian Lila Formation are inconsistent with Devonian Gondwana or stable South American Poles, but (like the Devonian strata of the Appalachians) they are consistent with a tilt-corrected overprint of Kiaman Superchron age. These results, together with previous results from the late Paleozoic Cas and Pular Formations are discordant from the Gondwana path only for the latest Cambrian-earliest Ordovician. The discordance in paleomagnetic data, together with regional geologic constraints, can be explained by a model in which the Arequipa block, representing a paraautothonous finger of Gondwana (like Japan or the Iberian Peninsula) rotated about a nearby pole but was then resutured during the Silurian. Such a scenario resolves much of the discrepancies in the models which have emerged from Peru, Chile, and Argentina. ¿ American Geophysical Union 1993