Palaeomagnetic poles for a stable continental block are typically defined from a combination of declination and inclination information from several temporally constrained studies. Poles from regions that have undergone vertical-axis rotation as a consequence of tectonics are excluded due to the absence of declination data. These poles, however, do contain useful information in their inclinations. We develop a simple but statistically rigorous technique allowing palaeomagnetic poles to be calculated from a mixture of declination and inclination data drawn from localities in the stable continental block and inclination data from the regions disturbed by vertical-axis rotation. Together this provides a larger data set of high-quality palaeomagnetic poles from which to calculate reference poles. The technique was used to define palaeomagnetic poles for South America for the Late Cretaceous–Cenozoic (120–5 Ma) period. Data from stable, cratonic South America, combined with data from Africa, rotated into a South America reference frame, and the Andean margin, yield reference poles, as well as mean poles for the Palaeogene and Neogene. Analysis of the data reveals systematic biases in the data set, and, in particular, the fit of the inclination data is poor for most time periods. In many cases, this situation is improved if the effect of inclination shallowing due to sedimentary depositional processes and subsequent compaction is removed. The best-fit poles define an apparent polar wander path for South America that is consistent with the global plate reconstruction parameters. Use of the new poles in studies of tectonic rotation should allow greater temporal and spatial resolution of vertical-axis rotation and offer the ability to identify smaller rotations in the Andean margin.