Soil mixing plays a significant role in contaminant transport, carbon sequestration, and landscape evolution, yet the rates and driving mechanisms are poorly constrained. Here we use depth profiles and advection-diffusion modeling of fallout nuclides to quantify differences in short-term (<100 yr) physical soil mixing across contrasting landscapes. We constrain advection in soils using the distribution of cosmogenic 7Be and weapons-derived isotopes, and quantify mixing with a steady-state model of vertical 210Pb transport. On a forested landscape in the Bega Valley in southeastern Australia and on grasslands in Marin County, California, where bioturbation is documented as the dominant sediment transport mechanism, we calculate diffusion-like mixing coefficients of 1–2 cm2 yr−1. In montane forest soils of northern New England, we observe little field evidence of short-term mixing, and find that the traditional advection-diffusion model fails to describe 210Pb profiles. Because nuclide profiles here can be described with a simple model of litterfall, organic matter decay, and radioactive decay, we argue that diffusion-like processes are barely active on short time scales, and that the advection-diffusion model overestimates diffusion-like transport. While animal bioturbation and soil freezing cycles have little effect on the fate of elements in New England, physical soil mixing drives transport at Bega Valley and Marin County. We suggest that the absence of soil stirring that we quantify in New England forests may explain the slow physical erosion here (∼0.2 cm/k.y.) relative to the actively bioturbated soils of Bega Valley and Marin County (5–10 cm/k.y.).