Tillage erosion, the net downslope translocation of soil by tillage, was measured on eight shoulder slope landscape positions within two fields in the upland regions of south–western Ontario. Translocation of soil by tillage was measured by labelling plots of soil with ¹³⁷Cs and measuring its displacement in response to tillage. Paired plots were utilized to compare soil translocation by upslope and downslope tillage. A single sequence of conventional tillage operations, consisting of mouldboard plough, tandem disc (double pass) and C-tine cultivator, translocated 90 kg soil m⁻¹ slope width when tillage was conducted upslope, and 142 kg m⁻¹ when conducted downslope. Net downslope translocation resulting from one upslope tillage sequence and one downslope, was 52 kg m⁻¹. Assuming one sequence of tillage operations occurs per year and is conducted upslope and downslope equally often, the rate of net downslope translocation would be 26 kg m⁻¹ yr⁻¹. The source area from which this soil loss occurred was defined by an average slope length of 5.2 m between paired plots and hillslope summit. The estimated rate of soil loss from this area was in excess of 5.4 kg m⁻² yr⁻¹ (54 t ha⁻¹ yr⁻¹). Tillage erosion accounted for at least 70% of the total soil loss from natural and tillage erosion as estimated by resident ¹³⁷Cs. Key words: Tillage translocation, tillage erosion, ¹³⁷Cs, complex topography, spatial variability