After its formation, a rill may remain in the field for months, often receiving lower flow rates than the formative discharge. The objective of this work was to evaluate the rill flow transport capacity of soil aggregates at discharges unable to erode the rill, and to analyse the influence of the rill macro-roughness on this transport process. A non-erodible rill was built in which roughness was reproduced in detail. In order to assess only the rill macro-roughness, a flat channel with a similar micro-roughness to that in the rill replica was built. Rill and channel experiments were carried out at a slope of 8 and at six discharges (8·3 × 10−5 to 5·2 × 10−4 m3 s−1) in the rill, and eight discharges (1·6 × 10−5 to 5·2 × 10−4 m3 s−1) in the channel. Non-erodible aggregates of three sizes (1–2, 3–5 and 5–10 mm) were released at the inlet of the rill/channel. The number of aggregates received at the outlet was registered. The number and position of the remaining aggregates along the rill/channel were also determined. The rill flow was a major sediment transport mechanism only during the formation of the rill, as during that period the power of the flow was great enough to overcome the influence of the macro-roughness of the rill bed. At lower discharges the transport capacity in the previously formed rill was significantly less than that in the flat channel under similar slope and discharge. This was determined to be due to local slowing of flow velocities at the exit of rill pools. Copyright © 2006 John Wiley & Sons, Ltd.