Groynes have been replaced by longitudinal training walls in an 11-km long pilot project to optimize training of the river Waal in the Netherlands. These train­ ing walls improve navigability, reduce flood levels, create a sheltered second channel with more favourable ecological conditions, and decrease the erosive action on the river bed that is responsible for large-scale bed degradation. River managers wish to assess whether longitudinal training walls could have similar advantages along other parts of the Dutch Rhine branches (without excessive increase of maintenance costs). The required maintenance dredging depends on the amount of sediment entering the sheltered channel over an entrance sill situated at the upstream edge of the longitudinal training wall. Currently operational morphodynamic models cannot reliably compute this sedi­ ment flux. We present laboratory experiments to study the passage of bed sediment at different discharge distributions between the main and sheltered channel, and different degrees of submergence.