Using experimental data and a model to predict size selection based on morphological data, we investigated size selection of redfish (Sebastes spp.) in the grid-based selection systems used in Northeast Atlantic trawls. We found that not all redfish make physical contact with the spacing between the bars in the sorting grids while they pass in the direction of the codend, and therefore the escapement of some undersized redfish depends on size selection in the codend. We estimate that most of the escapement in the combined selection system (consisting of a 55-mm grid and a diamond mesh codend) happens through the grid. We demonstrated that for one of the two grid systems investigated the increase in size selection obtained experimentally by increasing grid-bar spacing is well in line with what could be expected based on the morphology of redfish. However, the size selection observed experimentally was significantly lower than the size-selective potential of the grids estimated based on the morphology of redfish. By computer simulations, we show that a possible explanation for this difference could be that not all redfish that attempt to escape through the grid make their attempt with an optimal angle of attack.