Size selective grids were introduced into the Barents Sea demersal trawl shery in 1997. It was believed at the time that in comparison to standard diamond mesh codends, grids had better selective properties, i.e. narrower selection range. Few studies have directly compared the selective properties of the combined grid and codend with that of the codend only. In this study we describe two experiments carried out in 2002 and 2003, where we directly compared the properties of codend selectivity for cod (Gadus morhua ) with the combined selectivity of a Sort-V grid and a codend. In 2002 we used small-meshed covers, whereas the twin trawl method was used in 2003. In both experiments we compared the selectivity of a 135 mm codend only with that of a trawl tted with a 135 mm codend and a 55 mm grid, which is the mandatory con guration. In 2003 we also estimated the selectivity of a 155 mm diamond mesh codend, having an L 50 similar to that of the grid and 135 mm codend combination. The results presented no evidence that the grid and mesh combination had sharper size selection than codend meshes. The introduction of the mandatory use of grid in the shery in 1997 therefore only increased L 50 , and corresponded to a mesh size increase of around 20 mm. Mean selection length of the grid was inversely related to catch rates in the 2002 experiment when large catches were taken. Grid selection appeared less a ected by seasonal variations in degree of stomach fullness or condition than mesh selectivity. Other situations where grid selection might perform better than mesh selection are discussed.