In this study, we compared the size selectivity of a T90 codend (nettingturned by 90 degrees) with that of a standard codend made of similar netting. Sea trialswere conducted in a Norway lobster directed fishery in the Kattegat-Skagerrak area, wherethere is a need for improved selectivity because of a severe discard problem. The codendswere tested by fishing simultaneously with them in a twin trawl rig. Codend covers mountedwith Minisamplers were used, which made it possible to catch individuals escaping duringtowing and haul-back separately. Herein we proposed a model to assess the sequentialselection during towing and haul back. This model takes into account the parameterCtow, which can be interpreted as theproportion of fish that comes into contact with the codend meshes during towing and,thereby, has a chance of escape. Compared to the standard codend, the T90codend retained fewer Norway lobster both below and above the legal minimum landing size(40 mm, cephalothorax length), thereby causing a reduction of commercial catch. Thedifference was mainly due to a significantly higher escape rate during towing for theT90 codend. For plaice below minimum landing size (27 cm), theretention was slightly but significantly higher for the T90 codendcompared to the standard codend. A model developed for both codends showed that not allplaice are able to attempt escapement during the towing process. For cod, the resultsindicated an increased L50 (the length at which 50% of this species iscaught) for the T90 codend, but the effect was not statisticallysignificant, probably due to the limited number of cod retained during the sea trials. Theresults demonstrated that, for all three species, a significant proportion did escapeduring haul-back in both codends.