Escapement through the body or under the footrope of the sampling trawl used for the Mediterranean international trawl survey programme (MEDITS) was assessed i) by means of visual observations performed with a towed operating vehicle equipped with a high-sensitivity video camera, and ii) by attaching small bags to the outside of the trawl body and under the footrope. Due to the small size of most of the individuals present in the study area and to the towing speed, fish behaviour in front of the trawl could not be frequently observed. In contrast, their behaviour was easily observed inside the net, especially upstream and inside the codend. Some species, such as horse mackerel (Trachurus sp.), showed greater swimming endurance compared with others such as annular seabream (Diplodus annularis). In three fishing trips conducted in different seasons in 1996 and 1997 to obtain data on several species in different areas of the Adriatic, six small net bags were mounted outside the body of the MEDITS trawl to quantify escapement through the trawl belly and under the footrope. Significant escapement of many species occurred in the near-bottom part of the side panels. Norway lobster (Nephrops norvegicus) had the highest rate (64 %), mainly represented by small-size individuals. The escapement of greater forkbeard (Phycis blennoides) was also high (53 %). For hake (Merluccius merluccius), common pandora (Pagellus erythrinus) and poor cod (Trisopterus minutus capelanus), rates ranged from 10 to 16 %. The escapement of red mullet (Mullus barbatus) was very low despite their great abundance. Blue whiting (Micromesistius poutassou) escaped mainly through the mid-height part of the side panel. This great variability of escapement values could affect the proportion rates of the species sampled during a standard survey. Different escape behaviours among size classes were observed for M. poutassou, T. minutus capelanus and P. erythrinus, with smaller individuals tending to escape towards the bottom and larger ones towards the higher parts of the trawl body. In the species where the escape rate was size-dependent, therefore, the size-frequency distributions obtained from the codend catch may not reflect actual size-frequency distributions.