Selectivity studies usually describe the effects on target species, whereas information on by-catch and discards is scarce. Nevertheless, large quantities of undersized individuals and invertebrates are discarded in the Mediterranean multi-species bottom trawl fishery. The present work analyses the data from two surveys carried out on the shallow and deep continental shelf (50–78 m, and 147–189 m, respectively) off the Balearic Islands (western Mediterranean). In these surveys, the traditionally used 40 mm diamond mesh codend and an experimental square mesh codend were used under commercial conditions. Catch composition, yields, size selectivity of both target and by-catch species, and discards were compared between the two mesh types. The mean selection length (L 50) clearly increased for most species when using square mesh, escaping many more individuals under their minimum landing size. Yield of Spicara smaris was significantly lower by using the square mesh, changing the composition of the retained catch. Escapement ratio and economic loss were significantly higher with square mesh, although economic loss was almost negligible for both meshes on the deep continental shelf. The use of square mesh significantly reduced the discards of algae in the shallow waters and fish on the deep continental shelf. The results confirmed that square mesh codend reduces the fishing pressure on small specimens as well as the impact of trawling on the ecosystem. These benefits would not lead to a reduction of the yields neither of the main target species, the fishes Merluccius merluccius, Mullus surmuletus, Zeus faber, and the cephalopods Loligo vulgaris and Octopus vulgaris, nor of the rest of commercial categories, except for Spicara smaris.