In this paper we present a comparison between Remotely Operated Vehicle (ROV) and Agassiz trawling methods for sampling deep-water fauna in three submarine canyons of the Northwestern Mediterranean Sea and describe the behavioural reactions of fishes and crustacean decapods to ROV approach. 10 ROV dives, where 3583 individuals were observed and identified to species level, and 8 Agassiz trawls were carried out in a depth range of 750-1500 m. As noticed in previous studies, abundances of fishes and decapod crustaceans were much higher in the ROV videos than in Agassiz trawl samples, as the latter are designed for the retrieval of benthic, less motile species in permanent contact with the bottom. In our observations fish abundance was one order of magnitude higher with ROV (4110.22 ind/km2) than with Agassiz trawl (350.88 ind/km2), whereas decapod crustaceans were six times more abundant in ROV videos (6362.40 ind/km2) than in Agassiz samples (1364.52 ind/km2). The behaviour of highly motile fishes was analysed in terms of stationary positioning over the seafloor and avoidance or attraction to ROV approach. The most frequently occurring fish species Coelorinchus mediterraneus, Nezumia aequalis, Bathypterois dubius, Lepidion lepidion, Trachyrincuss scabrus and Polyacanthonotus rissoanus did not react to the presence of the ROV in most cases (>50%). Only B. dubius (11%), Lepidion lepidion (14.8%), P. rissoanus (41%) and T. scabrus (14.3%) reacted to ROV approach. More than 60% of less motile species, such as crustacean decapods, did not respond to ROV presence either. Only 33.3% of Geryon longipes, 36.2% of Munida spp. and 29.79% of Pagurus spp. were observed avoiding or defensively reacting to the ROV. The comparison of results obtained with ROV and trawl sampling is of ecological relevance since ROV can report observations in areas where trawling is technically unfeasible. The lack of reaction by most fish and crustacean decapod specimens further confirms that ROV surveying is an efficient technique to assess abundance and species composition in deep-sea waters for these motile species and questions trawl-based estimations.