The diversity and distribution of gelatinous zooplankton were investigated along the northern Mid-Atlantic Ridge (MAR) from June to August 2004.Here, we present results from macrozooplankton trawl sampling, as well as comparisons made between five different methodologies that were employed during the MAR-ECO survey. In total, 16 species of hydromedusae, 31 species of siphonophores and four species of scyphozoans were identified to species level from macrozooplankton trawl samples. Additional taxa were identified to higher taxonomic levels and a single ctenophore genus was observed. Samples were collected at 17 stations along the MAR between the Azores and Iceland. A divergence in the species assemblages was observed at the southern limit of the Subpolar Frontal Zone. The catch composition of gelatinous zooplankton is compared between different sampling methodologies including: a macrozooplankton trawl; a Multinet; a ringnet attached to bottom trawl; and optical platforms (Underwater Video Profiler (UVP) & Remotely Operated Vehicle (ROV)). Different sampling methodologies are shown to exhibit selectivity towards different groups of gelatinous zooplankton. Only ~21% of taxa caught during the survey were caught by both the macrozooplankton trawl and the Multinet when deployed at the same station. The estimates of gelatinous zooplankton abundance calculated using these two gear types also varied widely (1.4 ± 0.9 individuals 1000 m -3 estimated by the macrozooplankton trawl vs. 468.3 ± 315.4 individuals 1000 m -3 estimated by the Multinet (mean ± s.d.) when used at the same stations (n = 6). While it appears that traditional net sampling can generate useful data on pelagic cnidarians, comparisons with results from the optical platforms suggest that ctenophore diversity and abundance are consistently underestimated, particularly when net sampling is conducted in combination with formalin fixation. The results emphasise the importance of considering sampling methodology both when planning surveys, as well as when interpreting existing data.