We investigated meso- to bathypelagic (500–3,000 m) bacterioplankton assemblage composition at 19 locations in the North Atlantic Ocean beneath the offshore Amazon River plume, in the North Pacific Ocean near the Hawaiian archipelago, at the San Pedro Ocean Time Series (SPOTS) station off southern California, and in the Coral Sea off eastern Australia with a sensitive high-throughput fingerprinting approach, automated rRNA intergenic spacer analysis (ARISA), to examine variation between bacterial assemblages at different stations. Temperature and salinity were used to identify distinct water masses within gyres. ARISA fingerprints each contained 15–117 operational taxonomic units (OTUs) per assemblage; however, the OTU composition of fingerprints varied among stations, even within the same water mass. Fingerprints from 500 m at the SPOTS station over 4 yr shared on average a Sorensen Index (presence/absence similarity) of 0.68 6 0.01 and a Whittaker Index (proportional representation similarity) of 0.68 6 0.01, whereas at more oceanic stations at 500 m, fingerprints shared a Sorensen Index of 0.48 6 0.01 and a Whittaker Index of 0.38 6 0.01. At deeper depths (1,000 and 2,000 m), fingerprints were equally variable, sharing Sorensen Indices of 0.42 6 0.02 and 0.50 6 0.02 and Whittaker indices of 0.33 6 0.01 and 0.34 6 0.03, at 1,000 m and 3,000 m, respectively. Mesopelagic, moderately productive waters were more stable than those at less productive, open-ocean gyre stations, suggesting that variability in bacterioplankton communities at depth is influenced by organic matter supply and patchiness. Bacterioplankton in the world’s oceans are ecologically critical, processing typically 50% of primary production (Fuhrman and Azam 1982; Azam et al. 1983), comprising up to 70% of biomass in surface waters (Fuhrman et al. 1989), and mediating many global-scale nutrient transformations. Despite this, the richness and diversity of planktonic prokaryotic assemblages in the oceans have only recently been studied because of the advent of molecular techniques (Giovannoni et al. 1990; Fuhrman et al. 1992, 1993), which circumvent classical culture biases associated with .99% of bacterial taxa. These studies, and recent advances in molecular approaches to addressing