ABSTRACT: Marine bacterioplankton dominate microbial carbon biomass in surface waters of the oligotrophic ocean, yet there have been few studies examining rates of change in bacterioplankton assemblage composition in situ over time and across water masses. Temporal changes in bacterioplankton assemblage composition were investigated during 7 drifter studies of 24 to 360 h duration in the oligotrophic Gulf of Mexico, the North Pacific and the West Tropical Atlantic in 2001 to 2003, using an assemblage fingerprinting technique, automated rRNA intergenic-spacer analysis (ARISA). The similarity indices between assemblages collected over time in the same drifter changed on average by a Sørensen index of 0.12 d^-1 (comparing the presence/absence of operational taxonomic units, OTU) and a Whittaker index (comparing proportions in various OTU) of 0.17 d^-1 per fingerprint, across all surface drifter studies. Fingerprints generated from 7 replicate bacterioplankton DNA samples collected at each of 2 stations were remarkably similar to each other, sharing a Whittaker index >0.85. Despite this consistency over small spatial scales (<2 km), no clear relationship was observed between the separation distance of sampling locations and similarity between assemblage fingerprints in the North Pacific and Atlantic gyres over mesoscales (10 to 3000 km), oscillating around a mean of 0.38 to 0.47 for each gyre. Our results suggest that factors ultimately shaping assemblage composition are localized at spatial scales between a few kilometers and about 50 km, i.e. this is a typical horizontal 'patch size', within which communities are relatively homogenous, perhaps because physical mixing (e.g. by eddies) may dominate over biological interactions. Our results also suggest that, while similar environmental factors may cause bacterioplankton assemblages to share a small portion of OTU between different sampling locations, biological factors selecting for particular bacterial types (e.g. viral lysis, grazing, antagonism, nutrition) may cause assemblage composition variability over short geographic distances.