Mesocosm experiments are important for studies of the pelagic ecosystem, and reproducibility of population dynamics and the effect of enclosing the water mass are essential aspects of such experiments. Nine mesocosms manipulated with daily nutrient additions were divided into three treatment groups, each of three units. All units were filled on the same day, but the three treatment groups differed in the timing of the start of nutrient additions. Abundance of phytoplankton, viruses and bacteria was followed for 12 days after filling, and 80% of the variance was ascribed to the population dynamics initiated by the nutrient perturbation, only 3% to the difference in the starting time of the perturbation and 1% to the placement of individual mesocosms at the raft. Although within-group variance of nutrient perturbation was small, there was a continuous increase over time, demonstrating a tendency for mesocosms to slowly diverge. We conclude that the transient response induced by the experimental perturbation strongly dominates the observed abundance variations and that, although present, the effects of delaying the onset of nutrient perturbation and of differences in along-raft position of the mesocosms were comparatively much smaller. Consequences for design of numerical models of microbial population dynamics are discussed.