We examined succession of the rhizosphere microbiota of three model plants (Arabidopsis, Medicago and Brachypodium) in compost and sand and three crops (Brassica, Pisum and Triticum) in compost alone. We used serial inoculation of 24 independent replicate microcosms over three plant generations for each plant/soil combination. Stochastic variation between replicates was surprisingly weak and by the third generation, replicate microcosms for each plant had communities that were very similar to each other but different to those of other plants or unplanted soil. Microbiota diversity remained high in compost, but declined drastically in sand, with bacterial opportunists and putative autotrophs becoming dominant. These dramatic differences indicate that many microbes cannot thrive on plant exudates alone and presumably also require carbon sources and/or nutrients from soil. Arabidopsis had the weakest influence on its microbiota and in compost replicate microcosms converged on three alternative community compositions rather than a single distinctive community. Organisms selected in rhizospheres can have positive or negative effects. Two abundant bacteria are shown to promote plant growth, but in Brassica the pathogen Olpidium brassicae came to dominate the fungal community. So plants exert strong selection on the rhizosphere microbiota but soil composition is critical to its stability. microbial succession/ plant-microbe interactions/rhizosphere microbiota/selection.