Plants are colonized inside and out by a diverse array of microbes. These inhabitants can have pathogenic, beneficial, or commensal relationships with the plant. Historically, studies have focused on individual microbes and their effects on plant phenotypes and fitness; however, recent evidence points towards a microbial community context for these interactions. To fully describe host-pathogen interactions, we need to better understand not just the direct interactions between host and microbe, but also between microbe and microbe in the host-associated microbiota. Greater understanding of the ecological forces that structure these communities will lead to effective strategies for promotion of plant health. To uncover these forces, I have studied the bacterial communities associated with Arabidopsis thaliana and taken a multi-faceted approach involving observation of controlled and natural communities. In Chapter 2, I describe the benefits and pitfall of three different approaches for quantifying mixed species communities in vitro and in planta. In Chapter 3, I characterize the temporal dynamics and predictability of communities across two axes of complexity, environmental heterogeneity and species richness. In Chapter 4, I examine the spatial and temporal dynamics of communities associated with natural A. thaliana. Taken together these chapters demonstrate remarkable disparities in the ability to predict bacterial community succession under controlled versus field conditions. For mixtures of endophytic leaf isolates in vitro, I observed large dissimilarities in community compositions between replicates, even given a homogenous liquid media environment. In contrast, for field grown plants, I observed consistent differences in community composition by tissue that followed similar successional trajectories within tissue over two years at two sites. This predictability discrepancy is likely due to differences in the types, timing, and number of selective forces in controlled versus field environments. Under controlled conditions, bacteria encountered rich liquid or solid media or gnotobiotic plants rich in unoccupied niches. Bacterial competition for nutrients was likely the main force structuring the community, with variable outcomes for each replicated community. In contrast, bacterial communities associated with field grown plants followed remarkably repeatable successional patterns. Communities associated with vegetative roots initially diverged from soil communities in terms of composition, and upon flowering and senescence, root communities became more soil-like. Comparison of operational taxonomic unit (OTU) profiles between tissues indicated that the roots first filter OTUs from the surrounding soil that become differentially enriched in each aboveground tissue. The natural environment imposes a variety of selective forces on bacterial community composition, including fluctuations in abiotic conditions, host metabolite and defense factors, and interspecific competition. Based on my in vitro findings competitive outcomes can be difficult to predict a priori; however, host and environmental forces seem to impose stronger and more consistent selection on community composition. These findings have implications for the use of pro and prebiotics for microbiome manipulations. Given that bacterial competition is difficult to predict even under the simplest of conditions, introducing probiotic strains to compete with native strains is likely to be ineffective. Instead, selecting for the growth of native, desired taxa by specific nutrient addition may be more successful. Prebiotic strategies are more in line with natural processes in which the plant produces nutrient and defense metabolites that select for specific taxa. Manipulating microbiomes in the soil prior to planting is crucial. The roots will filter strains from the initial soil community that will then colonize the entire plant.