General linear models were used to compare bacterial growth across microcosms. For the microcosm growth experiments, the response variable was either the biomass (measured by optical density) or the log10 of the viable colony counts. The factors considered were time (weeks 1 to 4), microcosm composition (mono-, di- and tri-cultures), and their interaction. For the growth assay of evolved species experiment, the response variable was the log10 of the viable colony counts for each of the three constituent species and the factors considered were the strain inoculum origins (ancestral strains and re-isolated strains from microcosms). Post hoc testing was conducted using the Bonferroni pairwise comparisons method (95 % confidence level). Calculations were done in Minitab 17 (Minitab 17 Statistical Software, 2010).