Intraspecific trait change, including altered behaviour or morphology, can drive temporal variation in interspecific interactions and population dynamics. In turn, variation in species’ interactions and densities can alter the strength and direction of trait change. The resulting feedback between species’ traits and abundance permits a wide range of community dynamics that would not be expected from ecological theories purely based on species abundances. Despite the theoretical importance of these interrelated processes, unambiguous experimental evidence of how intraspecific trait variation modifies species interactions and population dynamics and how this feeds back to influence trait variation is currently required. We investigate the role of trait-mediated demography in determining community dynamics and examine how ecological interactions influence trait change. We concurrently monitored the dynamics of community abundances and individual traits in an experimental microbial predator–prey–resource system. Using this data, we parameterised a trait-dependent community model to identify key ecologically relevant traits and to link trait dynamics with those of species abundances. Our results provide clear evidence of a feedback between trait change, demographic rates and species dynamics. The inclusion of trait–abundance feedbacks into our population model improved the predictability of ecological dynamics from r2 of 34% to 57% and confirmed theoretical expectations of density-dependent population growth and species interactions in the system. Additionally, our model revealed that the feedbacks were underpinned by a trade-off between population growth and anti-predatory defence. High predator abundance was linked to a reduction in prey body size. This prey size decrease was associated with a reduction in its rate of consumption by predators and a decrease in its resource consumption. Modelling trait–abundance feedbacks allowed us to pinpoint the underlying life history trade-off which links trait and abundance dynamics. These results show that accounting for trait–abundance feedbacks has the potential to improve understanding and predictability of ecological dynamics.