To investigate the relative contributions of bottom-up (plant condition) and top-down (predatory mites) factors on the dynamics of the two-spotted spider mite (Tetranychus urticae), a series of experiments were conducted in which spider mites and predatory mites were released on bean plants. Plants inoculated with 2, 4, 8, 16, and 32 adult female T. urticae were either left untreated or were inoculated with 3 or 5 adult female predators (Phytoseiulus persimilis) one week after the introduction of spider mites. Plant area, densities of T. urticae and P. persimilis, and plant injury were assessed by weekly sampling. Data were analysed by a combination of statistical methods and a tri-trophic mechanistic simulation model partly parameterised from the current experiments and partly from previous data. The results showed a clear effect of predators on the density of spider mites and on the plant injury they cause. Plant injury increased with the initial number of spider mites and decreased with the initial number of predators. Extinction of T. urticae, followed by extinction of P. persimilis, was the most likely outcome for most initial combinations of prey and predators. Eggs constituted a relatively smaller part of the prey population as plant injury increased and of the predator population as prey density decreased. We did not find statistical evidence of P. persimilis having preference for feeding on T. urticae eggs. The simulation model demonstrated that bottom-up and top-down factors interact synergistically to reduce the density of spider mites. This may have important implications for biological control of spider mites by means of predatory mites.