Separate and combined releases of P. persimilis and N. californicus were tested for the control of T. urticae on greenhouse cut roses under integrated pest management conditions. Release strategies were carried out in separate greenhouses and were not replicated. Mite densities were controlled at regular intervals on 50 single, randomly collected leaves per greenhouse. Natural infestations of 23.9 T. urticae per leaf at the start of the trial were controlled throughout the growing season by the simultaneous, combined release of 26 P. persimilis/m² and 24 N. californicus/ m². Separate releases of each phytoseiid species - 32 P. persimilis/m² at an initial infestation of 0.16 T. urticae per leaf, or 51 N. californicus/m² in total at an initial infestation of 10.2 T. urticae per leaf - failed to adequately suppress spider mite populations. Additional treatments with selective acaricides were required. Reasons for the different control success of the three release strategies and their implications for practical use are discussed.