Although extensively investigated, the scaling-up of stirred vessels with respect to solids suspension is still a difficult matter. In view of this we carried out a series of model suspension experiments in a large stirred vessel (4.3 m in diameter). Complete and homogeneous suspensions were studied and the tests were repeated carefully at the laboratory in a small-scale vessel under identical conditions. From a comparison of the results it became clear that the scaling criterion for complete suspension is a constant specific power input. Homogeneity of the suspension is better on the large scale than on the laboratory scale. However, absolute homogeneity is unattainable in practice. Therefore the position of the outlet in a continuously stirred tank reactor can have an important effect on the solids hold-up in reactors. The thickness of the stirrer blades has an important influence on the mixing performance, hence the blade thickness of a stirrer in model experiments must be taken into account. Incorrect scaling of the blade thickness can easily cause too optimistic figures for the power consumption on a commercial scale. The wide variety in scaling rules for suspensions in the literature can be attributed to incorrect scaling of the stirrer blade thickness. Moreover, this effect can be used advantageously for the design of impeller blades. In our 80 1 vessel, a series of measurements was carried out using stirrer blades with a modified profile. Simple skewing off appeared to be an effective method of improving the performance of the stirrer or of maintaining a certain suspension quality at a lower power consumption. Savings of 17% in power consumption can easily be achieved in this way.