Water models using perspex have been built to study the fluid flow and recirculation patterns developed in the sump of a steel continuous casting machine and the influences these have on the behaviour of inclusions. An experimental method has been devised to simulate the behaviour of inclusions in the sump and to study the apportionment of the input flux of inclusions between the molten mould powder layer and the strand. The method entails the uses of finely dispersed coloured paraffin oil in the inlet stream together with a floating colourless paraffin layer on the top of the water in the model mould to simulate the molten powder layer on top of the molten steel. A theoretical model has been formulated which relates the inclusion separation in the sump to the fluid flow there. The inclusion removal ratio in the sump for a given continuous casting machine can be predicted using this theoretical model. The model, using the properties of liquid steel and practicable casting speeds, demonstrates that the removal of inclusions of small size (< 40 um) from the mould sump is less than 5% efficient. Inclusion agglomeration plays an important role in inclusion removal. It has been shown that deep submersion of the SEN enhances the agglomeration of inclusion particle. Under certain conditions, for example, the average particle diameter in the meniscus region has been found to be as much as three times its value at the SEN nozzle. The use of fine alumina flakes or small air bubbles, together with a plane light source, has been found to be very successful in studying the fluid flow patterns developed in three-dimensional models. Employing this method, the fluid flow patterns developed on different planes within the model mould have been viewed and recorded photographically. The photographs so obtained have helped to explain the results obtained for the removal of inclusions. The fluid flow patterns developed when small outside diameter nozzles with deep SEN submerged depths are used have been found to be of benefit to the removal of inclusions. Increasing the SEN submerged depth promotes inclusion agglomeration and hence increases the inclusion removal ratio. Reducing the nozzle outside diameter and the casting speed increases the inclusion removal ratio in the sump. But the infleunces of these latter changes are not very strong, so that inclusion removal consideration need not influence the design strategies used for the casting speed and nozzle outside diameter. The SEN port angle has a little effect on the inclusion removal when using deep SEN submerged depth. Although argon stream introduced into the tundish nozzle stream can protect the nozzle blockage, it is not beneficial to the inclusion removal in the sump.