This thesis investigates controls on and the nature of three-dimensional flow structures at river channel confluences. Previous work has shown that junction angle is an important control on the flow patterns at channel junctions since it affects the degree of curvature of flow from each tributary and sets up secondary circulation cells similar to those in meander bends. Recent work has highlighted the common occurrence of, and importance of, bed discordance at river confluences due to a significant difference in bed levels of one or both tributaries compared to the post-confluence bed level. In some cases, particularly where only one tributary is shallower, bed discordance appears to inhibit the formation of secondary circulation cells. In other cases, particularly those where bed discordance is due to local scour in the centre of the confluence, one or more cells have been identified. Thus, it is clear that both junction angle and bed discordance are important controls, but not how they interact or their relative effect under different circumstances. Such an investigation reqUIres detailed flow information in a wide variety of situations, difficult to obtain in either the laboratory or field. This thesis uses a three-dimensional numerical model, a relatively new tool in geomorphology, to investigate these questions. For example, a set of systematic experimental simulations in asymmetrical laboratory-style channels in which bed discordance was represented by making the angled tributary shallower than the main channel showed: (i) The secondary circulation pattern 111all experiments was characterised by a single, dominant cell on the side of the angled, shallower tributary, with upwelling along this bank. (ii) The presence of bed discordance acted to reinforce this helical circulation, not to destroy it. (iii) A comparable increase in the secondary circulation intensity was produced across a characteristic range of values of bed discordance, junction angle and velocity ratio. (iv) The secondary circulation in the absence of bed discordance, was driven by a cross-stream pressure gradient at the bed due to water surface slopes related to flow curvature and inertia. A zone of low pressure forms in the lee of a bed step which significantly enhances the pressure gradient due to flow curvature for even small degrees of bed discordance, thus reinforcing the circulation pattern. As well as describing flow structures, successful application of a sophisticated numerical model to confluence flow structures also allows the dynamics of the flow in both laboratorystyle experiments, and field applications, to be explained by consideration of pressure gradients. This enhanced understanding allows more general conclusions, for example, it is suggested in this thesis that a more critical approach is required to the commonly-made analogy between confluences and 'back-to-back' meanders: although it may be appropriate in Y-shaped (symmetrical confluences), in more asymmetrical confluences (the norm), it can only be applied over a very short distance as streamline curvature from the two tributaries in opposite directions cannot be maintained. Use of a numerical model has also allowed preliminary investigations into two other aspects of confluence flow: 1. Periodic flow characteristics, such as periodic upwelling of flow from a deeper channel within that from a shallower channel at a confluence with bed discordance, are often observed. Simulation of this phenonomen shows it to be related to fluctuations in the size and shape of the low pressure zone in the lee of the bed step, and therefore to fluctuating pressure gradients. This is important for mixing processes between the two water bodies. These results also suggest that the distinction between 'permanent' secondary circulation and transient turbulent eddies may not be as clear as often implied. 2. Examination of bed shear stress predictions indicate some implications of particular flow patterns for sediment transport and morphological change. For example, the comparison of model predictions for a field confluence with and without a scour hole suggests that: (ii)The presence of the scour hole promotes downwelling in this region, rather than downweIling promoting scour. (iii) The initiation of scour is probably associated with flow acceleration through the confluences and associated increase in turbulence levels. (iv) The formation and persistence of a scour hole and lateral bar, two characteristic features of confluence morphology, are closely linked. Future research should continue to address the implications of the flow dynamics at rIver confluences for mixing of the two fluids, and for morphological change.