The structure of mantle convection will greatly influence the generation and the survival of compositional heterogeneities. Conversely, geochemical observations can be used to obtain information about heterogeneities in the mantle and then, with certain model assumptions, information about the pattern of the mantle flow. In order to link the geochemical observations to the temporal and spatial pattern of convection it is necessary to develop an understanding of the mixing properties of convective flows. It is, for example, necessary to understand wether compositional heterogeneities can survive over significant times within a flow, or wether the presence of such heterogeneities requires a strict separation of convection cells. In the latter case the geochemical observation would indicate separate convecting layers within the Earth, while in the first case the presence of heterogeneities could be explained by a dynamic Earth model in which convection extends from the lithosphere to the boundary between mantle and core at a depth of about 3000km. It is the aim of this work to investigate the mixing properties of convective flows in order to bridge the gap between geochemical observations and the resulting implications for a dynamical model of the Earth's interior.