Organometallic complexes can be introduced onto metal oxide supports either by direct bonding or via an intervening ligand. The former method uses the diverse reaction sites available on oxide surfaces. The chemisorption processes may be considered in the same way as reactions of organometallic complexes in solution. For example, interaction of Rh(allyl)3 with silica causes an electrophilic cleavage of one metal-allyl bond to generate a new metal centre, viz.[Si]-ORh(allyl)2' Alternatively, interaction of RU3(CO)12 with silica proceeds via an oxidative addition reaction of a silanol group across one of the metal-metal bonds. Nuclearity is not always maintained during anchoring procedures, as in the case of the interaction of Rh4(CO)12 with alumina, in which mononuclear metal centres are produced. Some of the coordination centres derived from these grafted organometallic complexes have close analogues in mainstream organometallic chemistry; others appear to have no discrete counterpart at present and offer a challenge to the synthetic chemist. Organometallics also provide alternative routes to supported metal particles. In some instances these may be formed under much milder conditions than conventional reductions of metal salts. By virtue of the differing kinetics involved, different (often smaller) particle sizes, compositions, and size distributions become available. Ligandtethered complexes may be synthesised with high specificity. However, subsequent surface reactions occur to give new coordination centres. If the aim is to anchor a known homogeneous catalyst, then precautions must be taken to avoid these, either by protective silylation or by design of a stabilising tethering ligand. But the original tethered complex may be used as a catalyst precursor, and so these surface reactions may be of interest. Many new catalysts have been prepared by supporting organometallic complexes on oxide supports. Establishing the metal coordination site requires a combination of total product analysis and direct spectroscopic measurements characterizing the supported complex, preferably aided by reference to close model compounds. This is now considerably easier with the range of techniques presently available. For a high proportion of materials, the oxide surface binding site is not understood. Progress in this is necessary 47