Melt inclusions in phenocrysts are a potentially powerful tool in petrological research that can provide the only direct information available on the physical parameters ( P, T and melt composition) of crystallisation at various stages in the evolution of magmatic systems. However, melt inclusions also differ in principle from other parts of the magmatic system in that their composition, after trapping, may be controlled by the composition of the host phenocryst and therefore the direct application of our understanding of macro-scale magmatic processes to the interpretation of melt inclusion data can lead to erroneous conclusions. Our results indicate that the compositions of melt inclusions in early formed phenocrysts (olivine, pyroxene, plagioclase and spinel), often of most interest in petrological studies, can be affected by processes such as volatile dissociation, oxidation and/or partial re-equilibration with their host, both during natural cooling and homogenisation experiments. In particular, melt inclusions in all minerals are prone to hydrogen diffusion into or out of the inclusions after trapping and prior to eruption, and during homogenisation experiments. If not taken into account, this can significantly affect the crystallisation temperatures derived from the homogenisation experiments. Melt inclusions in highmagnesian olivine phenocrysts commonly have lower Fe contents compared to the initially trapped composition due to reequilibration with the host at lower temperatures. This often leads to the appearance of sulphide globules and in some cases high-magnesian clinopyroxene daughter crystals, and may cause an increase in the oxidation state of the inclusions. Homogenised melt inclusions in plagioclase phenocrysts in MORB usually have lower Ti and Fe, and higher Si contents compared to the melt composition at the moment of trapping. However, homogenisation experiments can provide reliable estimates of trapping temperature and the MgO, Al 2 O 3 , CaO, Na 2 O, and K 2 O contents of the host magma at the moment of trapping. Some of these processes can be identified by observing the behaviour of melt inclusions during homogenisation experiments using low-inertia visually controlled heating stages, and their effects can be minimised by using appropriate experimental conditions as determined by kinetic experiments, ideally completed for each phenocryst type in every sample. We also discuss general aspects of melt inclusion studies aimed at recovering H 2 O content of primary mantle-derived magmas and demonstrate that, in cases of low-pressure crystallisation, it is important to identify the 0009-2541/02/$ -see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 9 -2 5 4 1 ( 0 1 ) 0 0 3 6 9 -2