The Ekati property, Northwest Territories, Canada, hosts in excess of 150 volcanic kimberlite pipes occupied by a wide variety of rock types including coherent, magmatic material and a range of fragmental, volcaniclastic deposits of both pyroclastic and resedimented origin. Geochemical analysis of a suite of samples from several of these bodies provides valuable insight into the nature of their components and the processes by which they form. Observed variations in the bulk composition of coherent kimberlites correlate with their mode of emplacement and petrographic characteristics. High-volume bodies of coherent kimberlite, occurring within and in several cases dominating volcanic pipes (pipe-fill CK), are depleted in CO2 and key incompatible elements (e.g. Ti and Nb) but enriched in Ni and SiO2 relative to samples from narrow kimberlite dyke intrusions. The composition and certain textural features of pipe-fill CK can be interpreted to reflect formation by pyroclastic processes that involved a loss of volatiles and fines, and a concentration of olivine crystals relative to intrusive magmatic kimberlite. If this is the case, then the apparent coherent character of these rocks suggests that they represent coalescence of hot, fluidal pyroclasts (i.e. spatter) generated by fire-fountain style eruptions. Samples of massive pyroclastic kimberlite (PK) are geochernically similar to pipe-fill CK but generally show a greater degree of depletion of incompatible elements, suggestive of more explosive eruption processes and a greater degree of physical fractionation. The geochemical data support petrographic observations that Ekati PK contains very little xenolithic material. Resedimented volcaniclastic kimberlite (RVK) shows variable degrees of enrichment in Al2O3 relative to PK. The extent of Al-enrichment correlates with the proportion of dark, fine-grained elastic matrix material in the RVK samples and their composition can be explained as a mixture of a very olivine-rich, fines-depleted, juvenile component with varying amounts of argillaceous sediment likely derived from surface deposits present at the time of kimberlite eruption. However, the increased Al2O3 content in RVK is also positively correlated with Nb concentration indicating that the matrix consistently includes a small amount of kimberlitic ash. Geochemical trends in massive volcaniclastic kimberlite (MVK) and fine-grained volcaniclastic kimberlite (VK) from the Fox pipe are very different to those of other VK samples from Ekati and appear to reflect mixing of juvenile kimberlitic material with granodiorite. The data suggest very little if any incorporation of exotic fine-grained sediments into these rocks. Geochemical modelling suggests that the eruption(s) forming Fox MVK involved a significant loss of fines, similar to that interpreted for PK in other Ekati kimberlites. However, the large quantity of fragmented granodiorite (similar to 30 to 45%) within Fox MVK indicates that, in contrast to other localities, the Fox eruption was unable to completely clear its vent and disperse the excavated granodiorite prior to infilling the pipe. The composition of Fox VK provides no evidence for a loss of fines and suggests that these were either retained during eruption or reintroduced by mixing with ash deposits from previous eruptions. (c) 2008 Elsevier B.V. All rights reserved.