In the past few years, the so-called mélange model has been offered as an alternative to the long-standing model of enrichment of the subarc mantle by separate additions of fluid and sediment components from the subducting plate. In the mélange model, components from the subducting plate become physically mixed at the slab-mantle interface. Partial melting of the peridotite subsequently occurs after being hybridized by the mélange material that diapirically rises into hotter portions of the wedge. Here, we present the first Ba isotope study of lavas from the Tonga-Kermadec arc (southwest Pacific Ocean) and show that Ba isotopes distinguish between fluid and melt derived from different subducted components. This pro-vides fresh constraints on the debate. Remarkable along-strike Ba isotope variations were observed and are best explained by contributions from variable proportions of sediment and altered oceanic crust (AOC) fluid from the subducting plate. Combined Ba-Sr-Pb isotope rela-tionships indicate that sediment melt and AOC fluid were added to the source of the arc lavas separately at different times. This is inconsistent with the mélange model, at least in this arc.