A comparison of the crystal contents of selected sieve grades from 9 ignimbrites, with the yield of crystals from crushed pumice, shows that on average a tenfold concentration of crystals has taken place in the ignimbrites. This is explained by the selective loss of pumice by an amount equal to at least the present bulk of the samples studied. The pumice is thought to have been lost above the vent or the moving pyroclastic flow, though there is a possibility, yet untested, that it may be contained in the distil part of the ignimbrites. This paper also further characterises the relatively well-sorted basal layer, and the pipes (interpreted as due to primary fumaroles) which cut some ignimbrites. Basal layer and pipes show an even greater concentration of crystals than the body of the ignimbrites. One corollary of this study is that doubt is placed on the value of making chemical analyses of ignimbrites, the composition of which is likely to depart appreciably from that of the eruptive magma.