Halliday & Matthiessen (1999: 546):
We have shown that the motive for reconstruing experience in this way was in the first instance a textual one: in the grammars of these languages, when one is developing a reasoned chain of logical argument such that complex phenomena have to be given a clearly defined status in the organisation of information (the clause as "message"), such phenomena have to be constructed in nominal form. But there is no insulation between one part of the grammar and another, and this inevitably has ideational effects. Any semantic construct that appears as topical Theme has a function in transitivity; if it is formed as a nominal group, it is potentially a participant in some process, and therefore at some level it is an entity, a thing. If we say diamond is transformed into graphite, this is a process involving two things, diamond and graphite; if we reconstrue this as the transformation of diamond into graphite it has become one thing, transformation, with diamond and graphite serving only circumstantially to qualify it as a thing of a certain kind.