Research into clusters and bundles has a very brief history since the technology to pick them out of the raw data of language, of the morass of the linguistic trace, is a recent invention. Such research is nevertheless an excellent illustration of how the technological tools available to the researcher shape, dictate even, the object of study, as the radio-telescope in astronomy, the particle accelerator in physics and, of course, the concordancer in linguistics (Partington 1998: 144). Clusters are at the first level of abstraction from this morass – just like collocations with which, in fact, they have a great deal in common. Bundles lie at a further level of abstraction and seem often to have a grammatical-oid feel about them - indeed Biber at al. point out some grammatical correlates in their descriptions. Whether or not they constitute “missing links” on the chain or cline from the linguistic morass to the abstraction we call grammar and whatever their possible status is as units of meaning, it seems to us undeniable that they tell us a great deal about how speakers go about the construction of discourse. We might even go further and suggest that the revelations clusters and bundles supply about language structure seem to be confirming the ideas of Sinclair (1991, 1992, 1996 and elsewhere), Stubbs (1996, 2001), Hunston and Francis (2000) among others on the nature of grammar.