Sequence and order

Corpus linguists often attempt to avoid assumptions imported from pre-corpus studies, by using methods which could be called “inductive”, in so far as they proceed from observations about textual sequences to generalizations about order in the system. However, induction has been questioned for over 400 years (by Bacon, Hume, Popper and others), and the possibility of rigorous, theory-free induction is now generally rejected. One major phraseological model, proposed by Sinclair in the late 1990s, is certainly not a purely inductive generalization from raw corpus data. I will discuss this model using attested data on a particular construction and a distinction proposed by Firth, Halliday and Palmer between “sequence” (an observable feature of texts) and “order” (a feature of linguists’ models).