ABSTRACT
In the early days of computational lexicology, some 40 years ago, it was felt important to distinguish between a word and a lexical item. This was a period for emphasizing the complexity of language, and proposing abstract categories of language form to escape from the confines of surface phenomena; procedural models, like Immediate Constituent Grammar (e.g. Stageberg 1966: 262 ff.) were held in suspicion because position and sequence were non-negotiable. On the other hand, Firth (1957b) made a distinction between ‘sequence’, the physical positioning of linguistic events relative to each other, and ‘order’, realized mainly by sequence, which was an abstract representation of language form.1
