The phenomenon of lexical repulsion in text
The advances in computing technology which made possible the study of electronic text in large quantities opened the door to the study of lexis, and crucially of significant word collocation. The type and strength of preference or ‘attraction’ which obtains between two or more words has been the raw material out of which we have long been extracting secondary language resources by means of automated systems (e.g. A. Renouf, 1996; M. Pacey et al, 1998). Now, we turn to our vast data repositories once again, this time to try to find evidence for the existence of another ‘force’ in textual organisation, one which operates contrary to word ‘attraction’, and which we refer to as ‘repulsion’. Like attraction, repulsion is a universal phenomenon and we have an expectation that it will be discovered to be alive and well in language. By ‘repulsion’, we mean the rare or non co-occurrence of two words as neighbours. Since this is the case for the majority of words in the vocabulary, we begin with a focus on synonymous word pairs which, by virtue of their shared meanings, can be expected to provide more surprising, focussed and easily interpretable results. In this paper, we tentatively interpret from our results some of the ways in which repulsion seems to operate in text and to what extent it is proving to be amenable to objective measurement.