Searching for extended units of meaning—and what to do when you find them

Two of the key outcomes of corpus-linguistic research over the past 30 years have been the development of the idea that meanings are mostly constructed through context (undermining traditional notions of the individual word as an autonomous bearer of meaning); and the discovery that recurrence and regularity—our tendency to employ a limited number of conventionalized ways of expressing ideas—are essential features of the language system. Both findings have had a major impact on our understanding of how language works, and both have influenced the content of dictionary entries—contributing, for example, to improved word sense disambiguation, and to a greater emphasis on phraseology and collocation. However, there is still much to do. Ever-larger corpora and more powerful corpus-query tools reveal areas where we can further improve our description of languages, and thus provide better resources for users. In addition, the migration of dictionaries to digital media (removing space constraints) opens up new opportunities for doing this. In a characteristically far-sighted paper (Sinclair, Textus 9(1): 75–106, 1996 ), John Sinclair broadened the search for what he called “units of meaning” by investigating longer strings of words and identifying recurrent, and often quite extended, patterns of usage. Using this as a starting point, I will look at other examples in corpus data of the kinds of patterning Sinclair discussed, and we will see how current corpus-querying systems can help us identify these extended units of meaning. Finally, I will speculate about whether dictionaries should aim to describe these longer units, and if so, how this might work in practice.