On using spoken data in corpus lexicography

Corpora are increasingly used in lexicography in order to provide good evidence for dictionary statements: the inclusion of spoken data in corpora is generally considered important. This paper raises some issues connected with the use of spoken data. It points out that the extensive differences between written and spoken language have great consequences for dictionary-making. It argues that the repercussions have not yet been fully thought through, and suggests that new models for the lexicographical description of spoken language may have to be developed.