To gain a clear view of the possible range of logical metonymy data,
it is necessary to look to corpora of spoken and written text. This
investigation will give a sense of how widespread the use of logical
metonymy is, and how far qualia structure can go towards predicting
the interpretations that the examples are given in context. I have
therefore consulted both the Lancaster-Oslo/Bergen (LOB) Corpus and
the British National Corpus (BNC) for relevant data.
These are corpora of British
English, the LOB incorporating 500 written text samples of about 2,000
words each, the BNC consisting of over 100 million words of a wide
variety of written text (90%) and transcripts of spoken language
(10%). I focused mainly on logical metonymies with begin,
but also looked at those for finish.