8. Phrasal similes in the BNC
This study focuses on similes, a category of phraseological units that has been amply studied in the phraseology literature but suffers froma lack of corpus-based description. The concept of simile is defined and the differences between similes and comparisons, on the one hand, and similes and metaphors, on the other, are described in some detail with reference to the literature. Then four simile patterns are extracted from the British National Corpus: <i>as </i>Adj/Adv <i>as</i>, <i>is like a(n), is like </i>V-<i>ing a(n) </i>and V <i>like a(n). </i>The selection procedure consists in sifting out literal comparisons using the reversibility test and the degree of figurativeness. This process highlights interesting differences between the patterns in terms of frequency, lexicalization and register. The <i>as </i>Adj/Adv <i>as-</i>pattern is the most frequent but the vast majority of the instances represent literal comparisons. The few per cent similes tend to be lexicalized and there is little innovation. By contrast, the <i>is like a(n)</i>-sequences are much less common but regularly figurative and innovative. The V <i>like a(n)-</i>pattern also proves to contain a high proportion of similes, both lexicalized and innovative.