Luckily I was thinking of this this morning . I would suggest that you annotate your corpus for metaphorical expression. I used an approach of identifying metaphor in text in one of my papers to be uploaded soon.
That already tells me you subscribe to some variant of Lexical Metaphor (i.e. that metaphors are initiated by lexical items). This isn't necessarily the case though. Metaphors about the same concept can be initiated by different lexical items (eg the concept of 'pain' can have metaphors using such words as 'hurt', 'disturb', 'annoy', 'inflammatory', etc) or even non-lexical linguistic entities (eg. constructions, phrases). Consider if this is something you would like to include as well.
If you are going to annotate, come up with a scheme and stick with it, no matter what. It will be tempting to update the scheme as you go, that won't work out well. If attainable, get multiple annotators to work with your scheme and aim for a decent inter rater reliability.