On the sociolinguistic relevance of routine formulae