In the Linguistics lesson on formulaic phrases, I argued that non-canonical variants on idiomatic proverbs could be more frequent than the canonical forms. As an example of this, we all know the proverb 'An apple a day keeps the doctor away'. Entering "a day keeps the * away" into COCA produces 25 results. Of these in 17 the wildcard is filled by 'doctor' (other wildcard fillers include 'gerontologist', 'craving' and 'banana'!!). In the 17 'doctor' returns, in only 9 is the phrase preceded by 'apple' (alternatives include 'a clove', 'a slice of AppleCream Cheese Bundt Cake' and 'an orgasm'). In total then, only 9 of 25 (36%) of occurrences of the proverb are in full canonical form.
If this is representative of how people use idiomatic proverbs (and there are reasons to think that the number of non-canonical variants is underestimated given that these are probably common in informal conversation which is underrepresented in the corpus), a key question is: how do people know what the canonical form of the proverb is?
A second question is: do people perceive proverbs as potential gap-fill structures rather than fixed phrases?