Dylan Thurston <dpt@math.harvard.edu> points out my mistake:

...

Strictness alas matters. Here's the witness:

class Num a => ZeroList a where consZero :: a -> [a] consZero _ = 0:xs

Err, "Num a" is already a bad context by Simon's criterion because of "fromInteger", which is what ultimately causes the problem in this case.

Quite right. I overlooked the fact that the superclass constraint "Num a" implicitly includes all the methods of Num in ZeroList for these purposes. Olaf Chitil <olaf@cs.york.ac.uk> gives an example along the lines I was attempting (which I've corrected slightly):

instance Show MyType where shows _ = ("element of my type" ++ )

The important thing here is that the non-strict shows method can make up any result it likes. I do think some sort of extended defaulting mechanism as suggested by Thomas Hallgren would be useful. Yes, we will occasionally choose confusing interpretations of e.g. "" vs []. On the other hand, I'm sure we can warn when defaulting in this way. But its not the magic bullet against the monomorphism restriction one might have hoped for... -Jan-Willem Maessen