The exception proves the rule.
This seemed perfectly reasonable to us, especially since it confirmed Vin's suspicions. But then we got a second opinion from my personal linguist Ben Bergen:English has the noun 'proof', the verb 'to prove', and another verb 'to proof'. These all arise from a common Germanic form: compare for instance German 'prüfen' (to test), Norwegian 'å prøve' (to test), Danish 'at prøve' (to test), Swedish 'at pröve' (to test). There is also an Icelandic form (which I can't remember off the top of my head), and an Anglo-Saxon form (which I also can't remember right now). But anyway, all of these forms mean 'to test', as opposed to 'to have tested and have the proof ready to hand'. However, a weird thing happened with the English forms of the Germanic stem. The three principal forms split themselves between two different meanings. Two assumed a meaning of 'this has been proven, and I'm now showing you the proof, buddy', as in ... 'to prove': to show that something is correct, on the basis of proof that you have 'proof': to have information that shows that something is correct ... but one English form retained the meaning of 'to test, to undertake a test', rather than to have already tested and have the proof at hand, as in ... 'to proof': to examine and test (usually a piece of text) for mistakes, falsehoods, etc. It is only in this one form, 'to proof' that the real Germanic meaning survives. It is also with this form that the phrase 'The exception proves the rule should be read'. It really means that the exception is testing whether the rule is valid or not. A re-phrasing should really give 'The exception tests the rule', or even 'The exception proofs the rule', if you expand the meaning of 'to proof' beyond the correction of texts. Germanic is such fun ...
English 'prove' is a borrowing from Latin through Old French and Anglo-Norman. The American Heritage Dictionary gives the following etymology (Webster gives essentially the same): [Middle English prove, preve, from Anglo-Norman prove, and from Old French prueve, both from Late Latin proba, from Latin probre, to prove; see prove.] The expression 'the exception proves the rule' appears to be borrowed directly from Latin. Interestingly, it seems NOT to be that prove originally had the meaning 'test', despite the fact that 'proof' does mean 'test' in expressions like 'proof reader', 'the proof is in the pudding', and the like. The Dictionary of Modern English Usage cites the original legal sense of 'the exception proves the rule' as follows: "'Special leave is given for men to be out of barracks tonight till 11.0 p.m.'; 'The exception proves the rule' means that this special leave implies a rule requiring men, except when an exception is made, to be in earlier. The value of this in interpreting statutes is plain." The phrase seems to date from the 17th century. Anthony Cree, in _Cree's Dictionary of Latin Quotations_ (Newbury, 1978) says that the phrase comes from classical Latin (Latin spoken before A.D. 400). In full, the Latin seems to be 'exceptio probat regulam in casibus non exceptis.' ('The exception proves the rule in cases not excepted.') Or in a shorter version: 'exception probat regulam'. As for the Germanic cognates, I'm not a Germanicist, so I can't speak to them.
So that's the story so far. I'm not sure the matter has been conclusively settled (informed input welcome!).
More importantly, the following has been at the top of my whiteboard since September 2000:
NC is the exception to the rule that the exception proves the rule that proves it.
Nancy is the exception that proves that the exception proves the rule.