One of the most heated and prolonged debates in chemistry occurred over the so-called non-classical 2-norbornyl cation. Very specifically, during reactions, exo-2-norbornyl derivatives solvolysed about 60 times faster than the 2-endo ones. The endo derivatives behaved more or less as you might expect if the mechanism was SN2, but the exo ones behaved as if they were SN1, but there was an additional surprise: the nucleophile was about as likely to end up on C6 as C2. There were two explanations for this. Winstein suggested the presence of a non-classical ion, specifically the electrons in the C1-C6 bond partly migrated to form a "half-bond" between C2 and C6. Thus was born the "non-classical carbonium ion". On the other hand, Brown produced a sequence of papers arguing that there was no need for such an entity, and the issue could be adequately explained by more classical structures, and as often as not, by the use of proper reference materials.
 
That last comment refers to a problem that bedevils a lot of physical organic chemistry. You measure a rate of reaction, and decide it is faster than expected. The problem is, what was expected? This can sometimes border on being "a matter of opinion" because the structure you are working with is somewhat different from standard reference points. This problem is even worse than you might consider. I reviewed some of the data in my ebook Elements of Theory 1, and suggested that the most compelling evidence in favour of Brown's argument was that the changing of substitution at C1 made very little difference to the rates of solvolysis at C-2, from which Brown concluded there was no major change of electron density at C1, which there should be if the C1-C6 bond became a half-bond as Winstein's structure required. As it happened, Olah also produced evidence that falsified Brown's picture, and as I remarked at the end, each falsified the other, so something was missing.
 
In the latest edition of Science (vol 341, p 62- 64) Scholz et al. have produced an Xray structure of the 2-norbornyl cation, which was made from aluminium tribromide reacting with the exo 2-norbornyl bromide, and what we find is equal C1-C6 and C2-C6 distances, as required by the non-classical ion. Also, these are long, at about 180 pm. Case proved, right? Well, not necessarily. The first oddity is that the C1-C2 distance is 139 pm, or about the same length as benzene bonds. Which gets back to Brown's "falsification" of the non-classical ion: while the C1-C6 bond is dramatically weakened, the C1-C2 bond is strengthened, and the electron density about C6 may be not that much changed, despite the fact that the bond Brown thought he was testing was half-broken. Nobody picked that at the time.
 
What do I mean by, "not necessarily"? It is reasonably obvious this is not the classical structure that Brown perceived. That is correct, but there are two other considerations. The first one is that to get a structure, the structure must be in an energy well, which means it does not actually represent the activated state. To give an example, the cyclopropylcarbinyl system would presumably give, as an ion, Cyc-CH2+ would it not? The trouble is, the system rearranges and is consistent with that, as well as a cyclobutyl cation, and an allylcarbinyl cation. The actual cation is probably something intermediate. So the rate acceleration then may not be caused by the intermediate cation, but by whatever is happening on the reaction path. If this cation was the cause of the rate acceleration, it should also operate on the endo cation. Yes, the mechanism is different, but why? A product available to both cannot be the reason. There has to be something that drives the exo derivative to form the cation. My explanation for that is actually the same as that that drives the cyclopropylcarbinyl cation.
 
The second consideration is the structure itself: the two bonds to C6 are equal, and C1-C2 is remarkably short. There is one further way this could arise. Let us suppose we follow Winstein and break the C1-C6 bond. What Winstein, and just about everybody else, thought is that we replace that with two half σ bonds, but suppose no such σ bonds are formed? Instead, rehybridize C1 and C6 so we have two p orbitals. With two p orbitals and a carbenium centre we have the essence of the cyclopropenium cation, without two of the frame-work σ bonds. That gives us a reason why the cation is so stable: under this interpretation, it is actually aromatic, even if two of the bonds are only fractional π bonds.
 
Is that right? If it is, then there is a similar reason why ethylene forms edge-complexes with certain cations. Of course, it may not be correct, but as a hypothesis it seems to me to have value because it suggests further work.