Recall the experiment of Englert, Scully, and Walther. With shutters closed it seems at first that the presence of the photon in either cavity is a fact, and that this fact (i) indicates the slit taken by the atom and (ii) implies that the atom went through a single slit. If, however, the shutters are opened, this purported fact is "erased," for we now have evidence that the atom did not go through a single slit. This means, of course, that the presence of the photon in either cavity was never more than a possibility: the possibility of detecting the photon in either cavity. No fact has been "erased" since the presence of the photon in either cavity never was a fact. A possibility can be erased. A fact cannot. A possibility can cease to be a possibility. A fact cannot cease to be a fact.

Resonator Triptych (detail) by Eric J. Heller.

I don't say that what was possible yesterday may not be possible today (which is trivial), or that what was factual yesterday (for instance, it rained) must be factual today (which is nonsense). As long as nothing indicates either the slit taken by the atom or the manner in which the atom went through both slits (in phase or out of phase), both possibilities exist: the possibility that the atom went through a single slit and the possibility that it went through both slits. Once something indicates the slit taken by the atom, the former possibility is a fact and the latter possibility no longer exists. And once the former possibility is a fact, it will remain a fact, for it will always be true that the atom went through a single slit.

The need to emphasize the impossibility of "erasing" a fact may puzzle anyone who is not aware of the difficulties that we face in making sense of the quantum theory. On the other hand, you will understand it if you are familiar with Schrödinger's infamous cat. If quantum states did evolve deterministically, the cat would evolve into a "superposition" in which it is both alive and dead, and the experimenters would evolve into a superposition in which they both find the cat alive and find the cat dead, and the experimenters' colleagues would evolve into a state in which they are told both that the cat is alive and that the case is dead, and so on ad absurdum. I find it hard to understand that someone who arrives at such a conclusion does not seek the error in his argument. (The last sentence is courtesy of N.G. van Kampen, “Ten theorems about quantum-mechanical measurements,” Physica A 153, 97–113, 1988).