Well, we do know that memories aren't really localized - there's no neuron in your brain that means the color blue. And based on what we know about how neurons communicate with one another, the field of neural networks does shed some light on the basic constraints of neuroanatomy. I'm not needlessly speculating here; from what we do know about how the brain works, things like sensations and simple memories are localized in groups of neurons, but one individual neuron might play a part in the neural action of dozens of different such memories. When you take neurons out of a neural system, you don't remove individual memories - you diminish the system's ability to recognize stimuli in general. That's why I say it seems less likely that this technique is removing memories and is more likely overwriting or reassociating them; it sounds to me that the experiment perhaps leaves the mice with a more recent, tangible memory of hearing the tone for the first time and not getting shocked and that was what caused them to not fear the tone. (Because they had drug-induced amnesia, the tone would seem new even though it was not.) In any case, my real point was that it's hard to say what happened to individual memories, because all we are really observing is behavior.