Like many scientists in the field of memory, I used to think that a memory is something stored in the brain and then accessed when used. Then, in 2000, a researcher in my lab, Karim Nader, did an experiment that convinced me, and many others, that our usual way of thinking was wrong. In a nutshell, what Karim showed was that each time a memory is used, it has to be restored as a new memory in order to be accessible later. The old memory is either not there or is inaccessible. In short, your memory about something is only as good as your last memory about it. This is why people who witness crimes testify about what they read in the paper rather than what they witnessed. Research on this topic, called reconsolidation, has become the basis of a possible treatment for post-traumatic stress disorder, drug addiction, and any other disorder that is based on learning.
That Karim's study changed my mind is clear from the fact that I told him, when he proposed to do the study, that it was a waste of time. I'm not swayed by arguments based on faith, can be moved by good logic, but am always swayed by a good experiment, even if it goes against my scientific beliefs. I might not give up on a scientific belief after one experiment, but when the evidence mounts over multiple studies, I change my mind.