There have always been some anomalies that have sat uncomfortably with this assumption - for instance, memories don't stay in the hippocampus; it seems that after some days or weeks they are transferred to the cerebral cortex. But the community of memory researchers has mostly tried to ignore such inconvenient data.
Until recently, that is. Four years ago, Karim Nader and his colleagues in New York showed that if an animal was taught a particular task, and then days later was reminded of it by being put in the same context, the memory became labile once more - that means it could be disrupted by protein synthesis inhibitors. It was as if the reminder not only reactivated the old memory, but resulted in an entirely new memory being formed on top of it. Of course, we can intuitively recognise this; when we recall a past event, we are not recalling the event per se, but our memory of it from the last time we recalled it. This is why our autobiographical memories are being reshaped as we go through life.