A woman witnesses a crime and recounts it to a policeman. Months later she appears in court to testify. As her story unfolds, it begins to differ from the notes taken by the policeman. A journalist covering the case notices that her testimony includes things she could not have known at the time but that were later discovered and that appeared in his newspaper. Though intensely grilled by the DA, she sticks by her story.

Why did her memory change? Why didn't she know the difference between what she experienced and what she read in the paper? The short answer is that remembering is a dangerous affair in the life of a memory. A slightly longer answer requires that we delve into the mechanisms that store memories.

Memory formation occurs in stages. Initially, a temporary or short-term memory is formed. This memory is fragile and will dissipate unless it is converted into a long-term memory through protein synthesis inside the neurons that processed the experience. The new proteins stabilize the synaptic connections that constitute memory at the cellular level. If protein synthesis is disrupted in the hours following the experience, a long-term memory does not result. The conversion of short-term into long-term memory via protein synthesis is called consolidation.

It has also been found that disruption of protein synthesis after the remembrance of a fully consolidated long-term memory produces a loss of the memory. This is taken to mean that when memories are retrieved they have to be reconsolidated via protein synthesis in order to persist.

Reconsolidation is essentially an updating process. After consolidation, a memory remains unchanged until it is retrieved. At that point, the brain has the opportunity to incorporate new information into the memory, things that have been learned since the memory was stored initially. I haven't thought about the Edge Annual Question since last year, but now that I have been forced to remember it, my memory of it includes the new question.

So far so good. But considerable research now suggests that reconsolidation can overwrite previous memories. That is, the old memory is eliminated and the new one involves a collage of old and new information. This integration process determines what we will remember the next time. When our witness read the newspaper account, the old memory was retrieved and new information was integrated with the old information. She was unable to tell the difference between what she experienced and what she later learned because it was now one memory. Laboratory studies in fact show that people are not very good at remembering what they actually experienced, and often make mistakes that involve the insertion of new information into a memory.

The bottom line of reconsolidation research is that your memory of some experience is only as good as your last recollection of the experience. Each use of a memory changes the memory. Obviously, the changes are not always so dramatic as what I have described. But the fact is that memory can, at least to some extent, be changed by experience, and sometimes the changes can be striking.

There a number of practical implications of this research. One is that it might be possible to relieve emotional stress by having people remember their stressful experiences and then interfering with reconsolidation. This is pretty much what happened to Jim Carey's character in The Eternal Sunshine of the Spotless Mind. But there is also evidence that it works in real life situations with trauma victims. Studies in rats also suggest that this same approach can be used to reduce the ability of drug-related cues to produce relapse.

Memory works pretty well most of the time. But we should be careful as a society when we make significant decisions on the basis of one person's memory. The only way a memory remains "pure" and resistant to change is by never being used. The most accurate memories are indeed the ones never remembered. Be careful about what you remember.