Remembering what never happened

Most of us treat memory as a recording. We assume that an event is stored somewhere in more or less the state it arrived in, and that remembering means finding the file and playing it back. The feeling of remembering supports this: a vivid memory feels like watching rather than like guessing. Unfortunately the feeling is not a reliable guide to anything.

What laboratory work suggests is that a memory is rebuilt each time it is used, out of whatever material is available. Some of that material is the original event. Some of it is general knowledge about how such events usually go, and some of it is what has been said about the event since. The rebuilt version then replaces the old one, which is why a story told many times can drift a long way from the thing it started as, without the teller ever noticing a change.

The most unsettling demonstrations involve memories of events that did not occur at all. In a well-known kind of experiment, a researcher describes to an adult a childhood incident supplied by a relative, mixed in with real ones, and asks about it over several sessions. A substantial minority of people come to report the invented incident as their own, and they do not report it flatly. They add detail, weather, dialogue, and a feeling of what it was like, none of which came from anywhere except themselves.

This has consequences outside the laboratory, and the clearest is in how witnesses are questioned. A question that supplies information plants it. Asking how fast a car was going when it smashed into the other one produces higher estimates than asking the same question with the word hit, and it also makes witnesses more likely to recall broken glass that was never there. Police forces in several countries have changed their interview procedures for exactly this reason, and the change is not about honesty. The witnesses in these studies are trying to tell the truth.

None of this means memory is worthless, and the opposite conclusion is drawn too often. The gist of what happened survives rather well; it is the particulars that move. A reasonable working rule is to trust the shape of a memory and to distrust its details, especially the vivid ones, and especially those that have been rehearsed with other people who were there.