Can You Remember Something That Never Happened? Yes, and Vividly

Memory does not store and replay a recording. Retrieving a memory reconstructs it from fragments, from general knowledge about how such situations go, and from whatever the current context supplies, and the reconstruction is then stored again, which means each act of remembering can alter what is remembered. That architecture is efficient and is why memory works at all for ordinary purposes, since exact storage of everything would be unusable. It also means information encountered after an event can be incorporated into the memory of it without any sense that something has been added, which is the mechanism behind nearly every demonstration in this area and is not a malfunction so much as the normal operation seen from an awkward angle.

Several experimental methods produce false memories reliably:

• •Misleading questions after an event, which alter what witnesses later report seeing
• •Presenting a list of related words, after which people confidently recall a related word that was never shown
• •Suggesting a plausible childhood event to participants, a proportion of whom come to remember it
• •Doctored photographs, which produce detailed recollections of events they depict
• •Repeated imagining of an event, which increases confidence that it happened
• •Confidence and detail rise with repetition, so a false memory becomes more convincing over time

Elizabeth Loftus and colleagues published work from the 1970s onward showing that the wording of a question changes what witnesses report, with people asked how fast cars were going when they smashed into each other giving higher estimates and being more likely to report broken glass that was not there than those asked about cars that hit each other. Later work presented adults with accounts of childhood events supplied by family members, including one fabricated event about being lost in a shopping centre, and a substantial minority came to report remembering it, in some cases with added detail nobody had suggested. The technique has been replicated many times with different events, and the proportion who develop such memories varies with how plausible and how well supported the suggestion is.

The finding is frequently overstated in both directions and the limits are worth being precise about. Demonstrating that a memory can be false does not establish that any particular memory is, and using the research to dismiss a specific account is exactly the error it warns against. Most memories of most events are broadly accurate for the central details, and the distortions concentrate in peripheral information, in details supplied after the fact and in the exact wording of what was said. Traumatic events are not generally forgotten wholesale, and the claim that they routinely are is not well supported. Confidence expressed immediately after an event, before any feedback or discussion, correlates with accuracy considerably better than confidence expressed later, which is a practical finding with direct procedural consequences.

The practical consequences have been severe. A wave of prosecutions in several countries during the 1980s and 1990s rested on recovered memories of abuse elicited through suggestive therapeutic techniques including repeated imagining, hypnosis and pressure to produce an account, and a substantial number of those convictions have since been overturned, with several people having served long sentences. Eyewitness identification, which is among the most persuasive evidence a jury hears, has been shown to be unreliable in ways that correlate poorly with the witness's confidence, and a large proportion of convictions later overturned by genetic evidence involved mistaken identification. Reforms have followed in how interviews are conducted, how identification procedures are run and how juries are directed, and they arrived after the damage.

Remembering reconstructs an event from fragments and general knowledge and then stores it again, so later information can be incorporated without any sense that something was added. Misleading questions, related word lists and suggested childhood events all produce false memories reliably in experiments. Prosecutions resting on memories elicited by suggestive techniques produced convictions that were later overturned.