False Memory: Why You Remember a Word That Was Never There

Remembering is a reconstruction, not a playback. A memory is assembled at the moment of recall from fragments of what was encoded, the general gist of the episode, and whatever else is active at the time, including what was said afterwards. When something that was never there fits the gist well, it gets built into the reconstruction and arrives feeling exactly like the parts that are accurate.

The word list above shows the mechanism in its simplest form. Every word on it is an associate of one word that was never shown, so by the end of the list that word has been activated many times without ever appearing. At recall there is no record of where the activation came from, and the word is produced along with the rest. Roediger and McDermott described this as associative activation during study combined with a failure of what Johnson and Raye called reality monitoring: whether a word was heard or merely thought of is a question the memory system does not always answer.

What makes these errors important is how they feel. In the 1995 experiments the never-presented word was recognized as old at almost the same rate as words that had actually been studied, and when participants were asked to say whether they remembered the moment of hearing it or merely knew it had been on the list, they chose remembered most of the time. Confidence, detail and the sense of reliving are produced by the reconstruction, not by the accuracy of it, which is why inspecting the experience is not a reliable way to tell a false memory from a true one.

Everyday false memories have more routes in than the word list does. Misleading questions supply details that get absorbed. Retelling a story settles a version of it. Looking at photographs, hearing a sibling's account, or imagining an event in detail all add material that later cannot be sorted by source. Elizabeth Loftus and Jacqueline Pickrell showed in 1995 that a suggested childhood event that a family member said had not happened could be reported as a memory, with details added that nobody had supplied.

One limit deserves stating plainly. Ost and colleagues found in 2013 that how susceptible a person is to the word-list illusion has essentially no relationship to how susceptible they are to misleading suggestion, with correlations averaging about zero across a range of measures. False memory names a family of errors with different mechanisms rather than a single trait, so the demo above says nothing about how your memory handles anything else.

The word list method is older than its name. James Deese published a study in the Journal of Experimental Psychology in 1959 asking why some lists produce particular intrusions in immediate recall. He built 36 lists of twelve words, each list made of the primary associates of one word that was left out, and found that the probability of the missing word appearing as an intrusion was predicted by how strongly the list words brought it to mind, a correlation of .87 between the two measures.

Henry Roediger and Kathleen McDermott turned that observation into a method in 1995 with "Creating false memories: Remembering words not presented in lists." In their first experiment, 36 students at Rice University heard six of Deese's strongest twelve-word lists and recalled each immediately. They recalled 65 percent of the words that had been presented, and produced the missing word after 40 percent of the lists, while intrusions of any other word occurred after only 14 percent. On a recognition test afterwards, the never-presented words were called old 84 percent of the time against a hit rate of 86 percent for studied words, and participants were sure the word had been on the list in 58 percent of cases.

The second experiment extended the method to 24 lists of fifteen words, the length used in the demo on this page, with 30 students hearing sixteen of them. False recall rose to 55 percent of lists, which was slightly higher than recall of the words actually presented in the middle positions, 47 percent. On the recognition test participants also said whether they remembered the moment of presentation or simply knew the word had occurred, and for falsely recalled words the answer was remembered in most cases. The paradigm is now named after Deese, Roediger and McDermott, and it is used because it manufactures a vivid false memory in a minute with no deception and no distress.

The rows below are illustrations of where reconstruction adds material, not findings from the studies cited on this page.

Trying harder does not reliably make a memory more accurate, so the counters are about evidence and about how questions get asked.

• Write it down at the time. A contemporaneous note beats a confident recollection, and it is the only part of this list that works after the fact.
• Do not ask leading questions, and notice when one has been asked of you. "How fast was the car going when it smashed into the other one" supplies a detail that a later account may include.
• Keep accounts separate before comparing them. Once two people have talked, their memories stop being independent evidence.
• Do not let confidence stand in for accuracy. Confidence alone cannot establish that a memory is accurate: in the 1995 experiments people were sure about words that were never shown, and said they remembered hearing them.
• Be careful with imagining and rehearsing. Picturing an event in detail, repeatedly, makes it easier to recall and harder to place, the opposite of what you want for anything disputed.
• For anything consequential, build a record rather than a recollection: dates, documents, photographs, logs. This is a practice, not a judgment about anyone's honesty.

False memory is a broad label, and the research distinguishes several ways a memory can go wrong. This page is educational: it cannot establish whether any particular memory is accurate, and research on word lists settles nothing about disputed accounts of real events.

• It is not the same thing as the misinformation effect, which is one specific route to a false memory: misleading information supplied after an event. The word-list illusion on this page arises with no misleading input, and Ost and colleagues found in 2013 that susceptibility to the two is essentially unrelated, so neither stands in for the other.
• It is not source confusion, where the content is real but its origin is misattributed, as when something you read is remembered as something you were told. Source confusion is one of the mechanisms that produce false memories rather than a synonym for them.
• It is not confabulation, which in clinical use names the fluent production of false accounts seen in certain neurological conditions. The laboratory effect on this page happens to healthy people in ordinary conditions and is not a sign of anything.