Memory works like a recording device - once formed, memories are stable and can be recalled accurately.

When Hermann Ebbinghaus published his landmark study of memory in 1885, he established something that appeared to confirm the intuitive model: memories formed quickly and could be retrieved with remarkable precision if tested immediately, but faded in a predictable curve over time. The forgetting curve, as it became known, fit neatly with the dominant metaphor for memory, a recording that slowly degraded. Ebbinghaus worked with nonsense syllables rather than meaningful material, and his findings, influential as they were, concerned quantity of retention rather than the more troubling question of accuracy.

That question was taken up by British psychologist Frederic Bartlett, who published "Remembering: A Study in Experimental and Social Psychology" in 1932. Bartlett gave his subjects a Native American folktale called "The War of the Ghosts" and asked them to reproduce it from memory, first shortly after reading and then at increasing intervals. What he found was not degradation but transformation. Subjects did not simply lose details; they systematically changed them, rationalizing unfamiliar elements, substituting culturally familiar concepts, shortening and smoothing the narrative to make it more internally coherent. Memory was not playing back a recording. It was reconstructing a story.

Bartlett's work was largely forgotten in academic psychology for decades, supplanted by the behaviorist emphasis on learning and conditioning. In the postwar era, the recording metaphor remained dominant in popular understanding and in most textbook accounts of how memory worked. People spoke of photographic memory as a real and admirable thing. Courts treated eyewitness recollection as reliable access to stored experience.

What changed the scientific consensus was a sustained research program begun by Elizabeth Loftus in the early 1970s. In a landmark 1974 study with John Palmer, Loftus showed participants film footage of automobile accidents and asked them to estimate vehicle speeds. The precise verb used in the question, "contacted," "hit," "bumped," "collided," or "smashed", systematically altered the speed estimates, with "smashed" producing figures about nine miles per hour higher than "contacted." In a follow-up experiment, subjects asked with "smashed" were later more likely to report having seen broken glass in the footage, glass that had not been there. The leading question had not merely influenced an estimate; it had altered what subjects believed they had seen.

Loftus called this the misinformation effect. Over the next two decades, her laboratory and others documented it across dozens of experimental variations. Post-event information from other witnesses, from leading questions, from news coverage, or from simple suggestion consistently found its way into subjects' recall of original events. Loftus was able to implant entirely false memories, not just altered details but complete fictitious episodes, including childhood experiences of being lost in a shopping mall, using straightforward suggestion delivered by a trusted source. About one in four subjects in her false memory studies came to recall the planted event with apparent confidence and emotional specificity.

The mechanism behind the misinformation effect reflects a fundamental feature of memory encoding. Each time a memory is retrieved, it is briefly destabilized and then re-encoded, a process called reconsolidation. At each reconsolidation, new information present at the moment of recall can be integrated into the stored trace. Memory does not retrieve from a fixed record; it reassembles, and each reassembly is an opportunity for change.

This model ran directly against the intuition that forgetting and distortion were different kinds of failure. Forgetting was normal; distortion felt like something else, a trick, an anomaly. What the laboratory evidence accumulated by Loftus and her colleagues demonstrated was that distortion was not anomalous. It was the normal product of a memory system built for flexible, generative reconstruction rather than faithful reproduction.