Eyewitness memory: when later information enters the scene
Eyewitness memory can acquire details that were supplied after an event. In a 2019 experiment by Jessica LaPaglia and Jason Chan, volunteers watched a bank robbery video containing three warning shots. Some later encountered a description that reduced the number to two. On a final memory test, misleading details entered their recollections despite instructions to rely on the video. The experiment also measured confidence. In one condition, the effect persisted among responses participants rated with the highest levels of certainty.
The study's structure made the source of the error traceable. In its first experiment, 120 undergraduates watched a 25-minute excerpt from the television series Flashpoint. Half completed an immediate memory test containing no misleading information; the others played Tetris for the same interval. Later, all encountered a written description containing accurate contextual material and critical details that were either misleading or neutral. A neutral version referred to warning shots without giving a number. The researchers could therefore compare memory for details that had been contradicted with memory for details left unspecified.
That comparison measures how later misinformation affects reports of the original event. Both groups had viewed the same event, but only some details received a false replacement. Across conditions in the first experiment, participants recalled fewer correct details for misleading items than for neutral items. The average proportions were 39% and 52%. False information was also repeated much more often after a misleading description. The final test was assessing memory for the original video, so agreement with the later description counted as an error rather than successful learning of new material.
Earlier recall did not provide automatic protection. When the misleading material appeared within a fuller description of the event, participants who had taken the initial test repeated more misinformation than those who had not. Removing contextual material eliminated that additional disadvantage. In a third experiment, LaPaglia and Chan kept contextual information but changed its order. A coherent sequence produced the extra susceptibility following prior testing; a randomized sequence did not. These manipulations support the researchers' interpretation that processing the surrounding story helps determine how later misinformation affects recall.
The confidence findings require equal care. In the first experiment, the difference remained when analysis was restricted to responses rated four or five on a five-point confidence scale. The corresponding difference in the second experiment was not statistically significant for highly confident responses. That variation limits any sweeping claim about certainty. The study establishes that confident errors can occur under controlled conditions; it does not supply a universal error rate for witnesses who recognize a face, nor does it turn every confident identification into an unreliable one.
Repetition can also change certainty without an outside person supplying a false detail. In a separate 2022 study, Philip Gustafsson, Torun Lindholm and Fredrik Jönsson recruited 56 Swedish speakers to watch a staged assault. Participants were interviewed immediately and again two weeks later. Half were assigned four intervening exercises in which they wrote down their memories. The others had no such task. Each interview included confidence ratings for individual statements, allowing the researchers to distinguish confidence in correct details from confidence in mistakes.
Confidence generally predicted accuracy in that experiment, including at the later interview. Yet repetition increased confidence particularly for inaccurate recollections. Both findings belong in the assessment of eyewitness memory. Certainty carried information across the study's responses, while some errors became more confidently expressed. The researchers also emphasized limits: participants watched under laboratory conditions, knew an interview would follow and received no explicit misinformation. This was a study of rehearsal and retention, not a direct test of misleading police interviews or real-world lineups. Its confidence results cannot simply be transferred to every identification procedure.
LaPaglia and Chan also considered what their results meant for early interviews. They did not recommend waiting to question witnesses. Their discussion supported obtaining an early account because investigators cannot control every later opportunity for misinformation, including conversations between witnesses. The difficulty was understanding when the benefits of early recall might be accompanied by increased susceptibility to later suggestions. Their experiments varied the context of the misleading material rather than the seriousness of an actual crime. The findings therefore support a conditional account of eyewitness memory, with the effect of prior recall depending partly on how later information is encountered.
Public expectations have often been more absolute. Daniel Simons and Christopher Chabris reported a national telephone survey in 2011, using responses collected in 2009 and weighted for demographic representation. About 63% endorsed a video-camera model of memory, and about 48% endorsed the idea that a formed memory does not change. These figures measured beliefs, not the frequency of actual memory errors. They help explain why experimental changes in eyewitness memory can feel surprising: the experiments test a flexible recollective process, while many people expect a stable recording.
- Jessica A. LaPaglia and Jason C. K. Chan, Telling a good story: The effects of memory retrieval and context processing on eyewitness suggestibility, PLOS ONE, 2019 — journals.plos.org
- Philip U. Gustafsson, Torun Lindholm and Fredrik U. Jönsson, Eyewitness accuracy and retrieval effort: Effects of time and repetition, PLOS ONE, 2022 — journals.plos.org
- Daniel J. Simons and Christopher F. Chabris, What People Believe about How Memory Works: A Representative Survey of the U.S. Population, PLOS ONE, 2011 — journals.plos.org
- University of Washington, Human Memory and the Law lecture description identifying Elizabeth Loftus and her UC Irvine affiliation, February 13, 2013 — www.washington.edu