A witness sits across from a detective, years after the incident. Her voice is steady, describing someone seen that night with photographic detail. Yet, confidence is not accuracy. Research has confirmed it directly: a witness's confidence at trial is not a reliable predictor of identification accuracy. That steady voice is not evidence of a reliable memory. It may be evidence of a very well-rehearsed one. Consider the forensic risk: using inadequate methods on fragile evidence. This applies to human testimony too. Mishandling witness accounts doesn't just lose information; it corrupts it. The key idea here is that memory is not a recording. Researchers have confirmed that eyewitness memory is reconstructive — not a literal, permanent capture of an event. Every time a memory is retrieved, it is rebuilt. And every rebuild carries the risk of revision. Think of memory retention like a curve that drops fast and then levels off. Human forgetting is not linear. Retention tends to decline rapidly at first, then more gradually over time. A modern replication of the classic forgetting curve found this pattern holds — with a possible stabilization point around the 24-hour mark. That means the first day after an event is the most critical window for capturing accurate recall. Miss it, and you are not just working with less information. You are working with information that has already begun to reorganize itself. Cold-case witness work is risky because memory can be altered by post-event information like questions or media reports, known as the misinformation effect. The classic paradigm has three stages: witnessing an event, receiving post-event information, and later being tested on the original event. In the Yolanda case, witnesses lived inside a community for decades. Local rumor, news coverage, and repeated police questioning all became potential sources of contamination. Repeated misinformation can increase a witness's confidence in the resulting memory. That is the False Familiarity Trap: a witness mistakes repeated exposure for genuine recollection. It's often assumed traumatic events are remembered accurately, but research shows misleading information can distort even these memories. For highly arousing negative events, the distorting effect of misinformation can actually persist over time. [emphasis] Emotional intensity does not immunize a memory against contamination. It may extend how long that contamination holds. This matters enormously for any witness who was present near a violent crime scene and then spent thirty years inside a community talking about it. What should investigators do? The Cognitive Interview, based on memory-retrieval principles, is key. It involves reconstructing circumstances, reporting without filtering, and recalling events in varied orders. The Reverse-Chronological Recall tactic — working backward from the most recent memory toward the original event — disrupts the narrative scripts a witness has rehearsed over decades. Research summarized by the National Institute of Justice found the Cognitive Interview produced more information than a standard police interview without increasing confabulated details. A revised version elicited approximately 45 percent more information than the original in one laboratory comparison. Here are key safeguards. First: blind interviewing. Researchers found that interviewers who were blind to case details elicited more correct information than informed interviewers — even after a one-week delay. An interviewer's knowledge of case facts shapes their questions, and shaped questions shape answers. Second: capture confidence verbatim and immediately. The National Research Council is explicit — confidence recorded at the time of an early identification is far more informative than confidence measured later at trial. Video-record the procedure. Do not conduct a first identification in a courtroom. Delay and intervening procedures make courtroom identification a less accurate memory test. Remember this: a 30-year delay does not make a witness useless. Researchers confirm that accurate fragments can survive decades even when the overall narrative has become incomplete. The operational rule is to separate independently recalled details from later explanations — rather than treating the account as wholly accurate or wholly false. Apply the Cognitive Retrieval Method: use reverse-chronological sequencing, blind the interviewer to case details, capture confidence immediately, and prewar the witness that post-event information may have shaped their memory. That prewarning alone can reduce susceptibility to misinformation on a later test. The takeaway is this: a hesitant witness who clearly separates memory from uncertainty is often more reliable than a confident one who cannot. Confidence is not a compass. Process is.