Chapter 32 — Key Takeaways

A scannable one-page card. For the full argument and the worked examples, see index.md.

The core claims

• The most persuasive evidence is among the least reliable. A confident witness pointing across the courtroom convinces juries like nothing else — and mistaken eyewitness identification was present in a large majority of the DNA exonerations, more than any single discredited forensic method. Confidence persuades; it does not, by itself, indicate accuracy.
• Memory is reconstruction, not recording. We do not store and replay footage; we store fragments and rebuild the event each time, filling gaps with inference, expectation, and post-event information. The rebuild feels seamless and certain from the inside, so the witness cannot tell which parts are original. Two named consequences: the misinformation effect (post-event information rewrites the memory) and unconscious transference (a familiar face misattributed to the crime).
• Retrieval is not read-only. Each recall re-encodes the memory, so it can change — growing more vivid and confident at each telling. By trial, the jury sees the latest reconstruction, not the original perception, and the witness experiences all of it as one continuous, unaltered memory.
• Estimator variables cap reliability and cannot be fixed — only assessed. Lighting, distance, exposure time, stress/fear, weapon focus, the cross-race effect, retention interval, and disguise set a ceiling on any identification at the moment of the event. They degrade accuracy while leaving confidence largely untouched — sometimes even raising it. That mismatch is the engine of the whole problem.
• System variables are controllable — and done badly, they manufacture false certainty. How the lineup is built (fair fillers, one suspect), what the witness is told ("may or may not be present; you need not choose"), who administers it (blind), whether confidence is recorded immediately and verbatim, and whether feedback is given. The post-identification feedback effect ("good, that's our guy") inflates a tentative choice into trial-day certainty.
• Sequential vs. simultaneous matters less than the other safeguards. Simultaneous invites a "pick-the-closest" relative judgment; sequential pushes toward an absolute memory comparison. But the research is genuinely contested (sequential may also reduce correct identifications), and a fair, blind, well-instructed simultaneous lineup beats an unfair, non-blind sequential one. Don't let the debate distract from the undisputed safeguards.

The method-validity verdict (NAS 2009 / PCAST 2016 reasoning)

Where it sits: well below single-source DNA (which has earned the strong, exclusionary direction), and alongside the confident pattern-comparison disciplines — for the same structural reason: confident individualization asserted without a quantified basis. Human face memory is real and useful; the courtroom claim built on it outruns what the faculty can deliver.

What you can honestly say on the stand

• About the witnessing conditions (an expert): "The identification was made under conditions known to degrade accuracy — a brief, nighttime view, under high stress, across racial lines. I cannot say this particular witness is wrong; I can say these conditions are associated with elevated error rates, and the jury should weigh the identification accordingly." (An eyewitness expert testifies to the science, not to a verdict on a specific witness.)
• About confidence: "The witness's confidence at trial carries little information about accuracy, because confidence is inflated by feedback and repeated retrieval after the identification. The number that matters is the witness's confidence at the first identification, recorded verbatim before any feedback — if that record exists."
• About a clean first identification: "From a double-blind, fairly composed array, the witness's immediate statement was [verbatim words and certainty]. That is the measurement with evidentiary value — bounded by how good the witnessing conditions were."
• What you must NOT say (or accept): that a certain witness is therefore a correct one; that an in-court identification "proves" identity; that confidence is a reliable guide to accuracy; or that "the witness tried hard to remember" makes the memory reliable.

Key terms (one line each)

• Eyewitness identification — a witness's report that a particular person is the one they saw; persuasive to juries but, because memory is reconstructive, among the least reliable evidence and a leading documented cause of wrongful conviction.
• Lineup — a formal procedure in which a witness views the suspect with known-innocent fillers (live or a photo array); distinct from a single-suspect showup.
• Sequential vs. simultaneous — simultaneous shows all members at once (relative "pick-the-closest" judgment); sequential shows them one at a time with a forced yes/no (absolute memory comparison); order matters less than the other safeguards.
• Double-blind administration — neither the witness nor the administrator knows which member is the suspect, so the administrator cannot steer the witness or give confirming feedback.
• Estimator vs. system variables — estimator = witnessing conditions the system can only estimate and cannot change; system = the controllable choices about how the identification is collected.
• Confidence–accuracy relationship — the (weak, conditional) link between certainty and correctness: meaningful only for a pristine first-identification confidence statement, and largely worthless once inflated by feedback and retrieval by the time of trial.

The cold-case line

The neighbor's confident "tall stranger's truck" near the cabin at night, from a quarter-mile away, is an eyewitness lead discounted — logged, not erased, but down-weighted on witnessing-conditions grounds (night, distance, brevity, hindsight). It includes and excludes no one. We let the physical and digital evidence — collected under controllable, auditable conditions — carry the weight. Chasing it would repeat the Cotton/Thompson error in miniature.

The themes this chapter advanced

• Exclusion over proof — a confident identification can only include; it was a DNA exclusion that overturned Cotton's and Anderson's convictions. Eyewitness evidence constrains; it does not prove.
• The validity spectrum — the core claim ("this person, to the exclusion of others") sits low, with the confident pattern-comparison disciplines and well below DNA, because the confidence has no measured basis.
• Cognitive bias is the chief threat — double-blind lineup administration is the Chapter 31 blind-testing reform applied to a witness's memory; the steering happens unconsciously, so good intentions are not a safeguard.
• The CSI effect cuts both ways — juries over-trust a confident in-court identification (and convict), which is exactly the certainty the science says is least reliable.